nagus
2003-04-23, 13:03:27
Quelle: http://forums.anandtech.com/messageview.cfm?catid=31&threadid=1031972
--------------------------------------------------------------
Today we have another PCI video card comparison. Today’s benchmarks were done on an MX 440 PCI, a Radeon 9100 PCI, and a FX 5200 PCI. All the cards come equipped with 128mb DDR memory, and all are hardwired for different versions of Direct X. The first player is the Inno3D MX440 PCI, one of the best DX7 PCI video cards available. Visiontek’s Xtasy 9100 PCI is the new PCI card outfitted for DX8.1, and finally, the PNY FX 5200 PCI is the first mainstream DX9 PCI video card. For a baseline comparison, all the benchmarks that would support it were also run on Intel’s 845G “Extreme Graphics” integrated video.
On a side note, much of my benchmarking is limited by resolution factors. I currently have only a 15” CRT, so 1024x768 is the maximum resolution I can run. In some of the tests, the three video cards showed similar performance and the results would probably only begin to differentiate themselves at higher resolutions, where the tests would depend more on the video cards’ abilities and less on the processor speed. However, I would propose that most people who buy a PCI video card for gaming will probably not be playing graphically intensive games above 1024x768x32, since these cards are already slower (because of the restrictive PCI bus) than their “budget”-oriented AGP counterparts.
The benches were done using the latest drivers for each card, being the Catalyst 3.2 drivers for the 9100 PCI and the 43.45 Detonators for the 440 and FX 5200 PCI. Here is a list of the benchmarks used today:
3DMark2001SE
3DMark2003
Codecreatures
Dungeon Siege benchmark utility
DroneZ OpenGL benchmark
Comanche 4 Demo benchmark utility
Unreal Tournament 2003 Demo benchmark utility
System specifications:
Dell 4500S
2.4 P4 (400)
768mb pc2100
WinXP Home edition
Video cards tested:
Intel 845G integrated video “Extreme Graphics”, 64mb of system memory
Inno3D MX 440 PCI 128mb DDR with VGA & TV/out (s-video) - 270/200(400) – 5ns
PNY FX 5200 PCI 128mb with Dual-VGA & TV/out (s-video) – 250/150(300) – 6ns
Visiontek Xtasy 9100 PCI 128mb DDR with VGA, DVI-D & no TV/out – 250/250(500) – 4ns
3D Mark 2001 SE:
3DMark Score 1255
3DMark Score 4272
3DMark Score 4909
3DMark Score 7426
Game 1 - Car Chase - Low Detail 22.0 fps
Game 1 - Car Chase - Low Detail 106.6 fps
Game 1 - Car Chase - Low Detail 88.8 fps
Game 1 - Car Chase - Low Detail 110.7 fps
Game 1 - Car Chase - High Detail 10.1 fps
Game 1 - Car Chase - High Detail 41.5 fps
Game 1 - Car Chase - High Detail 40.2 fps
Game 1 - Car Chase - High Detail 48.6 fps
Game 2 - Dragothic - Low Detail 22.3 fps
Game 2 - Dragothic - Low Detail 55.3 fps
Game 2 - Dragothic - Low Detail 60.3 fps
Game 2 - Dragothic - Low Detail 108.4 fps
Game 2 - Dragothic - High Detail 10.7 fps
Game 2 - Dragothic - High Detail 17.9 fps
Game 2 - Dragothic - High Detail 27.3 fps
Game 2 - Dragothic - High Detail 62.6 fps
Game 3 - Lobby - Low Detail 22.0 fps
Game 3 - Lobby - Low Detail 80.6 fps
Game 3 - Lobby - Low Detail 87.2 fps
Game 3 - Lobby - Low Detail 115.9 fps
Game 3 - Lobby - High Detail 8.8 fps
Game 3 - Lobby - High Detail 33.0 fps
Game 3 - Lobby - High Detail 38.3 fps
Game 3 - Lobby - High Detail 53.4 fps
Game 4 - Nature Not supported by hardware
