VIA - Zhaoxin x86-Prozessoren [Archiv] - Seite 2

Archiv verlassen und diese Seite im Standarddesign anzeigen : VIA - Zhaoxin x86-Prozessoren

Seiten : 1 [2] 3 4

Loeschzwerg

2018-12-12, 07:00:35

Eine Zusammenfassung von Expreview zu "China"-GPUs:
https://www.expreview.com/65091.html

Google Übersetzung inkl. Bilder im Spoiler:

Super Classroom (160): Unknown domestic GPU, true domestic game card can be expected?

In the first half of the year, ZTE’s sanctions have caused everyone’s concern about processor chips. Not long ago, Fujian Jinhua Group was included in the list of entities, which involved DRAM memory chips. Processors and memory chips have also been strongly supported by the domestic development over the past few years. The semiconductor industry, because these two chips are imported from China, the storage chip alone imported more than 88.6 billion US dollars last year. In contrast, the output value of GPU chips is much smaller. The combined performance of AMD and NVIDIA's high-performance GPUs is about $10 billion, and even half of the domestic imports are only $5 billion.

Although the output value of the GPU market is small, but the threshold is extremely high, there are two AMD and NVIDIA manufacturers that can develop high-performance GPUs. No matter the technology, talents or patents, it is a serious obstacle to companies other than AN. . As for the domestic, the GPU is even worse. If the domestic CPU can still see the backs of leading vendors such as Intel and AMD, then the GPU may be uncertain where others lead.

Today's super-classroom will talk about domestic GPUs. Of course, the GPUs mentioned in this article mainly refer to high-performance general-purpose GPUs. The GPUs of mobile processors have many ready-made licenses available, which is not a problem. High-performance general-purpose GPUs are domestic. The embarrassment of development.

Domestic GPU Jingjia microjingmei series

In the domestic GPU camp, Jingjiawei should be the biggest one. The reason why this super-classroom is a big factor is because Jingjiawei announced the successful report of the new generation GPU JM7200. Micro has been described in various reports as a leading GPU manufacturer and the only manufacturer to independently develop GPUs. Although it is boastful, Jingjiawei is very representative on domestic GPUs, and they mainly face military and government. Enterprises and other markets.

According to the official website, Changsha Jingjia Microelectronics Co., Ltd. was established in April 2006. It has Beijing Maxwell Technology Co., Ltd., Changsha Jingmei IC Design Co., Ltd. and Shijiazhuang Branch. The company is committed to providing high-reliability, high-quality solutions, products and ancillary services to customers in the areas of information detection, information processing and information delivery. At present, it is the only company in China that successfully develops and manufactures domestically produced graphics processing chips (GPUs). In March 2016, Jingjiawei was listed on the Shenzhen Stock Exchange with stock code: 300474.

Jingjia Micro is a military-civilian integration company. Its products include GPU chips, military radars, etc. Among them, Jingmei series GPU chips are currently their core business, accounting for more than 70% of revenue and revenue. In 2014, the company developed Successfully developed the JM5400 GPU chip, which is claimed to be the first high-performance GPU chip with independent intellectual property rights in China. It can be widely used in graphics generation and display with high reliability requirements to meet the graphics system in airborne, shipboard and vehicle environment. The function and performance requirements fully replace the foreign chips such as M9, M54, M72, M96 and IMX6.

65005

Jingjiawei's graphics processor is mainly Jingmei JM series. The current main force is JM5400. Previously, the JM7200 was reported. The process technology was upgraded from 65nm to 28nm. Before JM7200, there is a JM7000, which is also 28nm process. There is no mention of the difference between the JM7000 and the JM7200. It is probably considered that the JM7200 is an advanced version of the JM7000. The architecture and specifications of the two should be similar.

65006

The introduction of JM7000 series has been deleted by Jingjiawei official website

Surprisingly, after being reported by the media on a large scale, Jingjiawei officially deleted the introduction of the JM7000 series GPU, including the above specification, and replaced it with JM7200, but the official website did not provide specific information about the JM7200.

So what level of GPU can Jingjiawei achieve? According to the official website and network information, we made the following specifications comparison, and compared the specifications of the weakest GT 640 graphics card with NVIDIA Kepler architecture, the results are as follows:

65007

From this comparison, the performance of Jingjiawei JM5400 is almost worth mentioning, and the specifications are weak. However, the significance of this product is not in performance. It is mainly for military equipment. In this field, performance is almost no problem. Whether it is self-controllable or not is the key. It can be said that the JM5400 solves the problem.

On the JM7200 with 28nm process, its rendering performance and floating-point performance can finally be seen. The specific data of the memory bandwidth can not be determined, but the official indicates that the data transmission capability is increased by several times. This should mean that the interface is upgraded from PCI to PCIe 2.0. X16, the only one that can increase dozens of times is it, and the maximum memory bandwidth is up to 128bit.

Compared with the GT 640, the rendering and floating point performance of the Jingjia Micro JM7200 is still worse, but considering the difference between the power consumption of 10W vs 50W, the performance of the JM7200 should be surprising, and its energy efficiency ratio is calculated. It can be much higher than the Kepler architecture, and then consider the official sudden support for the JM7000 series GPU specifications, which may be hidden.

In general, the performance of the JM5400 to JM7200 series GPUs has been greatly improved, but the overall technical level is still far away, still not as good as the low-end GPUs six or seven years ago, if you look at 3D technology support, the gap is even bigger. It is. However, because of the special industry demand, the JM5400/JM7200 series has extremely high requirements on temperature and reliability. This is not comparable to consumer GPUs.

Jingjia Micro's new generation GPU: 8,000 million, can fight GTX 1080 graphics card

The JM5400 and JM7200 series are at least 6-10 years behind AMD/NVIDIA. After solving the problem of high-performance GPUs, Jingjiawei is also developing a new generation of GPUs. The previous roadmap shows a new generation of high-performance general-purpose GPUs. The code names are JM9200, JM9400 and JM9800, but the official application last year began to raise funds of 1.3 billion yuan to develop a new generation of GPUs, MCUs and other chips, including high-performance GPU project research and development investment of 880 million yuan, codenamed JM9231 and JM9271, in response to the CSRC documents You can find the specific specifications of these two GPU projects:

65008

Compared with the JM5300/JM7200 series, the performance and specifications of Jingjiawei JM9231/9271 are increased by dozens of times. They will use the unified rendering architecture. The JM9231 supports PCIe 3.0 x16, the frequency is no less than 1.5GHz, and the pixel rendering capability is increased to 32GP/. s, the memory capacity has reached 8GB, the bandwidth is 256GB/s, and it will probably use GDDR5 memory.

As for the JM9271 GPU, Jingjiawei said that it will use the JM9231 architecture. "By increasing the number of computing units, increasing the memory bandwidth, bus and output interface speed, the scientific computing power has been greatly improved, and it can reach the international similar companies in 2017. The performance of high-end products is mainly for high-end applications such as artificial intelligence, security monitoring, speech recognition, deep learning, cloud computing, etc., which have high computational speed requirements. Based on JM9231, the scientific computing capabilities are greatly improved and improved. Develop computational libraries and high-performance computing platforms in the field of artificial intelligence to meet different application needs of customers."

According to the official specifications, the performance of the JM9231 does not lose the GTX 1080 graphics card, it can really reach the level of high-end graphics cards at the end of 2017, and it is likely to use HBM memory, because they have HBM-related technology in the procurement funds listed. And equipment costs.

Jingjiawei has plans for high-performance general-purpose GPUs on the GPU roadmap, and the prospects for this plan are very ambitious. The specifications of JM9231/9271 are different from the existing JM5400/7200. This technology spans the developers. The talent challenge is quite large, and the R&D investment is only 880 million yuan. The full budget is 150 million US dollars. When the NVIIDA releases the Pascal graphics card, it means that the research and development costs are as high as several billion dollars. Although Lao Huang’s statement is exaggerated, it is probably The cost of the past few years has been counted together, but the development of high-performance GPUs does burn money, and it is normal for GTX 1080 to spend $1 billion on this level of GPU.

In general, the domestic GPU roadmap given by Jing Jiawei looks exciting, but the GTX 1080 is not what I said. It is written in their application materials, but this plan is so big that people Can't believe that Jingjiawei did not make a clear statement on when to develop successfully. Considering that they have been postponed for nearly two years on the 28nm JM7200 GPU, it will not go smoothly in the JM9231/9271 GPU development process. There are still more follow-up questions.

Domestic GPU among the 716 ship JARI G12

In addition to the well-known Jingjiawei, there are still many domestic GPU units. Among them, there are two research institutes under the CSIC Group that have developed different GPUs - 716 developed in Lianyungang City, Jiangsu Province. JARI G12 GPU, the CSIC 709 in Wuhan, developed the Lingji GP101. Compared with the Jingjiawei listed company, it is necessary to disclose information. The two institutes under CSIC rarely disclose confidential information, especially they themselves. Strong military features, so the information on these two GPUs is very small, we first talk about JARI G12.

65009

According to a report released by CSIC, "JARI G12 is the most powerful domestic general-purpose graphics processor. The processor uses a hybrid rendering architecture that takes into account data bandwidth and rendering latency requirements, greatly enhancing the chip's Flexibility and adaptability; provide PCIe 3.0 bus, support x86 processor and domestic processor such as Godson, Feiteng, Shenwei; support 4 digital channels and 1 VGA output, provide general display such as DP, eDP, HDMI, DVI, etc. Interface, single channel digital channel maximum output resolution is 3840×2160@60fps, support extended, copy display and “extended+copy” display mode; built-in video codec hard core, support 2 channel 3840×2160 resolution video encoding , decoding function; support OpenGL 4.5 and OpenGL ES 3.0, meet high-performance 3D acceleration and VR display requirements; support OpenCL 2.0 to meet the needs of parallel computing and cloud computing; integrated tensor acceleration computing hard core, support AI computing acceleration. The GPU supports mainstream operating systems such as Windows, Linux, and VxWorks, and supports domestically-controlled and controllable operating systems such as the winning kylin, JARI-Works, and Tao. Environment system. "

Based on this information, JARI G12 can be judged to support PCIe 3.0, support DP, HDMI, DVI output, support 2 channels of 4K video decoding, encoding, graphics API support to OpenGL 4.5, and also support OpenCL 2.0 acceleration, but no DX support information, but The ARI G12 also integrates a tensor unit that supports AI operations.

https://www.forum-3dcenter.org/vbulletin/attachment.php?attachmentid=65010&thumb=1&d=1544594500 (https://www.forum-3dcenter.org/vbulletin/attachment.php?attachmentid=65010&d=1544594500)

The official also announced the JARI G12 architecture diagram, previously reported in the peer report that its performance is comparable to the GTX 1050 Ti, the reason is that the officially announced unit is huge, got the GTX 1050 Ti level, but this guess is unreasonable, this The architecture diagram does not really see the specific unit size, and it is even more difficult to know the performance level of the unit.

The 716's characterization of the JARI G12 is a generation of the same industry in the country. It has achieved a comprehensive transcendence in terms of functional performance and received trial and demand intentions from CLP Group, AVIC Group, National Defense Science and Technology University and other units.

Judging from the published indicators, the JARI G12's specifications and technology are indeed more advanced than Jingjiawei's JM7200 series, but so far, no ARI G12 GPU chip is designed, streamed or listed. Process technology and specific performance data are also unknown.

Domestic GPU, ship heavy industry 709, Lingjiu GP101

The 709 located in Wuhan is the first national key engineering research institute integrating computing technology, information technology, automatic control and mathematical theory research. It is mainly engaged in information command system, computer reinforcement technology, fault tolerance technology, parallel processing technology and network technology. Professional research and development work in software engineering, graphic image processing, mechatronics, and printed circuits.

On the GPU side, they developed the Lingji GP101, but this GPU is more mysterious on the Internet, not even a picture. In addition to a report by Xinhua News Agency, there is very little information on this GPU. According to the relevant person in charge of CSIC 709, Lingji GP101 graphics processor chip supports HDMI, DVI, VGA and other general display interfaces, supports 2D, 3D graphics acceleration and OpenGL ES2.0, and supports 4K resolution. At the same time, support for video decoding and hardware layer processing and other functions, can be widely used in military and civilian dual-use electronic equipment, industrial control, electronic information and other fields.

According to the official notification information, the specifications of the Lingjiu GP101 are not advanced, and the API supports OpenGL ES 2.0, which is an API specification for mobile processors. However, this GPU should also be used for military products, so performance is not a problem.

In addition to Lingjiu GP101, JARI G12 and Jingjiawei JM series, there are some domestic GPU sporadic reports, Shanghai Zhaoxin also has Elite 1000/2000 series GPUs, but Megacore GPU technology is also from VIA VIA, which is accurate VIA's S3 company, and these GPUs are mainly for low-power devices, more integrated in the SoC, unlike the other three GPUs.

Summary: Domestic GPU can only solve the problem now, consumer level is not expected

Compared with the CPU processor in the country, and achieved good results, the domestic GPU and the international level - or American companies, because in addition to the United States company in other countries on the high-performance GPU can not play. The domestic GPU technology is generally behind AMD/NIVIDA for at least 10 years. Now it can only solve the problem, and the performance and 3D technology are still early.

For a long time, domestic GPUs may be limited to the military, industrial and government enterprises. In the consumer market, there is no hope of the rise of domestic GPUs, at least for now. The new generation of GPUs developed by Jingjiawei is very powerful and the technology is advanced enough, but when mass production is still a problem, they may not be able to do this themselves.

Loeschzwerg

2018-12-12, 07:02:06

--Anhang--
65010

Tralalak

2018-12-14, 01:03:25

http://www.zhaoxin.com/Upload/201812131726290964.jpg
source: www.zhaoxin.com (http://www.zhaoxin.com/purchase.html)

gravitationsfeld

2018-12-14, 04:48:58

OpenGL 1.5? Soll das ein Witz sein?

Die groesseren Chips sehen vielversprechender aus. Mal sehen wie da die Performance aussieht.

Loeschzwerg

2018-12-14, 06:37:18

@Tralalak: There is a new "ZX" driver (v10.17.19.7 (http://download.windowsupdate.com/d/msdownload/update/driver/drvs/2018/12/82d94358-1e05-45eb-8cf6-2d93f09c4e53_54ed35c7b26d87bfd3c40cc5ec964a47e49a84b2.cab)) in the MS Update Catalog, featuring the ZX C960.

"ZX C960 GPU" = ZXIGP, PCI\VEN_1d17&DEV_3A04&SUBSYS_3A041d17
"ZX C860 GPU" = ZXIGP, PCI\VEN_1d17&DEV_3A03&SUBSYS_3A031d17
"ZX C860 GPU" = ZXIGP, PCI\VEN_1106&DEV_3A03&SUBSYS_3A031106
"ZX C860 GPU" = ZXIGP, PCI\VEN_1106&DEV_3A01&SUBSYS_3A011106

I'm going to to see if this one is still working with VX11/Chrome 645/640.

----

OpenGL 1.5? Soll das ein Witz sein?

Reicht vielleicht im militärischen Sektor :D

JM9231/9271 sind aktuell noch in der Entwicklung, dauert auch sicher noch seine Zeit und dann wird man diese in öffentlicher Hand vermutlich eh nie sehen.

Ich fand es prinzipiell ganz interessant zu lesen was da alles für Entwicklungen vorangetrieben werden.

Tralalak

2018-12-14, 11:31:25

@ Loeschzwerg:

"Support 4K ultra high definition video decoding Zhaoxin GPU performance is affirmed

Since the opening of the KX-5000 series processor, the Zhaoxin domestic x86 processor has adopted SOC design and integrated proprietary GPU. The SMIC C-860 and C-960 are integrated into the KX-5000 and KX-6000 series processors respectively. They support DX11.1 and Open GL 3.2 specifications and support H.264 and H.265 format video codecs. In terms of video output, the C-860 supports HDMI1.4b and VGA standards, while the C-960 is upgraded to HDMI2.0, DP1.2a and VGA standards.

With excellent GPU performance, Lenovo's open K6-5000 series processor's open-air M6200 desktop can simultaneously achieve 4K ultra-high definition video decoding and multi-channel output, the display effect is outstanding, and the CPU usage is extremely low, and got Many viewers are sure."

The KX-6000 series processor adopts 16nm FFC technology.
source: http://www.zhaoxin.com/InCenterContent.aspx?id=210

http://www.zhaoxin.com/app_upload/images/2(27).jpg
source: http://www.zhaoxin.com/InCenterContent.aspx?id=211

About TSMC 16/12nm Technology

"TSMC also introduced a more cost-effective 16nm FinFET Compact Technology (16FFC),which entered production in the second quarter of 2016. This process maximizes die cost scaling by simultaneously incorporating optical shrink and process simplification. Furthermore, 12nm FinFET Compact Technology (12FFC) drives gate density to the maximum, for which entered production in the second quarter of 2017. "

source: http://www.tsmc.com/english/dedicatedFoundry/technology/16nm.htm

Loeschzwerg

2018-12-14, 19:18:17

I'm going to to see if this one is still working with VX11/Chrome 645/640.

Nope, can't get the drivers to work. Swapping the KMD doesn't help.

I hope VIA shows up with some new products sooner or later.

gravitationsfeld

2018-12-14, 20:57:16

JM9231/9271 sind aktuell noch in der Entwicklung, dauert auch sicher noch seine Zeit und dann wird man diese in öffentlicher Hand vermutlich eh nie sehen.
Was ist denn sonst der Sinn?

Loeschzwerg

2018-12-14, 21:34:40

Hauptfokus des Unternehmens liegt wie gesagt im staatlichen und militärischen Bereich. Abwarten ob da was für normale Konsumenten abfällt.

gravitationsfeld

2018-12-15, 00:44:46

Hm, strategisch macht das vielleicht Sinn, die USA koennte irgendwann ein Emargo verhaengen. Kommt mir trotzdem komisch vor nur fuer's Militaer.

Benutzername

2018-12-20, 18:13:14

Hm, strategisch macht das vielleicht Sinn, die USA koennte irgendwann ein Emargo verhaengen. Kommt mir trotzdem komisch vor nur fuer's Militaer.