Game 4 - Nature Not supported by hardware
Game 4 - Nature 21.5 fps
Game 4 - Nature 39.3 fps
Fill Rate (Single-Texturing) 185.4 MTexels/s
Fill Rate (Single-Texturing) 434.5 MTexels/s
Fill Rate (Single-Texturing) 471.7 MTexels/s
Fill Rate (Single-Texturing) 746.4 MTexels/s
Fill Rate (Multi-Texturing) 341.2 MTexels/s
Fill Rate (Multi-Texturing) 804.1 MTexels/s
Fill Rate (Multi-Texturing) 718.6 MTexels/s
Fill Rate (Multi-Texturing) 1864.3 MTexels/s
High Polygon Count (1 Light) 3.7 MTriangles/s
High Polygon Count (1 Light) 20.4 MTriangles/s
High Polygon Count (1 Light) 22.3 MTriangles/s
High Polygon Count (1 Light) 22.6 MTriangles/s
High Polygon Count (8 Lights) 3.6 MTriangles/s
High Polygon Count (8 Lights) 6.5 MTriangles/s
High Polygon Count (8 Lights) 4.9 MTriangles/s
High Polygon Count (8 Lights) 8.5 MTriangles/s
Environment Bump Mapping Not supported by hardware
Environment Bump Mapping Not supported by hardware
Environment Bump Mapping 57.9 fps
Environment Bump Mapping 91.8 fps
DOT3 Bump Mapping 20.6 fps
DOT3 Bump Mapping 68.9 fps
DOT3 Bump Mapping 55.5 fps
DOT3 Bump Mapping 70.3 fps
Vertex Shader 18.9 fps
Vertex Shader 4.1 fps
Vertex Shader 40.6 fps
Vertex Shader 69.3 fps
Pixel Shader Not supported by hardware
Pixel Shader Not supported by hardware
Pixel Shader 72.2 fps
Pixel Shader 81.2 fps
Advanced Pixel Shader Not supported by hardware
Advanced Pixel Shader Not supported by hardware
Advanced Pixel Shader 21.5 fps
Advanced Pixel Shader 77.0 fps
Point Sprites 1.3 MSprites/s
Point Sprites 10.5 MSprites/s
Point Sprites 9.1 MSprites/s
Point Sprites 23.7 MSprites/s
Dungeon Siege benchmark utility 1024x768x32:
Average 26.04 fps
Average 72.19 fps
The FX 5200 would not run the Dungeon Siege benchmark. The screen went blank, but I could occasionally see weapons and spell effects on the screen (i.e. the flight path of an arrow or the glowing effects of a spell). I tried the 43.45, the 41.09 and the 40.72 drivers, but nothing worked. The same test ran beautifully on the MX 440 PCI with the same drivers, so I can only guess that this was a hardware issue.
Average 59.73 fps
DroneZ OpenGL benchmark at 1024x768x32:
Benchmark results
Rendered Frames: 9721
Minimum FPS: 12.35
Minimum FPS: 76.39
Minimum FPS: 85.49
Minimum FPS: 92.63
Maximum FPS: 54.46
Maximum FPS: 228.77
Maximum FPS: 349.30
Maximum FPS: 408.51
Average FPS: 23.4690
Average FPS: 113.5936
Average FPS: 137.8713
Average FPS: 156.3428
Minimum GL K-triangles: 0.88
Minimum GL K-triangles: 7.51
Minimum GL K-triangles: 13.73
Minimum GL K-triangles: 14.35
Maximum GL K-triangles: 232.27
Maximum GL K-triangles: 735.93
Maximum GL K-triangles: 832.26
Maximum GL K-triangles: 1062.56
Average GL K-triangles: 88.7187
Average GL K-triangles: 427.8348
Average GL K-triangles: 518.6028
Average GL K-triangles: 576.0433
Minimum T&L K-triangles: 0.95
Minimum T&L K-triangles: 5.38
Minimum T&L K-triangles: 9.57
Minimum T&L K-triangles: 10.25
Maximum T&L K-triangles: 441.46
Maximum T&L K-triangles: 1560.35
Maximum T&L K-triangles: 1779.33
Maximum T&L K-triangles: 2005.32
Average T&L K-triangles: 164.3632
Average T&L K-triangles: 792.3255
Average T&L K-triangles: 961.5839
Average T&L K-triangles: 1066.3693
Comanche 4 Demo benchmark utility at 800x600x32 & 1024x768x32:
Note: The demo requires hardware T&L and, therefore, the test would not run on the integrated Intel graphics.