Warum? Das Militär umfasst in China auch eine Menge an wissenschaftlichen instituten, Entwicklungsbüros, Fabriken etc. Die chinesische Volksbefreiungsarmee ist eigentlich ein riesiger Mischkonzern mit bewaffneter Abteilung.

zB fast alle Bibeln auf chinesisch werden von der Vorschriftendruckerei der Armee gedruckt. (die drucken auch andere Sachen als Auftrag, wenn Zeit ist).

Tralalak

2019-01-10, 21:56:36

8-core part of most likely 7nm KX-7000 have 16MB L3 cache

2:07:20 - 2:07:26
Z3noq_LzE0w
source: https://www.youtube.com/channel/UCHDr4RtxwA1KqKGwxgdK4Vg

_____________

The next generation of the 7nm KX-7000 will reduce this limit even further. Most likely, they will use their SMT, FMA, NN execution and other patents.

source: https://patents.google.com/patent/US20170364361A1
source: https://patents.google.com/patent/US20170097824A1/
source: https://patents.google.com/patent/US20170103300A1

BlackArchon

2019-01-10, 22:38:05

I still can't buy these things? No? Who would have thought...

Loeschzwerg

2019-01-11, 06:52:07

It's possible to order a Lenovo Thinkstation with KX-U5580. China only.

@Tralalak: Thanks for the YT link :)

Loeschzwerg

2019-01-13, 19:10:16

Geekbench 4 results of a KX-6840 8-core @ 3.0GHz (Ubuntu 18.04):
http://browser.geekbench.com/v4/cpu/11577482

Single: 1523
Multi: 7250

Tralalak

2019-01-30, 13:13:41

Lenovo M6200 with KX-U5580 1.8GHz OctaCore review
Source: https://www.chiphell.com/thread-1951970-1-1.html

Tralalak

2019-02-13, 15:44:26

VirtualBox 6.0.4 x86 virtualization software
29 Jan 2019
Version 6.0.4:
Virtualization core:
Support Shanghai/Zhaoxin CPUs.
source: https://www.macupdate.com/app/mac/24801/virtualbox#

Tralalak

2019-03-03, 11:32:49

Acer VN1000-AM
VIA Nano X3 L4500 1.4GHz + VN1000
source: https://www.bilibili.com/video/av44873710

Tralalak

2019-03-06, 18:26:48

New KX-6000 Series model
ZHAOXIN KaiXian KX-U6780A@2.7GHz (8C, 2.7GHz, 2 x 4MB L2)
source: http://ranker.sisoftware.net/show_run.php?q=c2ffcdfcdabbdae7dfefd8ecddfb89b484a2c7a29faf89fac7ff&l=en

P.S. Maxon Releases Cinebench R20 Benchmark

Tralalak

2019-03-07, 14:26:34

VIA QuadCore C4650 2.0GHz

CPU Multi Score: 284 cb.
CPU Single Score: 80 cb.
MP Ratio: 3.56 x

https://i.imgur.com/34M8Pdc.jpg

https://i.imgur.com/TH4kb91.jpg

https://i.imgur.com/aO3fkpy.jpg

Tralalak

2019-04-02, 19:09:20

Rumours: Zhaoxin KaiSheng KH-30000 series up to 16-core processor

http://imgsrc.baidu.com/forum/w%3D580/sign=e2e0a7004da98226b8c12b2fba82b97a/63619844d688d43f8eb0b487731ed21b0ef43b08.jpg

source: http://dq.tieba.com/p/6088224810

EDIT: 03.02.2019 / 13:20

The ZX-E series wins the KH-30000 series processor with a new micro-architecture, integrates 8 CPU cores on a single chip, provides 8MB cache, clock speed up to 3.0GHz, built-in DDR4 memory controller, SATA3, PCI-e 3.0 Controller, support ECC memory, support CPU hardware-assisted virtualization technology, SM3 and SM4 national secret algorithm. What's more worth mentioning is that the ZX-E KH-30000 series processor also supports dual-channel interconnection, which brings a greater improvement in performance including virtualization.

source: https://translate.google.com/translate?hl=sk&sl=zh-CN&tl=en&u=http%3A%2F%2Fwww.zhaoxin.com%2FInCenterContent.aspx%3Fid%3D218

konkretor

2019-04-03, 12:07:51

very nice

Tralalak

2019-04-03, 12:52:59

Zhaoxin completed the development of next-generation 7nm verification chip in 2019
source: http://www.chinanews.com/business/2019/04-03/8799500.shtml

8-core part of most likely 7nm KX-7000 have 16MB L3 cache

2:07:20 - 2:07:26
https://youtu.be/Z3noq_LzE0w
source: https://www.youtube.com/channel/UCHDr4RtxwA1KqKGwxgdK4Vg

_____________

The next generation of the 7nm KX-7000 will reduce this limit even further. Most likely, they will use their SMT, FMA, NN execution and other patents.

source: https://patents.google.com/patent/US20170364361A1
source: https://patents.google.com/patent/US20170097824A1/
source: https://patents.google.com/patent/US20170103300A1

BlackArchon

2019-04-03, 20:49:04

Are there any physical products to buy? No? Still only paper launches? Oh well... :frown:

Tralalak

2019-04-05, 14:25:05

KX-5000 (ZX~D, Wudaokou) vs KX-6000 (ZX-E, Lujiazui)
ZHAOXIN KaiXian KX-U5580 @ 1.80 GHz (8C/8T, 8MB L2 cache) vs. ZHAOXIN KaiXian KX-U6780A @ 2.70 GHz (8C/8T, 8MB L2 cache)

Single-Core Score: 1 239 (46,81%) vs. 1 819 (+46,81%)
Multi-Core Score: 5 553 (62,48%) vs. 8 888 (+60,06%)

source: https://browser.geekbench.com/v4/cpu/compare/9408079?baseline=12670754

C3000 (Denverton, Goldmont) vs. KX-6000 (ZX-E, Lujiazui)
Intel Atom C3758 @ 2.20 GHz (8C/8T, 16MB L2 cache) vs. ZHAOXIN KaiXian KX-U6780A @ 2.70 GHz (8C/8T, 8MB L2 cache)

Single-Core Score: 1 532 (84,22%) vs. 1 819 (+18,73%)
Multi-Core Score: 8 157 (91,78%) vs. 8 888 (+8,96%)

source: https://browser.geekbench.com/v4/cpu/compare/6597237?baseline=12670754

Benutzername

2019-04-18, 14:15:13

Are there any physical products to buy? No? Still only paper launches? Oh well... :frown:

Yes, how long can they sustain paper launches? What's the point of developing all these chips and not manufacturing them?

Loeschzwerg

2019-04-18, 18:21:20

Diese Fragen wurden euch doch längst beantwortet... für die restliche Welt gibt es Produkte wenn VIA meint dass es mal wieder an der Zeit wäre das Embedded-Lineup zu verjüngen.

Hier kannst eine Lenovo Kiste mit KX-U5580 kaufen:
https://vip.lenovo.com.cn/Campaign/ProductDetails/ProductDetails/2574

Produkte mit KX-6000 sind nur eine Frage der Zeit, dürften mit ziemlicher Gewissheit aber ebenso auf diesen Markt beschränkt sein.

konkretor

2019-05-24, 11:01:28

AMD sold special Ryzen in China

https://www.golem.de/news/thatic-sugon-verkauft-workstation-mit-chinesischer-ryzen-cpu-1905-141481.html

MSABK

2019-05-24, 11:25:04

AMD sold special Ryzen in China

https://www.golem.de/news/thatic-sugon-verkauft-workstation-mit-chinesischer-ryzen-cpu-1905-141481.html

Hmm schade, es ist leider nicht bekannt in welchem Verfahren der gefertigt wird.

konkretor

2019-05-24, 16:41:19

14 or 12

Tralalak

2019-05-24, 17:34:24

CPU-Z 1.89

● Zhaoxin KX-5000 and KX-6000 processors families support

http://img1.mydrivers.com/img/20190524/s_7c5c8956e39e4c629c7d01db9049bec7.png

source: https://www.cpuid.com/softwares/cpu-z.html
source: http://news.mydrivers.com/1/628/628482.htm

Screemer

2019-05-24, 17:54:11

AMD sold special Ryzen in China

https://www.golem.de/news/thatic-sugon-verkauft-workstation-mit-chinesischer-ryzen-cpu-1905-141481.html
Na schau an gibt's also doch nen ryzen von tatic. Sollte huawai so den Ban umgehen können, dann wäre das für amd natürlich top.

Edit: sorry ist OT.

Tralalak

2019-05-24, 23:45:21

HWiNFO v6.06

- Added support of Zhaoxin KaiXian CPUs

source: https://www.hwinfo.com/version-history/

Tralalak

2019-05-27, 16:53:58

Zhaoxin KX-6840 @ 2.7GHz benchmarked

• Geekbench 4.3.1 (Single-Core Score : 1 776 / Multi-Core Score: 8 558)
• SuperPi mod 1.8 WP (1M): 31,843 sec
• Ludashi 5.15. : 51 176

source: http://tieba.baidu.com/p/6144517155

Tralalak

2019-05-27, 23:52:34

Zhaoxin wächst weiter mit ökologischen und technologischen x86-Vorteilen.

http://editerupload.eepw.com.cn/201905/1558697019405136.png

source: https://translate.google.com/translate?hl=sk&sl=zh-CN&tl=de&u=http%3A%2F%2Fwww.eepw.com.cn%2Farticle%2F201905%2F400853.htm

MSABK

2019-05-28, 01:06:39

Na schau an gibt's also doch nen ryzen von tatic. Sollte huawai so den Ban umgehen können, dann wäre das für amd natürlich top.

Edit: sorry ist OT.

Kann die USA nicht einfach die x86 Lizenz verweigern?

Die VIA cpu wäre doch eine Alternative für Huawei.

Benutzername

2019-05-28, 06:44:01

Kann die USA nicht einfach die x86 Lizenz verweigern?

Im Prinzip ja, aber für den chinesischen Markt wäre das dann ja erstmal egal. Wird dann schwierig neuere Erweiterungen von x86 zu bekommen. Aber diese ganze trade war im moment ist sowieso Trumps Verhandlungstaktik.

Die VIA cpu wäre doch eine Alternative für Huawei.

Äh, werden die Kerne nicht immernohc oder zumindest teilweise in den USA entwickelt? Sind doch die Reste von Cyrix und Centaur iirc. Und irgendeine Lizenz für AMD64 werden die doch auch brauchen?

Loeschzwerg

2019-05-28, 07:03:54

Centaur ist ein eigenständiges Tochterunternehmen von VIA, hier würde ich zunächst kein Problem sehen. Die bisherige Lizenz für x86 lief inkl. Zusatzfrist bis 2016, wie da der aktuelle Stand ist, keine Ahnung.

Kriton

2019-05-28, 16:17:08

Warum nicht die X86-Lizenz aus dem chinesischen Joint Venture von AMD? Da hat man dann auch bessere Technik als bei VIA.

Tralalak

2019-05-28, 19:04:51

* In 2003, VIA settled its long-time patent and monopolisation disputes against Intel in the UK in exchange for an extensive cross-licence agreement with Intel for 10 years.
source: http://ec.europa.eu/competition/sectors/ICT/intel_provisional_decision.pdf

* In addition, the FTC settlement order will require Intel to:

• modify its intellectual property agreements with AMD, Nvidia, and VIA so that those companies have more freedom to consider mergers or joint ventures with other companies, without the threat of being sued by Intel for patent infringement;
• offer to extend VIA’s x86 licensing agreement for five years beyond the current agreement, which expires in 2013;
source: https://www.ftc.gov/news-events/press-releases/2010/08/ftc-settles-charges-anticompetitive-conduct-against-intel

_____________
About Shanghai Zhaoxin Semiconductor Co., Ltd. - 上海兆芯集成电路有限公司 ("Zhaoxin")

Zhaoxin (Shanghai Zhaoxin Semiconductor Co., Ltd., also goes by VIA Alliance Semiconductor Co., Ltd.) was established in April 2013 with a total registered capital of USD$250M. As a joint venture between Shanghai Alliance Investment Ltd. 80.1% ($200M) who is affiliated to Shanghai SASAC and VIA Technologies, Inc. 19.9% ($49.75M),

Zhaoxin has access to Centaur Technology's design team along with VIA's patent portfolio and intellectual property, including access to all of VIA's Centaur Technology x86 microarchitectures.

Zhaoxin has about 1000 employees and locates its headquarter at Zhangjiang of Shanghai with branches in Beijing, Hangzhou, Wuhan, Shenzhen, Taiwan, California and Texas of America.

With the forefront technologies and know-how in the design of CPU, GPU and chipsets, Zhaoxin is well known to provide high security, high performance, low power dissipation, and low cost SoC solutions.

_____________
Intel Buys Assets of VIA Telecom
source: https://news.softpedia.com/news/intel-buys-assets-of-via-telecom-493761.shtml

VIA Telecom Completes Sale of Part of Assets to Intel Corporation
October 1, 2015 Press Releases
Taipei, Taiwan, October 1, 2015 Taipei – VIA Technologies, Inc. today announced that VIA Telecom has completed the sale of part of its assets to Intel® Corporation. The final details of the transaction will not be disclosed.
source: https://www.viatech.com/en/2015/10/via-telecom-completes-sale-of-part-of-assets-to-intel-corporation/

P.S. rumours said... it appears that under unpublished conditions, the x86 license continues.

Loeschzwerg

2019-05-29, 07:10:06

The KX-6000 is manufactured on the most advanced 16nm process. It integrates CPU, GPU graphics core, chipset and up to eight CPU cores on a single chip. The maximum frequency is 3.0GHz, and it supports dual-channel DDR4 memory with a maximum capacity of 64GB. And has obtained PCIe 3.0 official certification, USB 3.1 Gen 1 official certification. It will be put into mass production in September this year.
http://www.zhaoxin.com/InCenterContent.aspx?id=232

Tralalak

2019-05-30, 14:57:59

AIDA64 Extreme (beta)
Version: 6.00.5107 beta (May 29, 2019)

Release notes:
● identification of Zhaoxin KaiSheng KH-20000 Series (aka WuDaoKou)
● identification of Zhaoxin KaiSheng KH-30000 Series (aka LuJiaZui)
● identification of Zhaoxin KaiSheng ZX-C+ (aka Zhangjiang)
● identification of Zhaoxin KaiXian KX-5000 Series (aka WuDaoKou)
● identification of Zhaoxin KaiXian KX-6000 Series (aka LuJiaZui)
● identification of Zhaoxin KaiXian ZX-C (aka Zhangjiang)
● identification of Zhaoxin KaiXian ZX-C+ (aka Zhangjiang)
● chipset information for Zhaoxin ZX-100, ZX-200
● chipset information for Zhaoxin ZX~D IMC

source: https://www.aida64.com/downloads/NzNhYWU2MDk=

konkretor

2019-06-03, 12:55:00

https://www.anandtech.com/show/14456/spotted-at-computex-let-bygones-be-bygons-with-a-sugon-hygon

More China CPU with x86

Tralalak

2019-06-16, 18:55:10

SisoftSandra Ranker
ZHAOXIN KaiXian KX-U6780A 2.7GHz

Processor Arithmetic: 71.77GOPS
Processor Multi-Media: 116.47Mpix/s

source: https://ranker.sisoftware.co.uk/show_device.php?q=c9a598c28acb84dc95dbfbb0d1b8e089e886a6edb598cdfbccf4c485a3c4f9 d4e5c3b18cbd9bf2cfffd9b18cbc9ae2dfefc9acc9f4c4e291ac9c&l=en

Edit:

Memory Bandwidth: 18.04GB/s

source: https://ranker.sisoftware.co.uk/show_run.php?q=c2ffcdffd9b8d9e4dde5d2ead8e0c6b489b99ffa9fa292b4c7fac2&l=en

Tralalak

2019-06-19, 13:54:19

ZHAOXIN introduces the new 16nm KaiXian KX-6000 and KaisHeng KH-30000 Series of x86/x86-64 processors

https://n.sinaimg.cn/sinacn20190619s/199/w600h399/20190619/0a28-hyrtarw3279396.jpg

source:
https://news.sina.cn/2019-06-19/detail-ihytcerk7993238.d.html?from=wap

Tralalak

2019-06-20, 14:12:19

Zhaoxin KX-6000 vs KX-5000 vs ZX-C+ FC Series processors

http://5b0988e595225.cdn.sohucs.com/images/20190620/3723439738ef47febd1434fdf7ae598d.jpeg

source: http://www.sohu.com/a/321758990_195713

KX-6880 3.0GHz (16nm, 8C/8T, 8MB L2 cache) vs KX-5680 2.0GHz (28nm, 8C/8T, 8MB L2 cache)

SPEC2006 INT: 29.2 vs 19.9 (+ 46,7%)
SPEC2006 INT RATE: 170 vs 115 (+ 47,8%)

KX-6880 3.0GHz (16nm, 8C/8T, 8MB L2 cache) vs ZX-C+ FC-1080 2.0GHz (28nm, 8C/8T, 4MB L2 cache)

SPEC2006 INT: 29.2 vs 16.4 (+ 78%)
SPEC2006 INT RATE: 170 vs 71,8 (+ 136,7%)

Testing agency: Ministry of Industry and Information Technology Software and Integrated Circuit Promotion Center Sapp evaluation center
Test platform: KX-U5680 + ZX-200 vs ZX-C+ FC-1080 + ZX-100S vs KX-5640 + ZX-200 vs ZX-C + 4700 4 core + ZX-100S
Frequency: 2.0GHz
OS: Ubuntu 14.04
Compiler: Intel C / C ++ / Fortrant compiler 15.0.2

https://i.imgur.com/bn8Dmyu.png

source: www.zhaoxin.com ('http://www.zhaoxin.com/ZXC.aspx?seriesid=15') vs www.zhaoxin.com ('http://www.zhaoxin.com/ZXC.aspx?seriesid=13') vs www.zhaoxin.com ('http://www.zhaoxin.com/ZXC.aspx?seriesid=14') vs www.zhaoxin.com ('http://www.zhaoxin.com/ZXC.aspx?seriesid=5')

Tralalak

2019-06-21, 07:38:40

Chinese-Produced Zhaoxin KX-6000 CPUs Purportedly Match Intel's Core i5-7400

source: https://www.tomshardware.com/news/china-zhaoxin-kx-6000-core-i5-7400,39694.html

MSABK

2019-06-22, 02:15:44

Kann es sein, dass sich das mit der CPU erledigt hat?

Auch das AMD-Joint-Venture THATIC wurde von der US-Regierung auf die Liste gesperrter Unternehmen gesetzt. Erst Anfang Juni war bekannt geworden, dass sich AMD aus dem Joint Venture mit der Investorengruppe Tianjin Haiguang Advanced Technology Investment Co., Ltd. (THATIC) teilweise zurückziehen möchte, den x86-Technologie-Transfer einschränken wird und keine Zen-2-basierten Prozessoren aus China kommen werden. Diese Pläne werden durch die US-Regierung nun schneller und härter umgesetzt als von AMD beabsichtigt.

https://www.computerbase.de/2019-06/usa-handelsverbot-china-supercomputer-amd-thatic/

Loeschzwerg

2019-06-22, 10:04:11

Für VIA/Zhaoxin sicherlich nicht.

Benutzername

2019-06-22, 11:26:37

Für VIA/Zhaoxin sicherlich nicht.

Naja, Centaur ist wohl noch unter dem Radar. die kommen vielleicht auchn och dran? Sind die da überhaupt noch groß involviert?:confused:

Das AMD joint Venture kann sich auch schnell wieder bilden, wenn sich die politische Wetterlage ändert. und nichts hindert die chinesen wirklich daran Zen unabhängig selbst weiter zu entwickeln. Genauso die Zhaoxin Prozessoren.

Loeschzwerg

2019-06-22, 11:42:04

Benutzername

2019-06-22, 12:44:27

Das CPU Design der Zhaoxin SoCs stammt immer noch von Centaur und Centaur wiederum ist ein eigenständiges Tochterunternehmen von VIA (Taiwan!). Insofern wüsste ich nicht was die USA hier zu melden haben, zumal die SoCs des Joint-Ventures Zhaoxin (KX-5000 und jetzt dann KX-6000) auch aktuell nur im asiatischen Markt verkauft werden.

Sitzen die nicht immernoch in den USA?

edit: ja tun sie:

Our mailing address is:
Centaur Technology, Inc.
7600-C N. Capital of Texas Hwy, Suite 300
Austin, TX 78731

https://centtech.com/contact/

Loeschzwerg

2019-06-22, 13:25:55

Spielt das eine Rolle wenn die sämtliche IP bei VIA und damit in Taiwan liegt?

Benutzername

2019-06-22, 14:04:48

Spielt das eine Rolle wenn die sämtliche IP bei VIA und damit in Taiwan liegt?

Ja, das IP können sie nicht blockieren, aber Centaur effektiv funktionsunfähig machen, wenn die US Regierung wirklich will, kann sie. Zum Beispiel die Steuerbehörde IRS Centaur auf den Hals schicken. Schönes Ingenieurbüro, daß sie da haben. Eine Schande, falls dem etwas passierte.

Loeschzwerg

2019-06-22, 14:13:53

Klar, so einen Mist kann man machen, aber warum sollten die Staaten das tun? Das ist Taiwan (ROC) und nicht China (PRC).

Letztendlich würde es Arbeitsplätze in den USA kosten und den KnowHow Transfer nur weiter antreiben.

Benutzername

2019-06-23, 15:56:17

Klar, so einen Mist kann man machen, aber warum sollten die Staaten das tun? Das ist Taiwan (ROC) und nicht China (PRC).

Uuups. Scheiße. Du hast natürlich recht. :redface::redface: Ich hatte Via in VR China verortet. Keine Ahnung, wieso. :freak:

Letztendlich würde es Arbeitsplätze in den USA kosten und den KnowHow Transfer nur weiter antreiben.

Ob trump das wirklich bremsen kann wird sich zeigen müssen.

Tralalak

2019-07-02, 10:46:48

Zhaoxin KX-6000 / KH-30000 Series SoC on Zhaoxin website

http://www.zhaoxin.com/Upload/201907011455430330.png

KaiXian KX-6000 series 4-core processor

KaiXian KX-6740A 2.7GHz
KaiXian KX-6640A 2.6GHz
KaiXian KX-6640MA 2.2GHz @ 2.6GHz Max turbo frequency (Adaptive Overclocking)

KaiXian KX-6000 series 8-core processor

KaiXian KX-U6880A 3.0GHz
KaiXian KX-U6780A 2.7GHz
KaiXian KX-U6780 2.7GHz

KaiSheng KH-30000 series 8-core processor

KaiSheng KH-37800 2.7GHz
KaiSheng KH-37800D 2.7GHz

source: http://www.zhaoxin.com/Solution.aspx?id=3

Loeschzwerg

2019-07-02, 12:58:17

Any infos on the TDP?

Tralalak

2019-07-02, 14:14:49

Only rumours: TDP is not, it is estimated that 4 cores SoC do not exceed ~30W, eight cores SoC do not exceed ~70W...

EDIT:

Zhaoxin KaiSheng KH-37800D 2.7GHz (8C/8T with 8MB L2 Cache) supports for inter-chip dual-channel interconnect technology (two-way interconnection) and supports dual channel DDR4 ECC UDIMM/RDIMM with up to 64GB or 128GB (dual)

Tralalak

2019-07-03, 11:20:25

SisoftSandra Ranker
ZHAOXIN KaiXian KX-U6780A 2.7GHz (8C/8T, 8MB L2 cache)

Processor Arithmetic: 71.77GOPS
Processor Multi-Media: 116.47Mpix/s

source: https://ranker.sisoftware.co.uk/show_device.php?q=c9a598c28acb84dc95dbfbb0d1b8e089e886a6edb598cdfbccf4c485a3c4f9 d4e5c3b18cbd9bf2cfffd9b18cbc9ae2dfefc9acc9f4c4e291ac9c&l=en

Edit:

Memory Bandwidth: 18.04GB/s

source: https://ranker.sisoftware.co.uk/show_run.php?q=c2ffcdffd9b8d9e4dde5d2ead8e0c6b489b99ffa9fa292b4c7fac2&l=en

SisoftSandra Ranker
ZHAOXIN KaiXian KX-6640A@2.6GHz (4C/4T, 4MB L2 cache)

Processor Arithmetic: 36,01GOPS
Processor Multi-Media: 57,90Mpix/s

source: https://ranker.sisoftware.co.uk/show_device.php?q=c9a598c28acb84dc95dbfbb0d1b8e089e886a6edb598ae98ac9cddfb9ca18c bd9be9d4e5c3aa97a781e9d4e4c2ba87b791f491ac9cbac9f4cc&l=en

Loeschzwerg

2019-07-19, 08:35:12

Von CJOYIN gibt es nun ein COM Express Modul mit KX-5000 SoC:
http://www.cjoyin.com/en/zx-c701-自主可控come核心.html

67185 67186

konkretor

2019-07-22, 08:17:54

Linux Support is incoming

http://lkml.iu.edu/hypermail/linux/kernel/1907.1/00156.html

https://www.phoronix.com/scan.php?page=news_item&px=Linux-5.3-x86-CPU-Changes

Tralalak

2019-07-22, 14:12:23

Zhaoxin KX-6000 Series SoC

http://www.zhaoxin.com/app_upload/images/1(64).jpg

http://www.zhaoxin.com/app_upload/images/3(26).jpg

http://www.zhaoxin.com/app_upload/images/2(41).jpg

http://www.zhaoxin.com/app_upload/images/4(16).jpg

http://www.zhaoxin.com/app_upload/images/5(16).jpg

source: http://www.zhaoxin.com/InCenterContent.aspx?id=241

Tralalak

2019-07-22, 16:32:16

ZPI2.0 with 3x Zhaoxin KX-37800D SoC

http://www.zhaoxin.com/app_upload/images/4(16).jpg

Per to per Senders
Full duplex
128Gbps pre socket

3x Eight cores KX-37800D 2.7GHz SoC => Mainboard with Socket 0 + Socket 1 + Socket 2 = 24 Cores ?

Tralalak

2019-07-22, 16:33:04

Blue Shield shares participating company Yunhai Qilin, together with Zhaoxin released the first megacore series two-way server YH-5210ZXv2 with Kaisheng KH-30000 series processor, supports ECC memory and also has dual-channel interconnected ZPI2.0 technology
source: https://translate.google.com/translate?hl=sk&sl=zh-CN&tl=en&u=http%3A%2F%2Fwww.hxxkw.org%2F2019%2Fzx_0708%2F82127.html

konkretor

2019-07-23, 14:48:07

https://www.golem.de/news/kh-30000-zhaoxin-spricht-ueber-2-sockel-plattform-1907-142714.html

Tralalak

2019-07-26, 16:59:46

Zhaoxin Kaixian KX-6000 series processor won three important awards in China's IC Design Achievement Award in 2019
source: https://translate.google.com/translate?hl=sk&sl=zh-CN&tl=en&u=http%3A%2F%2Fwww.elecfans.com%2Fd%2F1003181.html
source: http://www.zhaoxin.com/InCenterContent.aspx?id=242

Tralalak

2019-07-30, 10:14:11

In-depth understanding of the Zhaoxin KX-6000 / Kaisheng KH-30000 series processor

http://www.zhaoxin.com/app_upload/images/5(18).jpg

source: http://www.zhaoxin.com/InCenterContent.aspx?id=243

Loeschzwerg

2019-08-06, 07:16:47

Infos und Verlauf zu den GPUs von Jingjia Microelectronics Co:
https://news.mydrivers.com/1/639/639540.htm

67373

Google Übersetzung:
The most advanced GPU in China, the performance can be traced to GTX 1080 graphics card

In the field of CPU processors, China has been involved in a variety of instruction set architectures such as ARM, MIPS and even X86. Both self-developed and authorized have achieved success, but in the GPU field, domestic manufacturers are facing AMD and NVIDIA. Shadows can't catch up.

The threshold of the GPU market is too high, so there are many companies engaged in GPU independent research and development in China, but there are not many achievements. Changsha Jingjia Micro has started with military GPUs. The JM5400 GPU developed before has already achieved domestic replacement in the military field.

According to the information, Changsha Jingjia Microelectronics Co., Ltd. was established in April 2006. It is a leading military image insurance leader in China. Its main products are core modules and system products for military graphics insurance and small special radar. Designed FPGA verification, IP design technology. Its products are used in aviation, aerospace, navigation, automotive, industrial control and other fields.

The JM5400 GPU in the company's Jingmei series is manufactured using the 65nm process. The technology level is similar to ATI's 2003 M72 series GPU, but it is still available for military 2D display.

Last year, Jingjia Micro announced the JM7200 series GPUs with 28nm process technology and 1.2GHz core frequency. With 4GB DDR3 memory, the performance is similar to NVIDIA's GT 640 graphics card, but the overall power consumption is less than 10W, which is much lower than the latter's 50W TDP.

The JM7200 series GPU chips were streamed and verified last year. It was reported that the first batch of orders was obtained in March this year.

...

Then there is the JM9 series, which is still under development. The highest end is the JM9271 series, the core frequency is not less than 1.8GHz, supports PCIe 4.0 x16, with 16GB HBM memory, bandwidth 512GB/s, floating point performance up to 8TFLOPS The performance is close to the level of GTX 1080.

From the officially released indicators, if the JM9 series can reach the level of GTX 1080 and it will be launched in 2020, the gap between domestic GPU and advanced level can be shortened to 4 years, but the JM9 series is still under development, the specific situation Little is known.

Last weekend, Jingjiawei Company said in response to questions from the investor platform, "According to the company's current understanding, some research institutes in China are currently working on R&D of graphics processing chips. The company has many years of experience in graphics processing chip development and has established a professional. The R&D team with wide coverage and high R&D level has certain technical advantages in the field of domestic graphics processing chips. The company's next-generation graphics processing chip JM9200 is under preliminary research. Please pay attention to the company announcement regarding the specific conditions of the company's products.

Ghost1nTh3GPU

2019-08-06, 07:32:21

Ob die Karten auch Endkunden in die Hände bekommen? Die DYI-Szene ist in China mangels Konsolen ja relativ groß. Aber fragwürdig ob man so schnell diese Leistungsregionen erreicht. Wohl eher Bluffs für den Handelskrieg?

robbitop

2019-08-06, 08:38:54

Ob effiziente GPUs ohne Patentsverletzung möglich sind? Die richtig guten Ideen sind größtenteils bereits geschützt. Das ist ein riesen Hinderniss für einen Eintritt in den GPU Markt heutzutage. Wobei das ja etwas ist, was in China völlig ignoriert wird. Verkauf in Länder, die Patente achten wäre dann aber wohl schwierig.

ilPatrino

2019-08-06, 12:33:27

Loeschzwerg

2019-08-06, 13:51:32

Der Fokus dürfte hier rein auf den "internen" Nutzen liegen, also militärisch/staatlich. Insofern kratzen da weder Patente noch Treiber für Spiele.
Irgendwann fällt dann vielleicht auch was für die breite Masse bzw. den globalen Markt ab. Spätestens dann wenn die Leistung vollends konkurrenzfähig ist, würde ich vermuten. Dauert aber sicher noch ein paar Jahre.

Ghost1nTh3GPU

2019-08-06, 19:41:14

Die China GTX 1080 soll für 110 Mio. USD entwickelt werden: https://m.expreview.com/65091.html
Aber man will wohl definitv die Importe an US-Gaming-Grafikkarten reduzieren.

f-x-

2019-08-10, 01:56:37

gravitationsfeld

2019-08-10, 07:41:32

Die China GTX 1080 soll für 110 Mio. USD entwickelt werden: https://m.expreview.com/65091.html
Aber man will wohl definitv die Importe an US-Gaming-Grafikkarten reduzieren.
110 Mio waeren zumindest im Westen viel zu wenig?

Ghost1nTh3GPU

2019-08-10, 18:18:05

Wird auch im Text angesprochen. Aber ist der Entwicklungsaufwand für eine entsprechende GPU-Architektur wirklich so hoch? Viele Ansätze für das Design sind wohl da schon den Ings. im Sektor bekannt, wenn vielleicht auch durch Patente geschützt. Der größte finanzielle Aufwand liegt da wohl eher im Tapeout für so einen >200mm2 Chip und der Vermarktung.

Benutzername

2019-08-10, 18:37:26

Langlay

2019-08-10, 19:16:20

interessant wäre noch, ob man die karten (egal ob offiziell oder inoffiziell) außerhalb chinas zu kaufen bekommen wird (patente?) und wie der treiber-support für spiele aussehen wird, die in china nicht so verbreitet sind.

Über Aliexpess wird man die sicherlich auch im Westen kaufen können, wenn sie in China frei verfügbar sind.

Tralalak

2019-08-17, 00:01:32

VIA SLV-950 @ VIA/Zhaoxin C-OctaCore FC-1080@2.0GHz (8C 2GHz, 2MB L2)
VIA device with older 8-core processor
source: https://ranker.sisoftware.co.uk/show_system.php?q=cea598ab9ca594ac9eb8dfe2cffed8aa97a680e9d4e4c2aa97a781f9c4f4d2 b7d2efdff98ab78f&l=en

Tralalak

2019-08-23, 18:40:38

vMswfcT0hRA

Loeschzwerg

2019-08-26, 06:15:58

Eine Tabelle mit Angaben zur TDP sämtlicher Zhaoxin x86 SKUs:
67577

Loeschzwerg

2019-08-29, 06:52:25

Inoffizielle Aussage von Lenovo:
Es gibt aktuell keine Bestrebungen Produkte mit Zhaoxin Prozessoren außerhalb China einzuführen.

Die Geschichte bleibt vorerst also weiterhin eine lokale Angelegenheit.

Tralalak

2019-08-30, 12:01:24

hq-hq

2019-08-30, 12:59:11

ZHAOXIN KaiXian KX-U6780A@2.7GHz (8C, 2.7GHz, 2x 4MB L2)
HP 268 Pro G1 MT HP 268 (HP 86D1)
source: https://ranker.sisoftware.co.uk/show_system.php?q=cea598ab9ca593a795b3d4e9c4f5d3a19cad8be2dfefc9a19cac8af2cfffd9 bcd9e4d4f281bc84&l=en

Higher than 16.43% ranked results :freak:

bin mir jetzt nicht ganz sicher, aber sieht so aus als ob selbst ein 10 Jahre alter Turion II mehr Performance aufbieten kann

Lehdro

2019-08-30, 14:25:05

Higher than 16.43% ranked results :freak:

bin mir jetzt nicht ganz sicher, aber sieht so aus als ob selbst ein 10 Jahre alter Turion II mehr Performance aufbieten kann
Das Ding ist laut sisoftware ungefähr so schnell wie ein 4M8T Bulldozer oder ein i7 950. Also deutlich schneller als ein Turion II.

Tralalak

2019-08-30, 17:10:12

Higher than 16.43% ranked results :freak:

bin mir jetzt nicht ganz sicher, aber sieht so aus als ob selbst ein 10 Jahre alter Turion II mehr Performance aufbieten kann

:nono:

There is only one Individual Results for device HP 268 Pro G1 MT HP 268 (HP 86D1) in Processor Multi-Media and Global Rankv#133453 Position: Higher than 16.43% ranked results is average.

If you clicked only to Processor Multi-Media result is Higher than 42.50% ranked results or another better device with this cpu result is Higher than 43.18% ranked results.

Das Ding ist laut sisoftware ungefähr so schnell wie ein 4M8T Bulldozer oder ein i7 950. Also deutlich schneller als ein Turion II.

(y)

hq-hq

2019-08-30, 21:37:31

okay, sorry you are right, seems to be equal to an i7 950

i7 950 VS KX-U6780A
67.35GOPS VS 71.77GOPS
100.06Mpix/s VS 95.75Mpix/s

at least better than expected

to compare it with other notebooks
Intel Core i5-7260U is a bit slower and Core i5-8350U is bit faster?

Benutzername

2019-09-02, 14:39:16

Über Aliexpess wird man die sicherlich auch im Westen kaufen können, wenn sie in China frei verfügbar sind.

tomshw schreibt, daß das wohl nunmbercruncher für Luft- und Raumfahrt und Militär werden. Bisher haben die Treiber kein Vulkan oder DirectX12. Auch wenn jingjia Micro schon die Fühler in Richjtung Verbraucher austreckt mit einer kleinen 3D Karte.