800X600x32:
Frames per second: 31.51 avg
Frames per second: 31.30 avg
Frames per second: 33.79 avg
1024x768x32:
Frames per second: 29.98 avg
Frames per second: 26.70 avg
Frames per second: 31.56 avg
UT2K3 Demo benchmark utility at 800x600x32 & 1024x768x32:
800x600x32:
flyby: 15.98
botmatch: 11.85
flyby: 114.90
botmatch: 45.08
flyby: 105.77
botmatch: 45.42
flyby: 136.20
botmatch: 44.92
1024x768x32:
flyby: 11.40
botmatch: 8.15
flyby: 84.27
botmatch: 45.65
flyby: 76.34
botmatch: 43.83
flyby: 111.57
botmatch: 46.50
Codecreatures:
The Codecreatures benchmark runs tests at three increasing resolutions, starting at 1024x768x32. Being that this is my highest attainable resolution, I was only able to run at 1024x768. This is a DX8.1 benchmark, so only the 9100 PCI and the FX 5200 PCI could complete the test.
1024x768x32:
Average fps: 11.3
Average fps: 18.2
3D Mark 2003:
This test is mainly for DX9 video cards, although DX8.1 cards can run most of the tests. It absolutely would not run on the Intel graphics and would only run one test on the DX7-based MX 440 PCI, so I only included the 9100 and FX 5200 PCI cards. The test is run at 1024x768x32:
3DMark Score 1197
3DMark Score 1240
GT1 - Wings of Fury 63.3 fps
GT1 - Wings of Fury 65.5 fps
GT2 - Battle of Proxycon 5.4 fps
GT2 - Battle of Proxycon 9.3 fps
GT3 - Troll's Lair 5.0 fps
GT3 - Troll's Lair 8.9 fps
GT4 - Mother Nature 7.8 fps
GT4 - Mother Nature Not Supported
Feature Tests
Fill Rate (Single-Texturing) 428.4 MTexels/s
Fill Rate (Single-Texturing) 616.9 MTexels/s
Fill Rate (Multi-Texturing) 684.7 MTexels/s
Fill Rate (Multi-Texturing) 1736.8 MTexels/s
Vertex Shader 4.4 fps
Vertex Shader 8.5 fps
Pixel Shader 2.0 7.7 fps
Pixel Shader 2.0 Not Supported
Ragtroll 3.7 fps
Ragtroll 4.0 fps
In conclusion: The clear winner here today is the Visiontek 9100 PCI. It is based on the Radeon 8500 core, a tried and true product that has been polished over many driver releases since its introduction. In either OpenGL or D3D tests that are not CPU-bound, it is clearly the most powerful card of the three tested here. The FX 5200 PCI has a lot of useful features such as Dual-VGA, DX9 compliance, and TV/out. I am confused, however, why PNY would release such a lowly-powered PCI card. Normally, AGP versions of the FX 5200 come clocked at 250/200(400). The 5200 PCI sports 6ns DDR RAM rated at 150mhz, or effectively 300mhz. Even DDR MX 420 PCI video cards come equipped with 166(333)mhz DDR RAM. I am certain that the FX 5200 PCI’s performance will improve over time with future driver releases, but I doubt any driver release can make up the performance difference that the 9100 PCI enjoys, in part, because of its 200mhz faster memory.
If you are looking for a decent performing PCI video card for a good price, then the MX 440 PCI should meet your needs nicely. If you enjoy using the TV/out function, then either Nvidia-based card will do. If, however, you simply want the most-powerful and best performing PCI video card, then the 9100 PCI is it.