However, Jingjia Micro's hunger cannot be satisfied and has already started to promote its GPUs outside the military market. The JM7200 is the firm's most recent GPU. It's clocked at 1,200 MHz and comes with 4GB of DDR3 memory. For reference sake, the JM7200's performance is equivalent to a GeForce GT 640. However, the JM7200 has a TDP (thermal design power) that's less than 10W as opposed to the GeForce GT 640's 50W. The Chinese company is currently receiving orders for its JM7200 GPU and has gone back to the drawing board to plan its next-gen GPUs.

https://www.tomshardware.com/news/jingjia-micro-chinese-gpu-pcie-4.0-16gb-hbm-gtx-1080-performance,40217.html

Tralalak

2019-09-02, 16:29:19

PassMark - CPU Mark (Single Thread Performance)

Intel Atom C3958 @ 2.00GHz (16C/16T, 31W): 6 586 (791)
AMD FX-8100 Eight-Core (8C/8T, up to 3.7GHz 95W): 6 008 (1 097)
Intel Core i5-3335S @ 2.70GHz (4C/4T, up to 3.2GHz 65W): 5 788 (1 676)
ZHAOXIN KaiXian KX-U6780A @ 2.7GHz (8C/8T, 70W): 5 760 (877)
AMD Phenom II X6 1100T @ 3.3GHz (6C/6T, up to 3.7GHz, 125W): 5 702 (1 250)
Intel Core i7-950 @ 3.07GHz (4C/8T, up to 3.33GHz, 130W): 5 536 (1 327)
AMD Athlon 220GE @ 3.4GHz: (2C/4T, 35W): 5 302 (1 752)
Intel Core2 Extreme X9775 @ 3.20GHz (4C/4T, 150W): 4 646 (1 391)
Intel Atom C2750 @ 2.40GHz (8C/8T, up to 2.6GHz, 20W): 3 806 (584)
Intel Pentium Silver J5005 @ 1.50GHz (4C/4T, up to 2.8GHz, 10W): 2 909 (1 192)
VIA QuadCore C4650 @ 2.0GHz (4C/4T, 18W): 1 897 (527)
Intel Pentium J2850 @ 2.41GHz (4C/4T, 10W): 1 839 (548)

source: https://www.cpubenchmark.net/mid_range_cpus.html
source: https://www.cpubenchmark.net/midlow_range_cpus.html
source: https://www.cpubenchmark.net/singleThread.html

Tralalak

2019-09-05, 16:24:04

Shanghai Zhaoxin Semiconductor Co., Ltd. CHA PXP_MB
KX-7000 in Geekbench benchmark ?!?

Name: VIA QuadCore
Topology: 1 Processor, 8 Cores
Identifier: CentaurHauls Family 6 Model 71 Stepping 1
Base Frequency: 1.60 GHz
L1 Instruction Cache: 32.0 KB x 8
L1 Data Cache: 32.0 KB x 8
L2 Cache: 256 KB x 8
L3 Cache: 16.0 MB x 1

source: https://browser.geekbench.com/v4/cpu/14554775

ZHAOXIN KaiXian KX-U6780A @ 2.70 GHz vs VIA/ZHAOXIN "KX-7000" @ 1.60 GHz

Single-Core Score 1910 vs 1898
Multi-Core Score 8670 vs 10348

source: https://browser.geekbench.com/v4/cpu/compare/12754288?baseline=14554775

Loeschzwerg

2019-09-05, 17:22:38

Maybe a first hint about new VIA products. A low-power 8 core processor @ 7nm could fit nicely in a upcoming embedded lineup in 2020.

Loeschzwerg

2019-09-09, 06:18:20

Auf der IC China 2019 hat Lenovo einen Desktop-PC mit KX-6000, den Kaitian T630Z, vorgestellt:
http://www.eetrend.com/article/2019-09/100131526.html (Original)
https://translate.google.de/translate?hl=de&sl=zh-CN&tl=en&u=http%3A%2F%2Fwww.eetrend.com%2Farticle%2F2019-09%2F100131526.html (Google Übersetzung)

Tralalak

2019-09-09, 11:00:45

Core dynamics | domestic chip high-performance security platform butterfly

"In the case of full configuration, 20 multi-core domestic CPUs are used in a single system to maximize the performance of each chip and meet the urgent needs of users for high-performance security products."

source: http://www.zhaoxin.com/InCenterContent.aspx?id=252

Tralalak

2019-09-10, 01:10:35

Changing The Future Of Server & Scalable Computing
High Performance, x86 Compatible Microprocessors

source: https://centtech.com

P.S. Watch the video carefully and enjoy the images on the new high performance compatible x86 microprocessor :up:

Loeschzwerg

2019-09-10, 07:10:50

Thanks :) Didn't notice there was a site update.

A new LGA based CPU?
https://www.forum-3dcenter.org/vbulletin/attachment.php?attachmentid=67715&stc=1&d=1568091869 (from the video)

https://www.forum-3dcenter.org/vbulletin/attachment.php?attachmentid=67716&stc=1&d=1568092380
https://www.instagram.com/p/B1wBGHfBJUU/?utm_source=ig_web_copy_link

y33H@

2019-09-10, 07:29:43

hui, da kommt was =)

Tralalak

2019-09-10, 11:46:15

Thanks :) Didn't notice there was a site update.

A new LGA based CPU?

It looks like. :up:
For a couple of years ago, Zhaoxin has heard that they wants to sell LGA based CPU alone.

Tralalak

2019-09-15, 09:00:40

New Chinese x86 Zhaoxin processor leaked. Next-gen architecture has IPC like Skylake / Zen 2
source: https://translate.google.com/translate?hl=sk&sl=cs&tl=en&u=https%3A%2F%2Fwww.cnews.cz%2Fprocesor-via-zhaoxin-kx-7000-unik-geekbench-nova-architektura-vysoke-ipc%2F

Tralalak

2019-09-16, 16:23:52

Linux-Kernel 5.3 unterstützt x86-CPUs von Zhaoxin
Quelle: https://www.heise.de/ct/artikel/Linux-5-3-freigegeben-Prioritaeten-deckeln-und-Trouble-fuer-Nvidia-4470638.html?seite=3

Tralalak

2019-09-20, 00:00:13

Zhaoxin KaiSheng KH-37800D

With the aid of ZPI (Zhaoxin Processor Interconnect), KH-30000 is capable of building up a system using two processors which will maximize the processor threads and memory capability.
Support Dual-channel DDR4 ECC UDIMM/RDIMM memory, 128GB Max(Dual-way)

source: http://en.zhaoxin.com/ZXC.aspx?seriesid=18

Tralalak

2019-09-24, 11:44:38

FinalWire Rolls Out AIDA64 v6.10
Zhaoxin Benchmarks and AMD Radeon RX 5700 Series Support

AVX optimized 64-bit benchmarks for Zhaoxin processors
AIDA64 CPUID Panel, Cache & Memory Benchmark panel, GPGPU Benchmark panel, System Stability Test, and all cache, memory and processor benchmarks are fully optimized for Zhaoxin KaiXian KX-5000 Series and KaiXian ZX-C+ processors as well as for KaiSheng KH-20000 Series and KaiSheng ZX-C+ server processors, utilizing AVX, AES-NI and PadLock instructions. Detailed chipset information for Zhaoxin ZX-100, ZX-100S and ZX-200 south bridges, as well as for the integrated memory controllers of various Zhaoxin processors. PCI Express 4.0 controllers and devices support. Preliminary support for Zhaoxin KaiXian KX-6000 Series processors.

https://www.aida64.com/sites/default/files/shot4_cpuid_kx5000.png

https://www.aida64.com/sites/default/files/shot5_kx5000_aes_en.png

https://www.aida64.com/sites/default/files/shot6_cachemem_kx5000.png

AVX accelerated 64-bit benchmarks for Zhaoxin ZX-C+ and Zhaoxin KaiXian KX-5000 processors
Full support for Hygon C86 “Mukti/Dhyana” CPU
Enhanced RGB LED monitoring module
Matrix Orbital EVE3 LCD display support
Improved support for Intel “Comet Lake” and “Ice Lake” CPUs
Support for PCI Express 4.0 controllers and devices
Preliminary support for Zhaoxin KaiXian KX-6000 CPU
Advanced support for LSI RAID controllers
NVMe 1.4 support
Improvements for AMD X570 chipset based motherboards
GPU details for AMD Radeon RX 5700 Series
GPU details for nVIDIA GeForce RTX 2000 Super Series

source: https://www.aida64.com/news/aida64-v610-zhaoxin-benchmarks
source: https://www.aida64.com/news/pr/aida64-v610-press-release

Tralalak

2019-10-02, 17:17:40

Geekbench 4.4.0 Tryout for Windows x86 (64-bit)
ZHAOXIN KaiXian KX-6640MA@2.2+GHz (4C/4T, up to 2.6GHz Turbo, 25W TDP)

Single-Core Score : 1 698
Multi-Core Score : 3 794

source: https://browser.geekbench.com/v4/cpu/14725901v

Loeschzwerg

2019-10-02, 18:04:31

About Goldmont Level of performance I'd say.

Tralalak

2019-10-08, 20:27:27

Geekbench 5.0.2 Tryout for Linux x86 (64-bit)
2x Zhaoxin Kaisheng KH-37800D @ 2.7GHz CPUs (16C/16T)
source: https://browser.geekbench.com/v5/cpu/341928

Tralalak

2019-10-13, 01:51:37

ZHAOXIN KaiXian KX-U6780A@2.7GHz

https://i.imgur.com/fkc60nM.jpg

Tralalak

2019-10-24, 00:40:38

Tralalak

2019-11-01, 12:12:58

ZPI2.0 (Zhaoxin Processor Interconnect) 128 GBit/s per socket
CLOUD KIRIN YH-5210ZXv2 with 2x Kaisheng KH-30000 series 8-core CPUs (16C/16T)

http://r12.35.com/home/f/7/nmtfym/resource/2019/07/31/5d40f8199f8a6.png

source: http://www.cloudkirin.com/article-5077-38108.html

Tralalak

2019-11-04, 21:43:19

CentaurHauls Family 6 Model 71 Stepping 1
ZHAOXIN KX-7000 ZX-F OctaCore 2000MHz @ 2.00 GHz
1 processor, 8 cores

L1 Instruction Cache: 32 KB x 8
L1 Data Cache: 32 KB x 8
L2 Cache: 256 KB x 8
L3 Cache: 16 MB x 1

Single-Core Performance
Single-Core Score: 469
Crypto Score: 710
Integer Score: 446
Floating Point Score: 478

Multi-Core Performance
Multi-Core Score: 3 264
Crypto Score: 4 553
Integer Score: 3 133
Floating Point Score: 3 334

source: https://browser.geekbench.com/v5/cpu/526995
source: https://browser.geekbench.com/v5/cpu/526995.gb5

Tralalak

2019-11-07, 17:18:07

Centaur Technology's Deep-Learning Coprocessor Technology
source: http://www.ece.utexas.edu/events/centaur-technologys-deep-learning-coprocessor-technology?fbclid=IwAR1Hf8qo1eGyVG5DL_d1sa11HCDEU5OpZZ0UGeR0D3Aoa_sdB9FXdBusQog

Centaur Technology Reference Design v1.0 (8-core Centaur Integrated x86 CPUs with Centaur Integrated AI Coprocessor)
source: https://mlperf.org/inference-results
source: https://www.cdrinfo.com/d7/content/mlperf-releases-results-machine-learning-inference-benchmark

Loeschzwerg

2019-11-07, 17:25:14

8 Core @ 2.5GHz + AI Accelerator.
https://github.com/mlperf/inference_results_v0.5/blob/master/closed/CentaurTechnology/systems/0.json

{
"division": "closed",
"status": "preview",
"submitter": "CentaurTechnology",
"system_name": "Centaur Technology Reference Design v1.0",
"number_of_nodes": 1,
"host_memory_capacity": "32GB",
"host_processor_core_count": 8,
"host_processor_frequency": "2.5GHz",
"host_processor_model_name": "Centaur Integrated x86 CPUs",
"host_processors_per_node": 1,
"host_storage_capacity": "120GB",
"host_storage_type": "SSD",
"accelerator_frequency": "-",
"accelerator_host_interconnect": "-",
"accelerator_interconnect": "-",
"accelerator_interconnect_topology": "-",
"accelerator_memory_capacity": "4GB",
"accelerator_memory_configuration": "none",
"accelerator_model_name": "Centaur Integrated AI Coprocessor",
"accelerator_on-chip_memories": "-",
"accelerators_per_node": 1,
"framework": "TensorFlow commit f5ce1c00d4397875ff3d706881bd46430f4a9667 and f6fbfe013898dc16bd35ba380387ff02d0275ac3 + custom patches",
"operating_system": "ubuntu-18.04",
"other_software_stack": "Centaur ML Library Build 1223 and 1220",
"sw_notes": "The same hardware system is used for all mobilenet, resnet, ssd-small, and gnmt sumbissions, but the software is slightly different for gnmt.
Submissions for mobilenet, resnet, and ssd-small use TF commit f5ce1c00d4397875ff3d706881bd46430f4a9667 + custom patches to TF-Lite,
Centaur ML Library Build 1223, and host_processor_frequency set to 2.5GHz. Submissions for gnmt use TF commit f6fbfe013898dc16bd35ba380387ff02d0275ac3
+ custom patches for compiler/graph matching, Centaur ML Library Build 1220, and host_processor_frequency set at runtime via software to 2.3GHz."
}

Galaxie

2019-11-08, 02:24:41

Nur kann man nichts davon kaufen :(

Tralalak

2019-11-08, 20:46:28

First KX-6000 Series PCs

http://imgsrc.baidu.com/forum/pic/item/299f5867d0160924b45a9f6ddb0735fae6cd342f.jpg

http://imgsrc.baidu.com/forum/pic/item/a6b9bcfc5266d0161766a35e982bd40735fa352f.jpg

http://imgsrc.baidu.com/forum/pic/item/1ea933dab6fd5266cd540162a418972bd407362f.jpg

http://imgsrc.baidu.com/forum/pic/item/42a65bdb81cb39dbadbf5a20df160924ab18302f.jpg

source: http://tieba.baidu.com/p/6329337191

Tralalak

2019-11-09, 18:41:32

AMD FX-8300 3300MHz @ 2.00GHz
Single-Core Score: 306 (65,25%)
Multi-Core Score: 1 705 (52,24%)

vs

KX-7000 ZX-F OctaCore 2000MHz @ 2.00GHz
Single-Core Score: 469 (+53,27%)
Multi-Core Score: 3 264 (+91,44%)

source: https://browser.geekbench.com/v5/cpu/compare/526995?baseline=562975

___________________
https://i.imgur.com/srR1pdG.jpg
source: https://browser.geekbench.com/v5/cpu/562975.gb5
source: https://browser.geekbench.com/v5/cpu/562975

P.S. If you have Geekbench 5 results your CPU @ 2.00GHz so pls write here! Thank you.

woodsdog

2019-11-10, 08:40:17

Und wenn sie ihn gegen einen Pentium 100 benchen würden, könnten Sie 5000% dran schreiben... Was soll dieser Quatsch gegen eine 8 Jahre alte CPU, der man auch noch 40% des Takts klaut?? lächerlich.

robbitop

2019-11-10, 08:47:24

Es ging ihnen um IPC. Das ist nahezu Zen1 Level. (+57% ggü Bulldozer in st)

woodsdog

2019-11-10, 09:06:54

Zen@2ghz ist verfügbar und währe ein ehrlicher Vergleich.
"wir haben 50% mehr IPC verglichen mit einer 10 Jahre alten Arch" bedeutet nix anderes als "wir sind komplett nicht konkurrenzfähig".

Ist aber ja eh alles egal, weil man den ganzen Krempel selbst mit einem Anwendungsfall sowieso nicht kaufen kan v0v

robbitop

2019-11-10, 09:11:32

Tralalak

2019-11-10, 09:19:27

The Geekbench already has a 2GHz sample of the Chinese x86 Zhaoxin KX-7000 processor. 80% IPC Zen 2

source: https://translate.google.com/translate?hl=sk&sl=cs&tl=en&u=https%3A%2F%2Fwww.cnews.cz%2Fprocesor-via-zhaoxin-kx-7000-unik-geekbench-2ghz%2F

_________
P.S. If you have Geekbench 5 results your modern CPU @ 2.00GHz so pls write here! Thank you.

woodsdog

2019-11-10, 09:23:52

Es bringt nichts und trägt auch zum Thread nichts bei, die Via CPUs schlecht zu reden. Ja es gibt besseres. Aber interessant ist es dennoch, dass es x86 IP gibt, was von der IPC nicht mehr so weit entfernt ist und nicht von Intel und AMD stammt.
Den ganzen destruktiven Kram empfehle ich, hier wegzulassen, da kein Mehrwert für den Thread.

Ich rede hier keiner Hardware schlecht, ich bemängel den unsinnigen Vergleich gegen uralte Archs.

Wenn das hier aber ein VIA-Marketing-Thread ist, in dem ausschließlich positive Aspekte des Produktes angesprochen werden dürfen, tut es mir leid gepostet zu haben. Ist sonst im Forum nicht so.

woodsdog

2019-11-10, 09:25:53

Tralalak

2019-11-10, 09:37:40

Ich habe zu Hause keinen stärkeren Prozessor als den AMD FX-8300, den ich bei @ 2,0 GHz underclocked.
Deshalb bitte ich die Diskussionsteilnehmer, dies zu tun.

Loeschzwerg

2019-11-10, 12:05:07

Man nimmt eben was man hat, passt doch :) Wer sich nicht für das Thema interessiert hat doch hier im Forum genügend andere Spielplätze.

@Tralalak: Thanks for your informations, keep it up :up:

robbitop

2019-11-10, 19:11:51

Sehe ich auch so. In beiden Fällen.

Tralalak

2019-11-12, 00:43:56

Making a Microprocessor: Centaur Technology

https://www.facebook.com/cognitivefilms/videos/530671834177833/?t=1

Skysnake

2019-11-12, 20:38:56

They use Cadence Virtuoso for layout.

Tralalak

2019-11-17, 16:13:32

MLPerf Inference v0.5 Results
source: https://mlperf.org/inference-results

COCO object detection on SSD-MobileNet v1 (images/sec)

Intel Core i3-1005G1 1.2GHz up to 3.4GHz Ice Lake 2C/4T (Intel® UHD Graphics): 217,93 (33,43 %)
Centaur Integrated x86 CPUs @ 2.5GHz~2.3GHz 8C/8T (Centaur Integrated AI Coprocessor): 651,89 (299,13 %)

ImageNet image classification on MobileNet v1 (images/sec)

Intel Core i3-1005G1 1.2GHz up to 3.4GHz Ice Lake 2C/4T (Intel® UHD Graphics): 507,71 (8,40 %)
Centaur Integrated x86 CPUs @ 2.5GHz~2.3GHz 8C/8T (Centaur Integrated AI Coprocessor): 6 042,34 (1190,12 %)

ImageNet image classification on ResNet-50 v1.5, (images/sec)

Intel Core i3-1005G1 1.2GHz up to 3.4GHz Ice Lake 2C/4T (Intel® UHD Graphics): 100,93 (8,28 %)
Centaur Integrated x86 CPUs @ 2.5GHz~2.3GHz 8C/8T (Centaur Integrated AI Coprocessor): 1 218,48 (1207,25 %)

source: https://mlperf.org/inference-results
source: https://github.com/mlperf/inference_results_v0.5/blob/master/closed/Intel/systems/ICL-i31005G1_openvino.json
source: https://www.intel.ai/mlperf-nov2019/#gs.h01eu8
source: https://github.com/mlperf/inference_results_v0.5/blob/master/closed/CentaurTechnology/systems/0.json

Tralalak

2019-11-18, 18:53:25

THE INDUSTRY’S FIRST HIGH-PERFORMANCE X86 SOC WITH SERVER-CLASS CPUS AND INTEGRATED AI COPROCESSOR TECHNOLOGY

https://centtech.com/wp-content/uploads/MockPageAI-6-505x600.jpg

source: https://centtech.com/ai-technology/
source: https://centtech.com/wp-content/uploads/November-18-2019-press-release-1.pdf
source: https://centtech.com/wp-content/uploads/PRSlides_1118_Release.pdf

Screemer

2019-11-18, 19:17:57

They use Cadence Virtuoso for layout.
Dich gibt's auch noch? Mir fehlen deine technisch versierten Posts.

Loeschzwerg

2019-11-18, 19:20:03

Codename "NCORE", SoC called "CHA" and x86 core "CNS".
• Centaur developed a new x86 microprocessor with high instructions/clock (IPC)
• Microarchitecture designed for server-class applications with extensions such as AVX-512
• New x86 technology now proven in silicon with 8 CPU cores and 16MB L3 caches
• SoC architecture provides an extensible platform with 44 PCIe lanes and 4 channels of PC3200
• Including AI coprocessor, requires less than 195mm2
in 16nm TSMC
• Reference platform running at 2.5GHz today
• Simultaneous execution of x86 cores and 20 TOPS AI Coprocessor
• Delivers 20 peak terabytes/sec to AI Coprocessor from dedicated 16MB SRAM

AI Coprocessor is 34.4mm2 in 16FFC.

More technical details will follow soon:
Microprocessor Report article will be released on December 2, 2019. This report will be a deep dive into the technical details.

Tralalak

2019-11-18, 21:07:37

Centaur’s new CPU is the first x86 with an AI co-processor
They’re baaaack!
source: https://www.semiaccurate.com/2019/11/18/centaurs-new-cpu-is-the-first-x86-with-an-ai-co-processor/

Tralalak

2019-11-19, 10:03:56

VIA CenTaur Develops a Multi-core x86 Processor for Enterprise with in-built AI Hardware
source: https://www.techpowerup.com/261274/via-centaur-develops-a-multi-core-x86-processor-for-enterprise-with-in-built-ai-hardware#comments

VIA TECHNOLOGIES
Centaur zeigt x86-Chip mit AI-Block
Quelle: https://www.golem.de/news/via-technologies-centaur-zeigt-x86-chip-mit-ai-block-1911-145071.html

konkretor

2019-11-19, 10:48:32

someone was faster :D

Badesalz

2019-11-19, 12:48:36

Hört dieser AI Bullshit denn niemals auf?

Ghost1nTh3GPU

2019-11-19, 17:35:13

In China klassifizieren sie Bilder. ;)

x86 ist wohl ein starkes Argument, aber eine solche NPU sollte für AMD/Intel keine Herausforderung sein, sodass der Alleinstellungsfaktor begrenzt sein wird.

Eine China-APU mit VIA x86 und einer noch in Entwicklung befindlichen China-GPU-Arch für Konsolen bzw. Cloud sollte man wohl auch mittelfristig in Betracht ziehen.

Loeschzwerg

2019-11-19, 17:46:20

Nachdem VIA sich in letzter Zeit hauptsächlich auf AI Themen fokussiert hat, besteht sogar die Möglichkeit dass der neue Centaur Prozessor/Kern diesmal nicht ausschließlich in Asien versauert.

Tralalak

2019-11-19, 18:18:58

BYomh3f14Sw

scx

2019-11-20, 01:10:31

Hello!

Do you have any information about "VIA Nano X2 C4350AL" (40nm)? Is it CNC, CNQ or CNR? At first, I thought that it is CNR (maybe early sample of Isaiah II from 2012-2014), because it belongs to the C4000 series, but as far as I know, the last dual core Nano CPU from VIA was CNC. However, there is also Eden X1 C1050, which is a single core CNR, so maybe they released a dual core Nano CNR as well. On the other hand, it may be CNQ, because they belong to the 4000 series as well, although I have only heard about QuadCore L4000 series from CNQ.
Or maybe it isn't a VIA CPU at all, but an alternative name for Zhaoxin ZX-A? As far as I know, this CPU doesn't support AVX, so it should be a derivative of VIA Nano X2 CNC, right? But there is a little chance that VIA has developed VIA Nano X2 C4350AL (as a variant of Nano X2 L4350E?), and this is what ZX-A is based on. Right now, I am so confused.

So, do you have any idea what this could be?

Badesalz

2019-11-20, 01:38:03

https://youtu.be/BYomh3f14SwWas ist das für ein Kernschrott? Ich zuckte schon mit der Maus als in den Untertiteln direkt zu Anfang "blackchain" auftauchte. Und nu? Was hab ich erfahren?

AI und 5G. Und Lügen.

Loeschzwerg

2019-11-20, 07:22:40

Do you have any information about "VIA Nano X2 C4350AL" (40nm)? Is it CNC, CNQ or CNR?

Can you tell us the CPUID or the size of L1/L2 cache?

CN "A" => 6F2 (Family 6, Model 15, Stepping 2) e.g. Nano L2200
CN "B" => 6F8 (Family 6, Model 15, Stepping 8) e.g. Nano L3050
CN "C" => 6FA (Family 6, Model 15, Stepping 10)
CNQ => 6FC and 6FD (Family 6, Model 15, Stepping 12 and 13)
CNR => 6FE (Family 6, Model 15, Stepping 14)

CNR does have 2MB L2, every earlier "Nano" does have 1MB L2.

scx

2019-11-20, 12:20:05

Can you tell us the CPUID or the size of L1/L2 cache?

CN "A" => 6F2 (Family 6, Model 15, Stepping 2) e.g. Nano L2200
CN "B" => 6F8 (Family 6, Model 15, Stepping 8) e.g. Nano L3050
CN "C" => 6FA (Family 6, Model 15, Stepping 10)
CNQ => 6FC and 6FD (Family 6, Model 15, Stepping 12 and 13)
CNR => 6FE (Family 6, Model 15, Stepping 14)

CNR does have 2MB L2, every earlier "Nano" does have 1MB L2.

So, CN "A", CN "B" and CN "C" weren't official names? Was CNQ the first official name?

And what about 000006F1? Are CN "A" and CN without a suffix the same thing? Or maybe CN refers to 6F1 and CN "A" refers to 6F2? Can you provide any example of the 6F1 CPU? Does Nano L2100 belong to CN or CN "A"?
http://instlatx64.atw.hu/
https://github.com/InstLatx64/InstLatx64/blob/ddff8a9210c97b395d1463aa7e24c8bd56c9fcda/index.htm#L203-L205
https://github.com/InstLatx64/InstLatx64/tree/ddff8a9210c97b395d1463aa7e24c8bd56c9fcda/CentaurHauls

Back to the topic:
https://browser.geekbench.com/geekbench2/2480352
https://browser.geekbench.com/geekbench3/665443
Processor: VIA Nano X2 C4350AL @ 1.73 GHz 1 processor, 2 threads
Processor ID: CentaurHauls Family 6 Model 15 Stepping 13
L2 Cache: 1024 KB x 2

Model: LENOVO FFFFFFFFFF
Operating System: COS Desktop 3.8.13.13-cdos i686
Motherboard: LENOVO VT8611
BIOS: LENOVO O09KT11PUS

For comparison:
CNQ: https://browser.geekbench.com/geekbench2/2124912
CNR: https://browser.geekbench.com/geekbench3/3613798

See also:
https://openbenchmarking.org/s/C4350AL
https://browser.geekbench.com/geekbench2/search?q=C4350AL
https://browser.geekbench.com/geekbench3/search?q=C4350AL

The information available (CentaurHauls Family 6 Model 15 Stepping 13, 1 MB cache per core) clearly point to CNQ, made by VIA.
However, please look at other data: COS Desktop (China Operating System), Lenovo BIOS and motherboard. Moreover, I haven't seen a single motherboard with VIA Nano X2 C4350AL that would be available on the Western market.

There are several possibilities:
- It was a sample CPU made by VIA (no mass production).
- It was a special CPU made by VIA for Zhaoxin (or China government, probably Shanghai Municipal Government or Chinese Academy of Sciences).
- It was an early CPU made by Zhaoxin, based on VIA CNQ.

Loeschzwerg

2019-11-20, 17:00:16

Tralalak

2019-11-20, 18:09:34

Demo system with new CPU with AI coprocessor @ ISC East in New York!
source: https://www.instagram.com/p/B5GCSZkhrL2/

scx

2019-11-20, 18:23:02

To my knowledge CNA, CNB and CNC never were official, but could be some internal codenames. I don't know of any released 6F1 processor, I assume those were pure prototypes of the first Nano processors.

6F2 => VIA Nano 1000 and 2000 series; 65nm single core
6F8 => VIA Nano 3000 series; still 65nm single core but added SSE4 instruction-sets
6FA => VIA Nano X2 4000 series; 40nm dual core
6FC => VIA Quadcore (part of the 4000 series; Eden X2); two 40nm dual core on a substrate; mainly same processors DIE as used in VIA Nano X2 products but with tweaks and corrections
6FD and 6FE => VIA Quadcore-E (C4000, Eden X4 series and also Eden X1); a single 28nm quad core

I think that VIA didn't see any need for the C4350AL in Western market and made it available only in Asia region. After all it also wouldn't fit in the already existing lineup of Nano X2 processors (U and L).

Thanks for the clarification!

This is the specification of VIA Nano and Zhaoxin ZX processors according to my current knowledge:

VIA CPUs

CNA (2008) => 6F2 & 6F3 : CentaurHauls Family 6, Model 15, Stepping 2 & 3 AKA VIA Nano 1000/2000 Series, e.g. Nano L2100 - x86-64, SSSE3, 1 MB L2 cache per core, single core, 65nm
CNB (2009) => 6F8 : CentaurHauls Family 6, Model 15, Stepping 8 AKA VIA Nano 3000 Series, e.g. Nano L3025 - SSE4.1, VIA VT (compatible with Intel VT-X)
CNC (2011) => 6FA : CentaurHauls Family 6, Model 15, Stepping 10 AKA VIA Nano X2, e.g. Nano X2 L4050 - 2 cores, basically two Nano 3000 in the same die, 40nm
CNQ (2011) => 6FC & 6FD : CentaurHauls Family 6, Model 15, Stepping 12 & 13) AKA VIA Nano QuadCore, e.g. Nano QuadCore L4650E - 4 cores, basically two Nano X2 in a multi-chip module
CNR (2015, samples available in 2014) => 6FE (CentaurHauls Family 6, Model 15, Stepping 14) AKA VIA Nano QuadCore, Isaiah II, e.g. Nano QuadCore C4650 - SSE4.2, AVX, AVX2, AES-NI, 2 MB shared L2 cache, 4 cores on a single die, 28nm
CNS (2019) - AVX-512, NCORE AI coprocessor, 16 MB shared L3 cache, 8 cores, 16nm TSMC

Truncated versions of Nano (ULV) and Eden may be not included.

Zhaoxin CPUs

ZX-A (2014) => CentaurHauls Family 6, Model 15, Stepping 13 (KaiXian ZX-A), e.g. C4350AL - based on VIA CNQ (VIA Nano X2 C4350AL), 40nm TSMC; compatible with VX11(H) (Chrome 640/645 GPU - 3a01)
ZX-B (2014?) => CentaurHauls Family 6, Model 15, Stepping 13 (KaiXian ZX-B) - the same microarchitecture as ZX-A, 40nm HLMC (different fabrication plant, produced in mainland China)
ZX-C (2015) => CentaurHauls Family 6, Model 15, Stepping 14 AKA ZhangJiang (KaiXian ZX-C), e.g. C4610 - based on VIA CNR, 2 MB shared L2 cache, 4 cores, 28nm TSMC; compatible with VX11(PH) (Chrome 640/645 GPU - 3a01)
ZX-C+ (2016) => CentaurHauls Family 6, Model 15, Stepping 14 AKA ZhangJiang (KaiXian ZX-C+ & KaisHeng ZX-C+), e.g. C4701, FC-1081 - SM3, SM4, up to 4 MB shared L2 cache (2 x 2 MB), up to 8 cores (basically two quad-core CPUs in a multi-chip module, similar solution to that used in CNQ); compatible with ZX-100(S) (Chrome 320 GPU - 3a02)
ZX-D (2017) => CentaurHauls Family 6, Model 31, Stepping 12 AKA WuDaoKou (KaiXian KX-5000 & KaisHeng KH-20000), e.g. KX-U5680, KH-26800 - major redesign, full SoC, new uncore with northbridge moved on-die, new P2P high-speed interconnect crossbar that replaces FSB, PCIe 3.0, DDR4, integrated GPU (Chrome 860 GPU - 3a03), 4 MB L2 cache per cluster (up to 8 MB in total), up to 8 cores (2 quad-core clusters on a single die), 28nm HLMC & SMIC; compatible with ZX-200 (IOE chip for IO extensibility)
ZX-E (2018) => CentaurHauls Family 7, Model 11, Stepping 0 & 1 & 15 AKA LuJiaZui (KaiXian KX-6000 & KaisHeng KH-30000), e.g. KX-U6880A, KH-37800D - new integrated GPU (Chrome 960 GPU - 3a04), 16nm FF (FinFET) TSMC
ZX-F (2019) => CentaurHauls Family 6, Model 71 (KaiXian KX-7000 & KaisHeng KH-40000), Stepping 1 - PCIe 4.0 (planned), DDR5 (planned), 16 MB shared L3 cache (confirmed), 7nm TSMC (planned)

I haven't mentioned about the performance boost, but for sure it is significant.

Tralalak

2019-11-21, 21:20:50

01:10:39 – News: Centaur AI SoC

0vhenBFrEM0

source: https://pcper.com/2019/11/pc-perspective-podcast-565-3950x-reviews-google-stadia-thinkpad-p53/

Tralalak

2019-11-22, 14:09:14

VIA x86 is back
New Chip Technology for Video Analytic

https://scontent.fbts2-1.fna.fbcdn.net/v/t1.0-9/77215394_10156791053703927_1749705050017497088_o.jpg?_nc_cat=111&_nc_ohc=Ae-6Hqc-OJYAQnCXjVQelpoccLX0M05BP8y_JXZI8rFdnnxrKwHcXcILw&_nc_ht=scontent.fbts2-1.fna&oh=d137c1f08065eaa12b55ea11526402d6&oe=5E838670

source: https://www.facebook.com/viatechnologies/photos/ms.c.eJwzNDA0NTO3NDQwNTYzM7Y0MtczRBKxRBcxN7SwNDIHACHaCsY~-.bps.a.10156790880193927/10156791053693927/?type=3&theater

Tralalak

2019-12-03, 10:03:51

NEW
Centaur Adds AI to Server Processor
source: https://www.linleygroup.com/newsletters/newsletter_detail.php?num=6099

Loeschzwerg

2019-12-09, 06:54:19

Passmark Ergebnisse zum KX U6780A:
68535

Tralalak

2019-12-09, 17:47:29

But AMD FX-8300 is 8-core processor 3.3GHz (Maximum turbo frequency: 4.2GHz)

#1: 4200 MHz, 1.375V
#2: 3600 MHz, 1.2125V

Tralalak

2019-12-09, 17:52:58

Centaur Releases In-Depth Analysis from The Linley Group for World’s First x86 Processor with AI Coprocessor Technology
source: https://centtech.com/2019/12/09/centaur-releases-in-depth-analysis-from-the-linley-group-for-worlds-first-x86-processor-with-ai-coprocessor-technology/

_______

Centaur Unveils Its New Server-Class x86 Core: CNS; Adds AVX-512
source: https://fuse.wikichip.org/news/3099/centaur-unveils-its-new-server-class-x86-core-cns-adds-avx-512/

Tralalak

2019-12-09, 18:13:37

View Microprocessor Report. This report deep dives into the technical details.
source: https://centtech.com/wp-content/uploads/MPR_19_12_02_Centaur_Adds_AI_to_Server_Processor.pdf

Loeschzwerg

2019-12-09, 19:53:06

The core µarch and design choices sound quite nice. At least from a technological point of view Centaur seems to be back.

I extracted the DIE shot from the Linley report:
https://abload.de/img/centaur_cha_socqkjuv.jpg
The die requires 194mm2 in a TSMC
16FFC process. The Ncore DLA, which contains 16MB of
SRAM, is about half the size of the eight-CPU cluster.

Tralalak

2019-12-09, 21:34:12

The core µarch and design choices sound quite nice. At least from a technological point of view Centaur seems to be back.