----------------------------------------------------------------
--------------------------------------------------------------
Today we have another PCI video card comparison. Today’s benchmarks were done on an MX 440 PCI, a Radeon 9100 PCI, and a FX 5200 PCI. All the cards come equipped with 128mb DDR memory, and all are hardwired for different versions of Direct X. The first player is the Inno3D MX440 PCI, one of the best DX7 PCI video cards available. Visiontek’s Xtasy 9100 PCI is the new PCI card outfitted for DX8.1, and finally, the PNY FX 5200 PCI is the first mainstream DX9 PCI video card. For a baseline comparison, all the benchmarks that would support it were also run on Intel’s 845G “Extreme Graphics” integrated video.
On a side note, much of my benchmarking is limited by resolution factors. I currently have only a 15” CRT, so 1024x768 is the maximum resolution I can run. In some of the tests, the three video cards showed similar performance and the results would probably only begin to differentiate themselves at higher resolutions, where the tests would depend more on the video cards’ abilities and less on the processor speed. However, I would propose that most people who buy a PCI video card for gaming will probably not be playing graphically intensive games above 1024x768x32, since these cards are already slower (because of the restrictive PCI bus) than their “budget”-oriented AGP counterparts.
The benches were done using the latest drivers for each card, being the Catalyst 3.2 drivers for the 9100 PCI and the 43.45 Detonators for the 440 and FX 5200 PCI. Here is a list of the benchmarks used today:
3DMark2001SE
3DMark2003
Codecreatures
Dungeon Siege benchmark utility
DroneZ OpenGL benchmark
Comanche 4 Demo benchmark utility
Unreal Tournament 2003 Demo benchmark utility
System specifications:
Dell 4500S
2.4 P4 (400)
768mb pc2100
WinXP Home edition
Video cards tested:
Intel 845G integrated video “Extreme Graphics”, 64mb of system memory
Inno3D MX 440 PCI 128mb DDR with VGA & TV/out (s-video) - 270/200(400) – 5ns
PNY FX 5200 PCI 128mb with Dual-VGA & TV/out (s-video) – 250/150(300) – 6ns
Visiontek Xtasy 9100 PCI 128mb DDR with VGA, DVI-D & no TV/out – 250/250(500) – 4ns
3D Mark 2001 SE:
3DMark Score 1255
3DMark Score 4272
3DMark Score 4909
3DMark Score 7426
Game 1 - Car Chase - Low Detail 22.0 fps
Game 1 - Car Chase - Low Detail 106.6 fps
Game 1 - Car Chase - Low Detail 88.8 fps
Game 1 - Car Chase - Low Detail 110.7 fps
Game 1 - Car Chase - High Detail 10.1 fps
Game 1 - Car Chase - High Detail 41.5 fps
Game 1 - Car Chase - High Detail 40.2 fps
Game 1 - Car Chase - High Detail 48.6 fps
Game 2 - Dragothic - Low Detail 22.3 fps
Game 2 - Dragothic - Low Detail 55.3 fps
Game 2 - Dragothic - Low Detail 60.3 fps
Game 2 - Dragothic - Low Detail 108.4 fps
Game 2 - Dragothic - High Detail 10.7 fps
Game 2 - Dragothic - High Detail 17.9 fps
Game 2 - Dragothic - High Detail 27.3 fps
Game 2 - Dragothic - High Detail 62.6 fps
Game 3 - Lobby - Low Detail 22.0 fps
Game 3 - Lobby - Low Detail 80.6 fps
Game 3 - Lobby - Low Detail 87.2 fps
Game 3 - Lobby - Low Detail 115.9 fps
Game 3 - Lobby - High Detail 8.8 fps
Game 3 - Lobby - High Detail 33.0 fps
Game 3 - Lobby - High Detail 38.3 fps
Game 3 - Lobby - High Detail 53.4 fps
Game 4 - Nature Not supported by hardware