Let's rock and roll :weg:

Tralalak

2019-12-10, 00:42:14

Centaur Releases In-Depth Analysis from The Linley Group for its NCORE-Equipped x86 Processor
source: https://www.techpowerup.com/261979/centaur-releases-in-depth-analysis-from-the-linley-group-for-its-ncore-equipped-x86-processor

Tralalak

2019-12-11, 13:11:14

Zhaoxin unveils next-generation general-purpose processor design specifications

Kaisheng 16nm KH-40000 up to 64 cores, DDR4, PCIe 3.0 (officially launched in 2021)

Kaixian 7nm KX-7000 SoC with DX12 IGPU

source: http://www.zhaoxin.com/InCenterContent.aspx?id=282

Tralalak

2019-12-14, 09:27:08

Centaur CNS x86 for VIA / Zhaoxin processors. Haswell or Zen level performance?
source: https://translate.google.com/translate?hl=sk&sl=cs&tl=en&u=https%3A%2F%2Fwww.cnews.cz%2Fx86-procesor-centaur-cha-cpu-jadro-architektura-cns-detaily-rozbor%2F

Leonidas

2019-12-15, 15:28:48

Wird ja immer besser. Und den Markt bekommen sie nun auch geschenkt ...

Benutzername

2019-12-15, 15:36:01

Wird ja immer besser. Und den Markt bekommen sie nun auch geschenkt ...

Trump hat bestimmt Anteile an VIA und Huawei. :biggrin:

------------

interessanter Artikel auf der tschechischen(?) website über den eneun Centaur und was über ARM Serverchips haben sie auch.

Loeschzwerg

2019-12-16, 07:25:22

Eine Tabelle mit Angaben zur TDP sämtlicher Zhaoxin x86 SKUs:
67577

Das ist die zugehörige Roadmap aus der Präsentation Mitte des Jahres:
68657

Es wird also auch bei Zhaoxin ein LGA Sockel eingeführt.

Tralalak

2019-12-16, 09:40:35

Just like the Centaur Technology website.

link: https://centtech.com/

konkretor

2019-12-16, 15:29:32

https://www.heise.de/newsticker/meldung/Zhaoxin-KX-7000-Chinesische-x86-Prozessoren-mit-PCIe-4-0-und-DDR5-RAM-4616376.html

robbitop

2019-12-16, 15:41:12

Sind nach wie vor alle uArchs (auch die für 7 nm angekündigte die mit Zen 2 gleichziehen soll) noch von Centaur?

Wer hätte das mal gedacht, dass Centaur nochmal so interessante/performante uArchs in den 2020er Jahren bringen wird. Wenn man da an den C7 oder die Winchipdesigns zurückdenkt ... gruselig langsam. Dafür aber natürlich sehr sparsam.

Tralalak

2019-12-16, 15:58:25

Sind nach wie vor alle uArchs (auch die für 7 nm angekündigte die mit Zen 2 gleichziehen soll) noch von Centaur?

It's very probably (only x86 part SoC). The graphic part DX12 is completed by Zhaoxin (S3G). But it's more complicated. Because they will already have x86 Centaur CNS and SHA SoC adopted. Essentially, the KX-7000 will be Zhaoxin SoC. of course with "DNA x86 Centaur".

Wer hätte das mal gedacht, dass Centaur nochmal so interessante/performante uArchs in den 2020er Jahren bringen wird. Wenn man da an den C7 oder die Winchipdesigns zurückdenkt ... gruselig langsam. Dafür aber natürlich sehr sparsam.

It's over 24 years of x86 experience. Those people there are extremely wise. Glenn Henry and others can easily be compared to Jim Keller and others...
The client has changed and has different requirements.
They wants to do servers and machine learning, AI so the achitecture adapts to it.

Tralalak

2019-12-19, 18:27:51

New "old" model
ZHAOXIN KaiXian ZX-C+ C4901H @ 2.33 GHz

vs

VIA QuadCore C4650 @ 2.00 GHz

Geekbench 3 Score
VIA: 918 (85,32%)
ZX: 1 076 (+17,21%)

Geekbench 3 Multicore Score
VIA: 2979 (88,95%)
ZX: 3 349 (+12,42%)

Frequency
VIA: 2 000 (85,73%)
ZX: 2 333 (+16,65%)

source: https://browser.geekbench.com/geekbench3/compare/8808003?baseline=8832144

Tralalak

2019-12-20, 12:00:50

ASRock CRB VT8614G TBD (CRB VT8614C) with ZHAOXIN KaiXian ZX-C+ C4901H@2.33GHz (4C 2.33GHz, 2MB L2)
source: https://ranker.sisoftware.co.uk/show_system.php?q=cea598ab93a198ab99bfd8e5c8f9dfad90a187eed3e3c5ad90a086fec3f3d5 b0d5e8d8fe8db088&l=en

rasti666

2019-12-20, 21:25:26

What is the "ASRock CRB VT8614G TBD" ? Is this a motherboard ? The last motherboard with a VIA processor on sale is EPIA-M920-20Q. Maybe after years - VIA / ZHAOXIN will be available for purchase for a computer PC ? Recent years are just tables and plans of the ZHAOXIN company ;). Maybe something will go to stores for sale ? There will be some motherboards to buy after years - does anyone know ?

Tralalak

2019-12-21, 20:09:44

Maybe ASRock will kill two flies with one shot: it will fill the lack of Big Brother processors and bring Zhaoxin to the world market. But it's just speculation ...

BlackArchon

2019-12-21, 22:36:35

Wait, this is finally a product you can actually buy? Wow, this is new. But when a big player like Asrock is involved, I'm quite optimistic that this isn't vaporware like before. I'm impressed! :)

Tralalak

2019-12-23, 12:56:35

2x Zhaoxin KaiSheng KH-37800D 2.7GHz (16C/16T)

http://www.zhaoxin.com/app_upload/images/2(33).png

source: http://www.zhaoxin.com/InCenterContent.aspx?id=284

Tralalak

2019-12-30, 10:29:08

CPU-Z 1.91 with support of new Centaur CPU
This version also adds the preliminary support of the new Centaur x86 CPU with the AI coprocessor that was announced recently.
source: https://www.cpuid.com/news.html

konkretor

2019-12-30, 13:06:15

https://www.golem.de/news/loongson-3a4000-3b4000-chinesische-cpu-schafft-doppelte-leistung-1912-145782.html

Bitte hier weiter Diskutieren

https://www.forum-3dcenter.org/vbulletin/showthread.php?p=12180781#post12180781

Benutzername

2019-12-30, 17:31:13

https://www.golem.de/news/loongson-3a4000-3b4000-chinesische-cpu-schafft-doppelte-leistung-1912-145782.html

Das ist aber ein MIPS, kein x86er und kein Zhaoxin. Aber der Fred ist jetzt wohl sowieso allgemein für alles aus China? :D

Beeindruckend, wie sie sich dahintergeklemmt haben und jetzt doppelt so viel Leistung rausbekommen und sind so etwa bei Bristol Ridge. Ein großer Sprung nach vorn.

Loeschzwerg

2019-12-30, 18:04:32

Sollte trotzdem hier weiter diskutiert werden:
https://www.forum-3dcenter.org/vbulletin/showthread.php?t=595289

Tralalak

2020-01-05, 17:25:27

Sharpe RG-CT7800-1000, 2.4L volume, KX-6780A 2.7GHz 8-Core processor, 8GB DDR4, 256GB SSD
Rejet Desktop Completes Unified Operating System UOS Adaptation: ReplaceS Office, Smooth Play 4K

https://www.smalltechnews.com/wp-content/uploads/2020/01/frc-efa1eadcb0ead00546d796719dadbdcf.jpg
source: https://www.smalltechnews.com/archives/54337

Tralalak

2020-01-17, 10:34:51

Posted byu/CentaurHauls95 Centaur Technologies 1 day ago
x86 CPU Design House Centaur Technology will host an AMA next week, Thursday Jan 23, 2020 @ 12:00pm-3:00pm CST

Info

Hi from Austin TX!

Come ask Centaur Technology about how they design x86 CPU's and their new AI co-processor. Centaur engineers will be answering the questions.

Learn more about the industry's first high-performance x86 SOC with server-class CPUs and integrated AI coprocessor technology.

Technical details disclosed in Linley Group's Microprocessor Report (https://centtech.com/wp-content/uploads/MPR_19_12_02_Centaur_Adds_AI_to_Server_Processor.pdf)

Centaur even have a documentary: Rise of the Centaur (https://www.amazon.com/Rise-Centaur-Glenn-Henry/dp/B01FSZU6FK)

source: https://www.reddit.com/r/hardware/comments/ep0lf4/x86_cpu_design_house_centaur_technology_will_host/

Tralalak

2020-01-18, 19:40:24

[PATCH] x86/speculation/spectre_v2: Exclude Zhaoxin CPUs from SPECTRE_V2
source: http://lkml.iu.edu/hypermail/linux/kernel/2001.1/08763.html

Tralalak

2020-01-22, 15:51:12

[Fun Technology] Domestic CPU draws i5-7400? Zhaoxin KX-U6780A first evaluation! The rise of China's core! ��

I0N3sqnKIxA

gravitationsfeld

2020-01-22, 17:20:12

Sorry, aber ich glaub kaum, dass hier viele Leute Chinesisch koennen.

Loeschzwerg

2020-01-22, 17:26:05

Enough performance for office tasks and also "OK" for gaming.

I've extracted some benchmark results:
69068 69069 69070 69071 69072

Loeschzwerg

2020-01-22, 17:27:24

69073

3DMark with a dedicated GTX 1660Ti:
69074 69075

Igzorn

2020-01-22, 19:37:28

Scheint ja gut voran zu gehen.

Irgendwie hofft man das es doch noch einen dritten Hersteller gibt der etwas Performantes anbieten kann und hoffentlich dann auch hier in Europa verkauft und nicht nur für China entwickelt.

Badesalz

2020-01-22, 22:14:30

Sowas wollten sie eigentlich garnicht machen ;) Sie haben aber Angst, daß der Ami auch noch damit droht, sie von x86 abzuklemmen. Und Mips ist das Fallback für alles was man so mit ARM macht.
Jetzt ist aber auch RISC-V soweit. Da sind sie richtig heiss hinterher
https://www.eetasia.com/news/article/Chinese-OEMs-Look-to-Risc-V-to-Replace-Arm-MCUs

Das gleiche machen sie auch mit GPUs.
https://www.notebookcheck.net/Chinese-company-allegedly-developing-a-GTX-1080-class-GPU-with-faster-clocks-PCIe-4-0-and-high-bandwidth-HBM-memory.431309.0.html

Damit bauen sie sich ggf. dann ihre Workstations, Office, AiO-PCs, Router, Switche usw. usw. WENN es sein muss.

Tralalak

2020-01-23, 21:36:26

Centaur Technology started answering your questions @ www.reddit.com

CPU-Z screenshot
https://i.imgur.com/UU96AOP.png

Windows 10 screenshot
https://i.imgur.com/EgPKzRK.png

source: https://www.reddit.com/r/hardware/comments/ep0lf4/x86_cpu_design_house_centaur_technology_will_host/

Benutzername

2020-01-24, 14:09:07

Interesting reddit thread, although some questions have been left unanswered.Too bad they do not name it IDT winchip 4 as one poster proposed. I hope Dr. Cutress keeps his word and makes a review on Anandtech of the new CPU. :)

AI co-processor. Makes me wonder if that could be misused for other computing tasks? Decoding videos? (de)Compression?

Loeschzwerg

2020-01-24, 19:41:29

A second article from David Schor on the new Centaur processor:
https://fuse.wikichip.org/news/3256/centaur-new-x86-server-processor-packs-an-ai-punch/

Tralalak

2020-01-25, 13:02:53

Tralalak

2020-01-30, 00:02:21

Zhaoxin’s x86-Compatible CPUs for DIY Enthusiasts Now Available
source: https://www.anandtech.com/show/15446/zhaoxins-x86compatible-cpus-for-diy-enthusiasts-now-available

y33H@

2020-01-30, 09:43:02

I'll try to order one or to buy one in Taipei at Computex :D

Loeschzwerg

2020-01-30, 09:47:40

Ich bin auch schon an einer Bestellung dran ^^

Benutzername

2020-01-30, 16:27:33

Expected prices?

and what pattern cooler does it use? intel holes? AM3? AM4? Its own propietary mounting pattern?

Loeschzwerg

2020-01-30, 17:10:47

Well... quite pricy. About 570€.

The current revision of the board does have a proprietary mounting for heatsinks, but it seems that a later revision is planed with a more common solution. This later revision will also feature a regular ATX power connector and a PCB in white.

SKYNET

2020-01-30, 22:29:58

wenn das ding noch langsamer wäre, würds rückwärts rechnen X-D

scx

2020-01-31, 00:05:32

scx

2020-01-31, 00:49:52

[Self-Category] Is a new generation of domestic CPUs and motherboards ready to enter the DIY market? How about the performance of Zhaoxin ’s new flagship KX-6780A and Xinyingjie C1888 motherboards?

https://youtu.be/2CtMI4RkTbo

3:17:

ZHAOXIN is developing independent graphics cards and KX7000 series CPUs.

Is that so? A new dGPU based on the S3G IP? Which one was last one? VIA eH1 (2011)? As far as I remember, we had:
- S3G Chrome 540 GTX (2009) with 256MB gDDR3
- S3G 5400E (2009) with 512MB gDDR3, for embedded systems
- S3G 5400E x2 (2010), basically dual-GPU based on 5400E
- VIA eH1 (2011), based on 5400E with 512MB DDR3
Anything else from the Chrome 500-series (except the 520 iGPU from VIA VN1000)?

They didn't make dGPU based on Chrome 600 series, right? There was only Chrome 640/645 iGPU in the VX11/VX11H/VX11PH chipset. And of course, Zhaoxin produced some derivatives, but that's it.
- Chrome 640 (C-640) iGPU & Chrome 645 (C-645) iGPU from the VX11/VX11H/VX11PH chipset - Vendor Id: 1106 ("VIA Technologies, Inc."), Device Id: 3a01 ("VX11 Graphics [Chrome 645/640]")
- Chrome 320 (C-320) iGPU from the ZX-100S chipset - Vendor Id: 1d17 ("Zhaoxin"), Device Id: 3a02 ("ZX-100 C-320 GPU")
- Chrome 860 (C-860) iGPU from the ZX-D/KX-5000/KH-20000 (codename "WuDaoKou") SoC - Vendor Id: 1d17 ("Zhaoxin"), Device Id: 3a03 ("ZX-D C-860 GPU")
- Chrome 960 (C-960) iGPU from the ZX-E/KX-6000/KH-30000 (codename "LuJiaZui") SoC - Vendor Id: 1d17 ("Zhaoxin"), Device Id: 3a04 ("ZX-E C-960 GPU")
- Chrome iGPU from the ZX-F/KX-7000/KH-40000 SoC
However, there was no significant increase in performance, if any.
I thought they just gave up, focusing on the JM family (JM5400/JM7200/JM9231/JM9271), which was based on reverse engineered Radeon. Do you think that there is still a real chance to produce a new dGPU based on S3G technology?

Loeschzwerg

2020-01-31, 06:29:49

wenn das ding noch langsamer wäre, würds rückwärts rechnen X-D

Troll dich wieder...

You are wrong! It is only ~561 EUR (620 USD)! :wink:
https://twitter.com/IanCutress/status/1222881378641108993

:D I hope I can get mine before Ian.

Regarding S3G GPUs, you listed all the latest known products.

Zoetrope -> AlphaChrome
Columbia -> DeltaChrome, GammaChrome and Chrome S20 series
Destination -> Chrome 400 and 500 series
Excalibur -> Chrome 600 series

C-960 is still based on the Excalibur architecture, only with some minor tweaks, mainly on media/video support.

scx

2020-01-31, 08:03:45

:D I hope I can get mine before Ian.

I remember that getting 540 GTX wasn't a problem, but VIA eH1 was almost impossible to get.
A similar situation was with VIA motherboards/PCs: there was a lot of hardware with C3 (mini-ITX boards, thin clients, etc.), a little bit with C7, some with Nano, getting latest products was almost a miracle.
Anyway, good luck with it.

Regarding S3G GPUs, you listed all the latest known products.

Zoetrope -> AlphaChrome
Columbia -> DeltaChrome, GammaChrome and Chrome S20 series
Destination -> Chrome 400 and 500 series
Excalibur -> Chrome 600 series

C-960 is still based on the Excalibur architecture, only with some minor tweaks, mainly on media/video support.

Wow, until now, I didn't know that they have codenames!
https://www.phoronix.com/forums/forum/linux-graphics-x-org-drivers/x-org-drm/1026981-it-s-still-going-to-be-tough-getting-the-openchrome-via-kms-driver-in-the-linux-kernel?p=1027179#post1027179

So, all Unichrome graphics belong to Zoetrope, and Chrome9 GPUs belong to Columbia, right? BTW, I thought Savage XP/AlphaChrome was never released.

Do you know anything about Linux driver for Excalibur? For sure, "via_chrome9_dri.so"/"via_drv.so" can't handle it. And I am afraid that the old "s3g_dri.so"/"s3g_drv.so" driver supports only Destination. OpenChrome barely handles 2D mode for Chrome9 HD from VX900. Both VN1000 and VX11 are completely unsupported by it.
I know that Zhaoxin has published some drivers for Windows, but I have not heard of Linux drivers from them that would be publicly available. But they must have something.
https://www.catalog.update.microsoft.com/Search.aspx?q=Zhaoxin

Anyway, maybe Destination and Excalibur were fine in 2008-2012, but today they are a little bit outdated.
I wonder if Zhaoxin has some serious plans for the next generation.

SKYNET

2020-01-31, 12:10:49

Troll dich wieder...
.

ist halt so, das ding ist langsamer als die A10 CPU in meinem uralt laptop X-D

und das ding ist zu langsam um zu surfen... :ulol:

Loeschzwerg

2020-01-31, 14:30:41

Schau, wenn du dich nicht für alternative Technik interessiert ist das deine Sache, aber erspare dem Rest hier bitte deine dämlichen Kommentare.

SKYNET

2020-01-31, 14:48:35

Schau, wenn du dich nicht für alternative Technik interessiert ist das deine Sache, aber erspare dem Rest hier bitte deine dämlichen Kommentare.

das problem ist, dass das teil auch "für gaming" beworben wird, das teil hat aber nichtmal genug leistung um flüssig zu surfen... oder anders gesagt: das teil hängt 10 jahre hinterher.

Screemer

2020-01-31, 15:00:22

das ist für domestic china. du scheinst nicht wirklich ne ahnung zu haben was und wie die da daddeln. lass es einfach. der thread ist für dich einfach uninteressant.

Benutzername

2020-01-31, 16:03:32

das ist für domestic china. du scheinst nicht wirklich ne ahnung zu haben was und wie die da daddeln. lass es einfach. der thread ist für dich einfach uninteressant.

Ja. Bis auf ein paar reiche Kids der richtig Reichen, daddelt man in der VR China auf richtig alten Möhren und die Leute kaufen sogar recyclte Server CPUs und Laptop CPUs und stecken die in diese Wundertütenboards von Aliexpress. Spielen tun die dann ja auch sowas wie World of Warcraft und dessen asiastische NAchahmer und Visual Novels, also schnell genug sind die Dinger. Und surfen und paar Browserspiele sollten auch gehen. Surfen kann ich sogar noch brauchbar auf meinem alten Pentium-M Laptop. Kommt halt darauf an, was man erwartet und welche Seiten man anguckt.

---------------------

Well... quite pricy. About 570€.

Ouch. That requires too much enthusiasm for me to buy. :(

The current revision of the board does have a proprietary mounting for heatsinks, but it seems that a later revision is planed with a more common solution. This later revision will also feature a regular ATX power connector and a PCB in white.

The CPU is at a snails pace, but they make a fancy PCBs. Does it come with RGB? ;)

SKYNET

2020-01-31, 16:40:39

das ist für domestic china. du scheinst nicht wirklich ne ahnung zu haben was und wie die da daddeln. lass es einfach. der thread ist für dich einfach uninteressant.

in china werden die webseiten sicherlich nicht viel anders sein als hier(macnhmal glaube ich, leute denken das WWW in china aisht aus wie hier zu ISDN zeiten), entsprechend wird das ding auch da langsam laufen... weil selbst nen Q6600 rennt dem ding gnadenlos davon, und die bekommst im gebrauchtmarkt sicherlich auch in china billiger als die krücke von VIA.

Benutzername

2020-01-31, 16:44:22

in china werden die webseiten sicherlich nicht viel anders sein als hier(macnhmal glaube ich, leute denken das WWW in china aisht aus wie hier zu ISDN zeiten), entsprechend wird das ding auch da langsam laufen... weil selbst nen Q6600 rennt dem ding gnadenlos davon, und die bekommst im gebrauchtmarkt sicherlich auch in china billiger als die krücke von VIA.

Daß der q6600 schneller ist stimmt auf jeden Fall. Aber das Teil ist sowieso mehr ein erster Testlauf und für Programmierer zum üben für die nachfolgenden Chips. Für Arbeitseinsatz noch nicht wirklich geeignet. Vielleicht für den eigenen mailserver zu Hause.

Und Regierungskunden nicht vergessen. China hat berechtigterweise Angst vor Backdoors der Amerikaner. Deswegen ja chinesische Hardware.

Loeschzwerg

2020-01-31, 17:08:28

I remember that getting 540 GTX wasn't a problem, but VIA eH1 was almost impossible to get.

I do own a GTX 540 and eH1. The eH1 was easy to obtain, got it when it was still available at some special retailers for embedded hardware. The GTX 540 was a tough one and also a board with VN1000 chipset.

Do you know anything about Linux driver for Excalibur? For sure, "via_chrome9_dri.so"/"via_drv.so" can't handle it. And I am afraid that the old "s3g_dri.so"/"s3g_drv.so" driver supports only Destination. OpenChrome barely handles 2D mode for Chrome9 HD from VX900. Both VN1000 and VX11 are completely unsupported by it.

I don't know of any freely available Linux drivers, that support VN1000 and newer GPU IP. Maybe there are some drivers integrated in Chinese Linux distributions like Deepin, Kylin or UOS.

Anyway, maybe Destination and Excalibur were fine in 2008-2012, but today they are a little bit outdated.
I wonder if Zhaoxin has some serious plans for the next generation.

At least the GPU of the KX-7000 SoC will get a feature update and support DX12. No information on performance so far. I think they will mainly focus on the CPU part.

---

das problem ist, dass das teil auch "für gaming" beworben wird, das teil hat aber nichtmal genug leistung um flüssig zu surfen...

Alles was nicht Highend ist oder eines deiner Systeme schlägt passt dir nicht in den Kram und du MUSST das irgendwie bei jeder Gelegenheit heraushängen lassen. Ich sehe hier nur ein Problem ;)

robbitop

2020-01-31, 17:09:38

Sollte der neue Core nicht ~Zen IPC haben?

Loeschzwerg

2020-01-31, 17:11:39

Erst der KX-7000, falls du den meinst. KX-6000 basiert immer noch irgendwie auf Isaiah, natürlich mit Tweaks hier und da.

Benutzername

2020-01-31, 17:16:13

Erst der KX-7000, falls du den meinst. KX-6000 basiert immer noch irgendwie auf Isaiah, natürlich mit Tweaks hier und da.

Das schreit nach einem Duell mit den Hygon Dhyana (oder wie die lizenzierten Zen hießen).

SKYNET

2020-01-31, 21:37:51

Daß der q6600 schneller ist stimmt auf jeden Fall. Aber das Teil ist sowieso mehr ein erster Testlauf und für Programmierer zum üben für die nachfolgenden Chips. Für Arbeitseinsatz noch nicht wirklich geeignet. Vielleicht für den eigenen mailserver zu Hause.

Und Regierungskunden nicht vergessen. China hat berechtigterweise Angst vor Backdoors der Amerikaner. Deswegen ja chinesische Hardware.

naja, aber wenn du dann nen servercluster von den dingern brauchst um an die leistung von einem 3950X ranzukommen, hilft einem das auch nicht weiter :freak:

das teil ist einfach unbrauchbar nach dem aktuellen stand, und wie du schon sagtest, für mehr als nen mailserver eher ungeeignet... und selbst da, gibts sachen die VIEL leistungsfähiger und trotzdem sparsamer sind ;)

Alles was nicht Highend ist oder eines deiner Systeme schlägt passt dir nicht in den Kram und du MUSST das irgendwie bei jeder Gelegenheit heraushängen lassen. Ich sehe hier nur ein Problem ;)

nö, es muss einen praktischen nutzen erfüllen, und das kann das teil leider nach heutigem standard nicht...

Screemer

2020-01-31, 22:46:35

@skynet: reichts jetzt langsam? btw. schreib ich das grad auf nem pentium n3540 auf dem nebenher noch ein 1080p h264 stream, ein bisschen musik und knapp 20 tabs in ff laufen.

robbitop

2020-02-01, 10:45:02

scx

2020-02-01, 23:25:42

ZX-D GPU: C-860 AKA CHX001

Release notes

Supported chips: ZX-D B0&B1 Chip
Supported OS: 方德桌面操作系统 3.0_64bit - http://www.nfschina.com/html/product/center/base/250.html
Supported Linux: Linux Kernel 4.9
Supported Board: HX001EH & HX001EB & HX001EK
Driver version: 19.00.46 & 21.00.13

Frequency

ECLK - 350Mhz
ICLK - 700Mhz
VPP CLCK - 350MHz
VCP CLCK - 350MHz

Display Support

Supported Device: CRT, DP1/HDMI1/DVI1, DP2/HDMI2/DVI2

Not snoop only and FB size=256M/512M with dual channel memory, can support up to three multi-view.

Single view
1. Integrated DisplayPort1/DisplayPort2: 3840x2160@30Hz, 1920x1080@60Hz
2. Integrated CRT: 2048x1536@60Hz
3. Integrated HDMI1/HDMI2: 3840x2160@30Hz, 1920x1080@60Hz
4. Integrated DVI (through Integrated HDMI1/HDMI2): 1920x1080@60Hz

Multi-view
17. Integrated CRT + Integrated DisplayPort1/HDMI1/DVI1 + Integrated DisplayPort2/HDMI2/DVI2: 1920x1080@60Hz

Drivers

GLX

Version: 1.4

/usr/lib/x86_64-linux-gnu/xorg_extra/libglx.so

OpenGL

Version: 3.2

/usr/lib/x86_64-linux-gnu/zx/libGL.so.1.2
/usr/lib/x86_64-linux-gnu/dri/zx_dri.so
/usr/lib/x86_64-linux-gnu/xorg_extra/libglx.so

driver 19.00.47c:

OpenGL vendor string: Shanghai Zhaoxin Semiconductor Co., Ltd
OpenGL core profile version string: 3.0 19.00.47c
OpenGL core profile version string: 3.1 19.00.47c
OpenGL core profile version string: 3.2 19.00.47c

driver 21.00.13:

OpenGL vendor string: Shanghai Zhaoxin Semiconductor Co., Ltd
OpenGL core profile version string: 3.0 21.00.13
OpenGL core profile version string: 3.1 21.00.13
OpenGL core profile version string: 3.2 21.00.13

OpenGL ES

Version: 2.0

/usr/lib/x86_64-linux-gnu/dri/zx_dri.so

driver 19.00.47c:

OpenGL ES profile version string: OpenGL ES-CM 1.1
OpenGL ES profile version string: OpenGL ES 2.0 19.00.47c

driver 21.00.13:

OpenGL ES profile version string: OpenGL ES-CM 1.1
OpenGL ES profile version string: OpenGL ES 2.0 21.00.13

X.Org 2D driver

/usr/lib/xorg/modules/drivers/zx_drv.so

DRI driver

/usr/lib/x86_64-linux-gnu/dri/zx_dri.so

VDPAU

/usr/lib/x86_64-linux-gnu/zx/libvdpau_zx.so

VA-API

/usr/lib/x86_64-linux-gnu/dri/zx_drv_video.so

OpenCL

Version: 1.1

/etc/OpenCL/vendors/s3.icd
/usr/bin/s3asm
/usr/bin/s3eopencc1
/usr/bin/s3exgcc
/usr/bin/s3oclcompiler
/usr/lib/x86_64-linux-gnu/s3g_driocl.so
/usr/lib/x86_64-linux-gnu/s3oclcompiler.so
/usr/lib/x86_64-linux-gnu/zx/libOpenCL.so
/usr/lib/x86_64-linux-gnu/zx/libOpenCL.1.1.so

Decoding Capabilities

1
Format: H.265 HEVC
Performance: 1920x1080p30@70Mbps; 1920x1080p120@25Mbps; 4096x2160p30@70Mbps
Profile: MP

2
Format: H.264
Performance: AVC 1920x1080p30@70Mbps; 1920x1080p120@25Mbps; 3840x2160p30@50Mbps
Profile: HP

3
Format: VP8
Performance: 1920x1080p30@10Mbps
Profile: /

6
Format: M-JPEG
Performance: 1920x1080p30@80Mbps (M-JPEG); Baseline 8kx4k, 130Mpps (JPEG)
Profile: Baseline

7
Format: VC-1 / WMV9
Performance: 1920x1080p30@72Mbps
Profile: Level:AP@L3.0

8
Format: AVS
Performance: 6.0.3.08.60(base),1920x1152p60@65Mbps; 6.0.0.08.60(broadcast),1920x1152p60@25Mbps
Profile: /

Note:
Smplayer hardware decode support: H265, H264 and VC1

Encoding Capabilities

1
Format: H.265 HEVC
Performance: 1920x1080p50
Profile: MP/I, P frames/CABAC

2
Format: H.264 AVC
Performance: 1920x1080p50
Profile: HP/I, P frames/CAVLC

Files

Linux driver files

/etc/modprobe.d/zx_gfx.conf
/etc/X11/xorg.conf.d/zx_xorg.conf
/usr/lib/x86_64-linux-gnu/dri/ZXEApp.cfg
/usr/lib/x86_64-linux-gnu/dri/zx_dri.so
/usr/lib/x86_64-linux-gnu/dri/zx_drv_video.so
/usr/lib/x86_64-linux-gnu/udev/rules.d/90-pulseaudio.rules
/usr/lib/x86_64-linux-gnu/xorg_extra/libglx.so
/usr/lib/x86_64-linux-gnu/zx/libGL.so.1.2
/usr/lib/x86_64-linux-gnu/zx/libvdpau_zx.so
/usr/lib/xorg/modules/drivers/zx_drv.so

Added in 21.00.13:

/usr/lib/x86_64-linux-gnu/libkeinterface_zx.so.0.0.0

Notes:

#1. 2d driver: zx_drv.so /usr/lib/xorg/modules/drivers
#2. for old private driver architecture, we need provide glx: libglx.so /usr/lib/xorg/modules/externsions
#3. 3d client wrap driver: libGL.so.1.2 /usr/lib/$DRIVER_ARCH-linux-gnu/zx
#4. 3d opengl driver: zx_dri.so /usr/lib/$DRIVER_ARCH-linux-gnu/dri
#5. app patch: ZXEApp.cfg /usr/lib/$DRIVER_ARCH-linux-gnu/dri
#6. video vdpau driver: libvdpau_zx.so /usr/lib/$DRIVER_ARCH-linux-gnu/zx
#7. video va driver: zx_drv_video.so /usr/lib/$DRIVER_ARCH-linux-gnu/dri
#8. x11 config: zx_xorg.conf /usr/share/X11/xorg.conf.d

Extra drivers

OpenCL:

/etc/OpenCL/vendors/s3.icd
/usr/lib/x86_64-linux-gnu/s3g_driocl.so
/usr/lib/x86_64-linux-gnu/zx/libOpenCL.so
/usr/lib/x86_64-linux-gnu/zx/libOpenCL.1.1.so

OpenCL compiler:

/usr/bin/s3asm
/usr/bin/s3eopencc1
/usr/bin/s3exgcc
/usr/bin/s3oclcompiler
/usr/lib/x86_64-linux-gnu/s3oclcompiler.so

Kernel module sources

/usr/src/zhaoxin_kernel/.built-in_64.o.cmd
/usr/src/zhaoxin_kernel/built-in_64.o
/usr/src/zhaoxin_kernel/core_errno.h
/usr/src/zhaoxin_kernel/core_module.c
/usr/src/zhaoxin_kernel/kernel_import.h
/usr/src/zhaoxin_kernel/kernel_interface.h
/usr/src/zhaoxin_kernel/Makefile
/usr/src/zhaoxin_kernel/os_interface.c
/usr/src/zhaoxin_kernel/os_interface.h
/usr/src/zhaoxin_kernel/os_shared.c
/usr/src/zhaoxin_kernel/s3.h
/usr/src/zhaoxin_kernel/s3gfb.h
/usr/src/zhaoxin_kernel/s3gfx_trace.h
/usr/src/zhaoxin_kernel/s3g_buffer_object.h
/usr/src/zhaoxin_kernel/s3g_cec.h
/usr/src/zhaoxin_kernel/s3g_debugfs.h
/usr/src/zhaoxin_kernel/s3g_def.h
/usr/src/zhaoxin_kernel/s3g_driver.h
/usr/src/zhaoxin_kernel/s3g_ioctl.h
/usr/src/zhaoxin_kernel/s3g_ioctl_compat.h
/usr/src/zhaoxin_kernel/s3g_led.h
/usr/src/zhaoxin_kernel/s3g_screen.h
/usr/src/zhaoxin_kernel/s3g_sync.h
/usr/src/zhaoxin_kernel/s3g_types.h
/usr/src/zhaoxin_kernel/s3g_version.h
/usr/src/zhaoxin_kernel/zx1.c
/usr/src/zhaoxin_kernel/zxfb.c
/usr/src/zhaoxin_kernel/zxfb_ops.c
/usr/src/zhaoxin_kernel/zx_buffer_object.c
/usr/src/zhaoxin_kernel/zx_cec.c
/usr/src/zhaoxin_kernel/zx_debugfs.c
/usr/src/zhaoxin_kernel/zx_driver.c
/usr/src/zhaoxin_kernel/zx_ioctl.c
/usr/src/zhaoxin_kernel/zx_ioctl_compat.c
/usr/src/zhaoxin_kernel/zx_keymap.c
/usr/src/zhaoxin_kernel/zx_led.c
/usr/src/zhaoxin_kernel/zx_pcie.c
/usr/src/zhaoxin_kernel/zx_screen.c
/usr/src/zhaoxin_kernel/zx_sync.c

Copyright

* Copyright 1998-2014 VIA Technologies, Inc. All Rights Reserved.
* Copyright 2001-2014 S3 Graphics, Inc. All Rights Reserved.
* Copyright 2013-2014 Shanghai Zhaoxin Semiconductor Co., Ltd. All Rights Reserved.

As you can see, this in not an open source driver, but rather an open source glue around the binary blob.

/usr/src/zhaoxin_kernel/built-in_64.o: ELF 64-bit LSB relocatable, x86-64, version 1 (SYSV), not stripped

Compiled kernel modules:

/lib/modules/${KERNEL_VER}/kernel/drivers/gpu/drm/zx/zx.ko
/lib/modules/${KERNEL_VER}/kernel/drivers/gpu/drm/zx/zx_core.ko

modprobe:

/etc/modprobe.d/zx_gfx.conf

Supported devices - driver version 19.00.46:
* Chrome 640 (C-640) iGPU & Chrome 645 (C-645) iGPU from the VX11/VX11H/VX11PH chipset - Vendor Id: 1106 ("VIA Technologies, Inc."), Device Id: 3a01 ("VX11 Graphics [Chrome 645/640]")
* Chrome 860 (C-860) iGPU from the ZX-D/KX-5000/KH-20000 (codename "WuDaoKou") SoC - Vendor Id: 1d17 ("Zhaoxin"), Device Id: 3a03 ("ZX-D C-860 GPU")

static struct pci_device_id pciidlist[] =
{
{0x1106, 0x3A01, PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0},
{0x1106, 0x3A03, PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0}, //chx001 qt
{0x1d17, 0x3A03, PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0}, //chx001
{0, 0, 0}
};

Supported devices - driver version 21.00.13:
+ Chrome 960 (C-960) iGPU from the ZX-E/KX-6000/KH-30000 (codename "LuJiaZui") SoC - Vendor Id: 1d17 ("Zhaoxin"), Device Id: 3a04 ("ZX-E C-960 GPU")

CHX002

Additional software

MPlayer and SMPlayer:

/usr/local/bin/smplayer_1490_x64
/usr/local/bin/vaapi_mplayer_x64
/usr/local/etc/mplayer/config_x64
/usr/local/etc/smplayer/smplayer.ini_x64

woodsdog

2020-02-02, 08:50:52

Ich sehe das auch so. Ein dritter Mitspieler im x86 bringt Potenzial und ist interessant. Dass da nicht sofort gleichwertige Ergebnisse kommen sollte klar sein. Ein Bashing bringt uns nicht weiter und stört die Diskussion. Ich plädiere für konstruktive Beiträge oder ein Fernbleiben.
Dabei bin ich nicht gegen Kontroversen. Da es offensichtlich ist, dass diese CPU nicht gegen die aktuellen Kontrahenten ankommt, ist keine Basis für eine sinnvolle Kontroverse vorhanden.