Game 4 - Nature Not supported by hardware
Game 4 - Nature 21.5 fps
Game 4 - Nature 39.3 fps
Fill Rate (Single-Texturing) 185.4 MTexels/s
Fill Rate (Single-Texturing) 434.5 MTexels/s
Fill Rate (Single-Texturing) 471.7 MTexels/s
Fill Rate (Single-Texturing) 746.4 MTexels/s
Fill Rate (Multi-Texturing) 341.2 MTexels/s
Fill Rate (Multi-Texturing) 804.1 MTexels/s
Fill Rate (Multi-Texturing) 718.6 MTexels/s
Fill Rate (Multi-Texturing) 1864.3 MTexels/s
High Polygon Count (1 Light) 3.7 MTriangles/s
High Polygon Count (1 Light) 20.4 MTriangles/s
High Polygon Count (1 Light) 22.3 MTriangles/s
High Polygon Count (1 Light) 22.6 MTriangles/s
High Polygon Count (8 Lights) 3.6 MTriangles/s
High Polygon Count (8 Lights) 6.5 MTriangles/s
High Polygon Count (8 Lights) 4.9 MTriangles/s
High Polygon Count (8 Lights) 8.5 MTriangles/s
Environment Bump Mapping Not supported by hardware
Environment Bump Mapping Not supported by hardware
Environment Bump Mapping 57.9 fps
Environment Bump Mapping 91.8 fps
DOT3 Bump Mapping 20.6 fps
DOT3 Bump Mapping 68.9 fps
DOT3 Bump Mapping 55.5 fps
DOT3 Bump Mapping 70.3 fps
Vertex Shader 18.9 fps
Vertex Shader 4.1 fps
Vertex Shader 40.6 fps
Vertex Shader 69.3 fps
Pixel Shader Not supported by hardware
Pixel Shader Not supported by hardware
Pixel Shader 72.2 fps
Pixel Shader 81.2 fps
Advanced Pixel Shader Not supported by hardware
Advanced Pixel Shader Not supported by hardware
Advanced Pixel Shader 21.5 fps
Advanced Pixel Shader 77.0 fps
Point Sprites 1.3 MSprites/s
Point Sprites 10.5 MSprites/s
Point Sprites 9.1 MSprites/s
Point Sprites 23.7 MSprites/s
Dungeon Siege benchmark utility 1024x768x32:
Average 26.04 fps
Average 72.19 fps
The FX 5200 would not run the Dungeon Siege benchmark. The screen went blank, but I could occasionally see weapons and spell effects on the screen (i.e. the flight path of an arrow or the glowing effects of a spell). I tried the 43.45, the 41.09 and the 40.72 drivers, but nothing worked. The same test ran beautifully on the MX 440 PCI with the same drivers, so I can only guess that this was a hardware issue.
Average 59.73 fps
DroneZ OpenGL benchmark at 1024x768x32:
Benchmark results
Rendered Frames: 9721
Minimum FPS: 12.35
Minimum FPS: 76.39
Minimum FPS: 85.49
Minimum FPS: 92.63
Maximum FPS: 54.46
Maximum FPS: 228.77
Maximum FPS: 349.30
Maximum FPS: 408.51
Average FPS: 23.4690
Average FPS: 113.5936
Average FPS: 137.8713
Average FPS: 156.3428
Minimum GL K-triangles: 0.88
Minimum GL K-triangles: 7.51
Minimum GL K-triangles: 13.73
Minimum GL K-triangles: 14.35
Maximum GL K-triangles: 232.27
Maximum GL K-triangles: 735.93
Maximum GL K-triangles: 832.26
Maximum GL K-triangles: 1062.56
Average GL K-triangles: 88.7187
Average GL K-triangles: 427.8348
Average GL K-triangles: 518.6028
Average GL K-triangles: 576.0433
Minimum T&L K-triangles: 0.95
Minimum T&L K-triangles: 5.38
Minimum T&L K-triangles: 9.57
Minimum T&L K-triangles: 10.25
Maximum T&L K-triangles: 441.46
Maximum T&L K-triangles: 1560.35
Maximum T&L K-triangles: 1779.33
Maximum T&L K-triangles: 2005.32
Average T&L K-triangles: 164.3632
Average T&L K-triangles: 792.3255
Average T&L K-triangles: 961.5839
Average T&L K-triangles: 1066.3693
Comanche 4 Demo benchmark utility at 800x600x32 & 1024x768x32:
Note: The demo requires hardware T&L and, therefore, the test would not run on the integrated Intel graphics.