Hier im Thread kommt aber auch jedes Mal gleich der "bleib halt weg"-Beißreflex wenn man nicht super positiv über die Via Dinger postet. Das ist dann auch immer gleich automatisch Bashing, trolling und "nicht konstruktiv".

Ja - es ist interessant... aber es bleibt doch trotzdem ein Neues Produkt, welches Mühe hat mit 10 Jahre alter "westlicher Technik" mitzuhalten. Und das seit immer. „nicht sofort gleichwertig“ ist maßlos geschönt.
Warum man das im ggs. zu allen anderen Hardware Threads im 3DC ausgerechnet hier nicht schreiben können soll, weil es ja so offensichtlich ist und die "Diskussion stört" ist mir nicht klar... was diskutiert ihr eigentlich? Sehe wenig Diskussion hier.

Dieses Produkt ist DOA sofern man kein Chinesischer ITler ist, der für die Regierung Hardware ranschaffen muss. Das ist der einzige Grund warum die noch existieren. Für den westlichen Markt abseits der Kuriosität weder interessant noch konkurrenzfähig.
Die wollen 560 Eur für ein Board+CPU (+Ram?). Ein kleiner 3600 dreht Kreise um das Ding und ist billiger inkl. Board… und das muss ja wohl gesagt werden dürfen.

robbitop

2020-02-02, 10:37:28

Ja es ist langsam und ja P/L stimmen nicht. Wurde nun zum xten Mal gepostet. Noch dazu herrscht darüber offenbar allgemeiner Konsens. Ist also redundant und trägt somit nichts Konstrktives zur Diskussion bei. Hat mit Beißreflex nichts zu tun. Das Produkt existiert und es bekommt einen Nachfolgerkern mit Zen IPC. Wenn es für Spezialmärkte ist, kann das interessant sein.

Jetzt bitte BTT. Rest kann per PN geklärt werden.

scx

2020-02-02, 12:24:02

Similarities between ZX-F [1] and CHA/CNS [2].
- CPUID: [00040671] => CentaurHauls Family 6 Model 71 Stepping 1 (hint: 71 = 47h)
- 16 MB L3 cache
- 8 cores on single die area
What is more, we can see "CHA001" in the Zhaoxin motherboard name ("Shanghai Zhaoxin Semiconductor Co., Ltd. CHA001 MB"), while Centaur Technology uses "CHA" as the SoC name. Moreover, both processors are supposed to use LGA.

Differences:
- ZX-F will be produced in 7 nm TSMC, while CHA/CNS in 16 nm TSMC
- ZX-F will have a DX12-capable GPU, while CHA/CNS has no iGPU at all

Hard to say:
- NCORE (AI processor) is probably a Centaur-thing, because Zhaoxin didn't say a word about it
- Centaur Technology boasted of support for AVX-512, while Zhaoxin is silent on this subject
- Zhaoxin processors support SM3/SM4, while Centaur Technology doesn't mention it
- Centaur Technology wants to stick with PCI Express 3.0, while Zhaoxin probably wants to use a more recent generation
- Centaur Technology wants to use DDR4, while Zhaoxin probably wants to switch to DDR5 when it becomes available

By the way:
- KH-40000 will continue to utilize DDR4 memory and PCIe Gen 3 as well as 16 nm process

[1] - https://browser.geekbench.com/v5/cpu/526995
[2] - https://i.imgur.com/UU96AOP.png

My question is: Is it possible that ZX-F and CHA/CNS are at least similar? Until now I thought that Zhaoxin processors are developed independently from VIA/Centaur, or at least that VIA has no access to what Zhaoxin has developed. But if these chips have something in common, then Zhaoxin's IP must be available to VIA, at least in a limited way.
Even if these cores are not the same, we may have a similar situation as for CNR (C4650) and ZX-C+ (C4710): almost the same design, the same CPUID, and only small differences in terms of supported instructions (ZX-C+ has added support for SM3/SM4).
What do you thing about it?

Windi

2020-02-02, 13:08:16

Sobald man Zen1 Leistung erreicht wird es doch interessant.
Dann reicht der Prozessor für vieles aus und man kann das Teil in vielen LowEnd Kisten verbauen.

Es ist natürlich wichtig, das sie möglichst günstig werden, aber dafür kann man sorgen.
So ein Silizium-Chip kostet von TSMC wohl kaum 30 Dollar. Das teure sind die Entwicklungskosten.
Wenn nun die chinesische Regierung dafür sorgt, das die Firma genug Geld für die Entwicklung bekommt, dann kann man die Prozessoren für den freien Markt günstig anbieten.
Eine Möglichkeit wäre es wie SpaceX zu machen. Für die Regierung verkauft man Produkte für den doppelten oder dreifachen Preis und erwirtschaftet somit große Überschüsse mit denen man die restlichen Produkte subventionieren kann. (Begründung: Besondere Sicherheitsfeature für Behörden-PCs, die den doppelten Preis rechtfertigen)

Es ist alles eine Frage des Willens der Regierung. Falls es China wirklich will, kann man leicht mehr Entwicklungsbudget wie AMD bekommen. Und dann ist es nur eine Frage der Zeit, bis man zur Konkurrenz aufholt.

Tralalak

2020-02-03, 17:27:50

https://i.imgur.com/KKozKVU.jpg

Tralalak

2020-02-18, 12:31:01

Centaur CHA x86 AI CPU pictured
source: https://www.semiaccurate.com/2020/02/17/centaur-cha-x86-ai-cpu-pictured/

Tralalak

2020-02-19, 13:43:22

VIA AMOS-3006 Ruggedized System with high performance

https://scontent.fbts2-1.fna.fbcdn.net/v/t1.0-9/p960x960/87055583_10157040926298927_663687908837818368_o.jpg?_nc_cat=110&_nc_ohc=x4aVnth-fhcAX_AyZmn&_nc_ht=scontent.fbts2-1.fna&_nc_tp=6&oh=0c44b56962fdf87fb32bbbc79887ebb4&oe=5EFCA61A

source: https://www.viatech.com/en/2020/02/notes-from-the-field-via-amos-3006/

Tralalak

2020-02-21, 11:12:17

Stanford Seminar - Centaur Technology's Deep learning Coprocessor

5Z7cmyYakAw

Tralalak

2020-02-25, 22:34:24

This mini PC is powered by a Zhaoxin x86 processor
source: https://liliputing.com/2020/02/this-mini-pc-is-powered-by-a-zhaoxin-x86-processor.html

Loeschzwerg

2020-02-27, 22:50:01

https://i.imgur.com/KKozKVU.jpg

Without censored information on ZX-G:
https://www.forum-3dcenter.org/vbulletin/attachment.php?attachmentid=69318

16 cores and 64 PCIe 4.0 lanes @ 7nm.

konkretor

2020-02-28, 10:06:59

https://www.anandtech.com/show/15493/hygon-dhyana-reviewed-chinese-x86-cpus-amd

Loeschzwerg

2020-02-28, 10:28:31

Hatten wir schon:
https://www.forum-3dcenter.org/vbulletin/showthread.php?p=12229086#post12229086
https://www.3dcenter.org/news/hardware-und-nachrichten-links-des-27-februar-2020

scx

2020-03-01, 08:00:38

Without censored information on ZX-G:
https://www.forum-3dcenter.org/vbulletin/attachment.php?attachmentid=69318

16 cores and 64 PCIe 4.0 lanes @ 7nm.
It seems that ZX-F is going to be extremely similar to CHA/CNS:
- the same CPU ID: [00040671] => CentaurHauls Family 6 Model 71 Stepping 1 (hint: 71 = 47h)
- 8 cores on single die area
- 16 MiB shared L3 cache (2 MiB per core, shared across all cores)
- DDR4 (4 channels)
- PCI-e 3.0 (44 lanes)
- TSMC 16 nm process (16FFC)
- 2.5 GHz core frequency

There is still no confirmation of AVX-512 support in ZX-F, but it seems almost certain.
The main difference is probably the NCORE AI coprocessor (including NDU and NPU) in CHA and a Chrome-based GPU in ZX-F.
Zhaoxin shifted their plans to use PCIe 4.0 and TSMC 7nm process to ZX-G in 2021. What's more, contrary to the original assumption, they don't intend to use DDR5 in the near future.

Tralalak

2020-03-20, 09:22:44

Lenovo Computers with Zhaoxin Technology

http://www.zhaoxin.com/Upload/202003161613460458.jpg
source: http://www.zhaoxin.com/InCenterContent.aspx?id=299

Tralalak

2020-03-23, 22:47:15

AIOSTAR Computers with Zhaoxin Technology
source: https://translate.google.com/translate?hl=sk&sl=zh-CN&tl=en&u=http%3A%2F%2Fwww.ds-360.com%2Fnews%2F2020%2F3%2F2020_1_35988.htm

Tralalak

2020-03-26, 11:30:37

Tongfang Computers with Zhaoxin Technology
source: https://translate.google.com/translate?hl=sk&sl=zh-CN&tl=en&u=http%3A%2F%2Fwww.zhaoxin.com%2FInCenterContent.aspx%3Fid%3D300

Tralalak

2020-03-26, 17:56:35

Folding@home for Fighting Covid-19 with VIA QuadCore 2.0GHz (CPU) and NVIDIA GeForce GTX1050Ti (GPU)

4xnNKUZYr58

Benutzername

2020-03-26, 18:41:38

Folding@home for Fighting Covid-19 with VIA QuadCore 2.0GHz (CPU) and NVIDIA GeForce GTX1050Ti (GPU)

https://youtu.be/4xnNKUZYr58

Nice endurance test for the Zhaoxin CPU. Inform us, if it bursts into flames. ;)

Tralalak

2020-03-26, 19:06:09

Nice endurance test for the Zhaoxin CPU. Inform us, if it bursts into flames. ;)
It is a high quality VIA EPIA M920-20Q embedded board for 24/7/365 operation. :weg:
So don't worry :comfort:

Benutzername

2020-03-30, 17:14:41

It is a high quality VIA EPIA M920-20Q embedded board for 24/7/365 operation. :weg:
So don't worry :comfort:

Well, yes, it is an industrial controller computer and supposedly built to last, but is it going to take 24/7/365 100% workload?

Loeschzwerg

2020-03-31, 17:40:30

Die Centaur CPUs durchlaufen entsprechende Last-/Haltbarkeits-Prozesse, wie auch bei AMD/Intel. Wird auch in der Centaur Doku gezeigt.

davidzo

2020-04-01, 22:24:55

Well, yes, it is an industrial controller computer and supposedly built to last, but is it going to take 24/7/365 100% workload?

Jedenfalls eine bessere Garantie als der 9900KS ;D

Tralalak

2020-04-02, 11:29:37

Haier Computers with Zhaoxin Technology

http://www.zhaoxin.com/app_upload/images/4(32).jpg

source: www.zhaoxin.com (https://translate.google.com/translate?hl=sk&sl=zh-CN&tl=en&u=http%3A%2F%2Fwww.zhaoxin.com%2FInCenterContent.aspx%3Fid%3D302)

BlackArchon

2020-04-02, 13:18:22

Gamersnexus hat einen Fertig-PC mit Zhaoxin-Prozessor getestet: https://www.youtube.com/watch?v=-DanhnASClQ

Loeschzwerg

2020-04-02, 14:00:13

Kein gutes Video von GN. Die getestete Hardware basiert komplett auf alter VIA Technologie, die man regulär in EU/US kaufen kann (VIA EPIA M920-20Q). Seit 2015 btw...

y33H@

2020-04-02, 15:31:05

Der ZX-C+ bzw der C4701 ist doch ein alter 28 nm 4C mit 2 GHz und kein wirklich neues Design von Zhaoxin?

http://www.zhaoxin.com/ZXC.aspx?seriesid=5

scx

2020-04-02, 15:45:30

Der ZX-C+ bzw der C4701 ist doch ein alter 28 nm 4C mit 2 GHz und kein wirklich neues Design von Zhaoxin?

http://www.zhaoxin.com/ZXC.aspx?seriesid=5
We believe that ZX-C is basically CNR (Isaiah II), and ZX-C+ only added support for SM3/SM4. They are so similar that they have have the same CPUID. ZX-D is the first major redesign.

See also:
https://www.forum-3dcenter.org/vbulletin/showthread.php?p=12148644#post12148644
https://www.forum-3dcenter.org/vbulletin/showthread.php?p=12205567#post12205567
https://www.forum-3dcenter.org/vbulletin/showthread.php?p=12231741#post12231741

y33H@

2020-04-02, 16:05:07

Got ya, thanks.

=Floi=

2020-04-02, 23:59:32

Der ZX-C+ bzw der C4701 ist doch ein alter 28 nm 4C mit 2 GHz und kein wirklich neues Design von Zhaoxin?

http://www.zhaoxin.com/ZXC.aspx?seriesid=5

Für die kühlung aber auch wieder ok.
Bei der leistung schlägt der geringe takt voll zu.

Tralalak

2020-04-03, 10:01:09

Dr. Ian Cutress with Zhaoxin KX-6000 Series processor and mainboard

I know I keep teasing it, but here's 8 cores of Chinese x86 on 16nm, built for 2020. Let me get a couple deadlines out the way over the next couple of weeks, and we'll do a full exposé. Die size is ~150mm2

https://pbs.twimg.com/media/EUnRG8YWsAATW7E?format=jpg&name=small

source: https://twitter.com/IanCutress/status/1245756699400638466

Loeschzwerg

2020-04-03, 10:28:21

That's an engineering/reference board directly from Zhaoxin. Nice.

A few retail boards beside the known board from Cjoyin:
69653 69654 69655 69656

Tralalak

2020-04-05, 14:36:27

47:42 - 55:15

1Pcss7Kbi9M

I supported him with this drivers on youtube

P.S. here is drivers to Zhaoxin/S3G C-960 iGPU
https://support.hp.com/vn-en/drivers/selfservice/hp-268-pro-g1-mt/32591011/model/32591012

And here is drivers to Zhaoxin/S3G C-860 iGPU
https://www.catalog.update.microsoft.com/Search.aspx?q=Zhaoxin

And here is latest S3G drivers
https://www.catalog.update.microsoft.com/Search.aspx?q=S3G

P.S. to Loeschzwerg: is are some other official C-960 frivre available?

Loeschzwerg

2020-04-05, 15:09:29

The "HP" driver is version 10.17.19.20. Beside this one I do have an older version 10.17.19.7.

I've uploaded it here:
https://www.file-upload.net/download-13982880/10.17.19.7_Win10x64.cab.html

CHANGE HISTORY:

10.17.19.0020-37.00.20(09/27/2019)
1. Add Brightness control by GFX PWM
2. HDMI Electrical Testing Issue in 297MHz(4K x 2K @30)
3. Paint 3D occurs blurred screen when zoom out to below 33%
4. The UUT pops up the error box when running 3DMARK. Line:728 ERROR information

10.17.19.0017-37.00.17(06/20/2019)
1. Test with Win10 19H1

10.17.19.0015-37.00.15(04/29/2019)
1. Eclk=300MHz

10.17.19.0007-37.00.07(11/30/2018)
1. RC1 release with RS5
2. Eclk=350MHz

10.17.19.0005-37.00.05
1. RC release with RS4

10.17.19.0004-37.00.04
1. Beta release with RS4
2. Change GPU name
3. Update GPU clk

10.17.19.0002-37.00.02
1. Alpha release
2. Add 4k@60Hz support

10.17.19.0000-37.00.00
1. Initial release

Tralalak

2020-04-10, 16:27:36

Zhaoxin KX-U6780A x86 CPU Tested: The Rise of China's Chips
source: https://www.tomshardware.com/features/zhaoxin-kx-u6780a-x86-cpu-tested

Loeschzwerg

2020-04-11, 10:53:11

CPU performance is as expected, IGP performance is strange... I get the same performance numbers in Dota2 with the C-645.

y33H@

2020-04-11, 13:01:52

Honestly I was expecting a more in depth analysis but maybe that's Ian's board and they have to send it back ... but yes, performance as expected at least from a CPU perspective.

Leonidas

2020-04-12, 07:47:30

Zhaoxin KX-U6780A x86 CPU Tested: The Rise of China's Chips
source: https://www.tomshardware.com/features/zhaoxin-kx-u6780a-x86-cpu-tested

Auswertung jener Benchmarks hier:
https://www.3dcenter.org/news/hardware-und-nachrichten-links-des-1011-april-2020

Tralalak

2020-04-12, 12:54:54

Nice desktop with Zhaoxin Eight Core processor

https://support.hp.com/doc-images/771/c06537734.png

HP 268 Pro G1 MT
ZhaoXin KaiXian KX-U6780A Processor
2.7 GHz
8 MB L2 cache, 8 cores
8 threads
ZhaoXin C960 UHD Graphics
70 W
Supports DDR4 memory up to 2666 MT/s

source: https://support.hp.com/in-en/product/hp-268-pro-g1-mt/32591011/document/c06537715#AbT2
source: http://h10032.www1.hp.com/ctg/Manual/c06558314

=Floi=

2020-04-12, 15:39:47

Ist der vertreib des HP PC international?

In der zusammenfassung des THG test wird er ja ganz schön auseinandergenommen.