800X600x32:
Frames per second: 31.51 avg
Frames per second: 31.30 avg
Frames per second: 33.79 avg
1024x768x32:
Frames per second: 29.98 avg
Frames per second: 26.70 avg
Frames per second: 31.56 avg
UT2K3 Demo benchmark utility at 800x600x32 & 1024x768x32:
800x600x32:
flyby: 15.98
botmatch: 11.85
flyby: 114.90
botmatch: 45.08
flyby: 105.77
botmatch: 45.42
flyby: 136.20
botmatch: 44.92
1024x768x32:
flyby: 11.40
botmatch: 8.15
flyby: 84.27
botmatch: 45.65
flyby: 76.34
botmatch: 43.83
flyby: 111.57
botmatch: 46.50
Codecreatures:
The Codecreatures benchmark runs tests at three increasing resolutions, starting at 1024x768x32. Being that this is my highest attainable resolution, I was only able to run at 1024x768. This is a DX8.1 benchmark, so only the 9100 PCI and the FX 5200 PCI could complete the test.
1024x768x32:
Average fps: 11.3
Average fps: 18.2
3D Mark 2003:
This test is mainly for DX9 video cards, although DX8.1 cards can run most of the tests. It absolutely would not run on the Intel graphics and would only run one test on the DX7-based MX 440 PCI, so I only included the 9100 and FX 5200 PCI cards. The test is run at 1024x768x32:
3DMark Score 1197
3DMark Score 1240
GT1 - Wings of Fury 63.3 fps
GT1 - Wings of Fury 65.5 fps
GT2 - Battle of Proxycon 5.4 fps
GT2 - Battle of Proxycon 9.3 fps
GT3 - Troll's Lair 5.0 fps
GT3 - Troll's Lair 8.9 fps
GT4 - Mother Nature 7.8 fps
GT4 - Mother Nature Not Supported
Feature Tests
Fill Rate (Single-Texturing) 428.4 MTexels/s
Fill Rate (Single-Texturing) 616.9 MTexels/s
Fill Rate (Multi-Texturing) 684.7 MTexels/s
Fill Rate (Multi-Texturing) 1736.8 MTexels/s
Vertex Shader 4.4 fps
Vertex Shader 8.5 fps
Pixel Shader 2.0 7.7 fps
Pixel Shader 2.0 Not Supported
Ragtroll 3.7 fps
Ragtroll 4.0 fps
In conclusion: The clear winner here today is the Visiontek 9100 PCI. It is based on the Radeon 8500 core, a tried and true product that has been polished over many driver releases since its introduction. In either OpenGL or D3D tests that are not CPU-bound, it is clearly the most powerful card of the three tested here. The FX 5200 PCI has a lot of useful features such as Dual-VGA, DX9 compliance, and TV/out. I am confused, however, why PNY would release such a lowly-powered PCI card. Normally, AGP versions of the FX 5200 come clocked at 250/200(400). The 5200 PCI sports 6ns DDR RAM rated at 150mhz, or effectively 300mhz. Even DDR MX 420 PCI video cards come equipped with 166(333)mhz DDR RAM. I am certain that the FX 5200 PCI’s performance will improve over time with future driver releases, but I doubt any driver release can make up the performance difference that the 9100 PCI enjoys, in part, because of its 200mhz faster memory.
If you are looking for a decent performing PCI video card for a good price, then the MX 440 PCI should meet your needs nicely. If you enjoy using the TV/out function, then either Nvidia-based card will do. If, however, you simply want the most-powerful and best performing PCI video card, then the 9100 PCI is it.
----------------------------------------------------------------