Questionable optimizations have certainly landed NVIDIA in a lot of trouble over the past year, and I was a little shocked that the company was so forthcoming about the fact that it had optimized its drivers yet again. Clearly, NVIDIA isn’t worried about these optimizations being labeled cheats, but do they actually work? Let’s find out.
What to look for in the benchmark results
We’ll be looking for two things today—whether GeForce FX is actually any faster on nForce3, and whether NVIDIA’s platform optimizations have hurt the performance of ATI’s Radeon 9800 XT. To expose any performance differences, I’ve tested a GeForce FX 5950 Ultra and Radeon 9800 XT on nForce3 250Gb and K8T800 platforms. By looking at the performance differences between the K8T800 and nForce3 250Gb with each graphics card, we can see if NVIDIA’s optimizations make the GeForce any faster or the Radeon any slower.
To keep things manageable, I’ve limited our testing to a subset of our standard chipset benchmarking suite. We’ll be looking at gaming and 3D workstation graphics performance exclusively.
Our testing methods
As ever, we did our best to deliver clean benchmark numbers. Tests were run at least twice, and the results were averaged.
Our test systems were configured like so:
Processor | Athlon 64 3200+ 2.0GHz | |
Front-side bus | HT 16-bit/800MHz downstream HT 16-bit/800MHz upstream |
|
Motherboard | Abit KV8-MAX3 | NVIDIA reference |
North bridge | VIA K8T800 | NVIDIA nForce3 250GB |
South bridge | VIA VT8237 | |
Chipset drivers | VIA Hyperion 4.51 | NVIDIA 4.08 |
Memory size | 512MB (1 DIMM) | |
Memory type | Corsair XMS3500 DDR SDRAM at 400MHz | |
Hard drive |
Western Digital Raptor WD360GD Serial ATA hard drive |
|
Audio | VT8237/VT1616 | nForce3 250GB/ALC655 |
Graphics | ATI Radeon 9800 XT with Catalyst 4.3 drivers GeForce FX 5950 Ultra with ForceWare 56.72 drivers |
|
OS | Microsoft Windows XP Professional | |
OS updates | Service Pack 1, DirectX 9.0b |
Thanks to Corsair for providing us with memory for our testing. If you’re looking to tweak out your system to the max and maybe overclock it a little, Corsair’s RAM is definitely worth considering. We ran all the systems with 2-7-3-3 memory timings.
The test systems’ Windows desktops were set at 1280×1024 in 32-bit color at a 75Hz screen refresh rate. Vertical refresh sync (vsync) was disabled for all tests.
We used the following versions of our test applications:
- Cinebench 2003
- SPECviewperf 7.1.1
- FutureMark 3DMark03 build 340
- Comanche 4 demo
- Quake III Arena v1.32
- Serious Sam SE v1.07
- Unreal Tournament 2003 demo
Our tests and methods are generally publicly available and reproducible. If you have questions about our methods, hit our forums to talk with us about them.
Cinebench 2003 rendering and shading
The only Cinebench test to use graphics hardware acceleration is the OpenGL hardware shading test.
Not much to see here. The 9800 XT is faster than the 5950 Ultra, and moving from the K8T800 to nForce3 250Gb doesn’t give the GeForce much of a boost.
SPECviewperf workstation graphics
SPECviewperf simulates the graphics loads generated by various professional design, modeling, and engineering applications.
The 5950 Ultra gets a slight performance boost moving to the nForce3 250Gb in viewperf’s 3dsmax-02 test. The 9800 XT takes a negligible performance hit switching platforms.
In the drv-09 test, the GeForce FX sees a massive 37% increase in performance moving from the K8T800 to the nForce3 250Gb. The Radeon isn’t so lucky; it’s 2% slower on the nForce3 than on the K8T800.
The GeForce FX sees more modest performance gains in viewperf’s dx-08 and light-06 tests. The Radeon continues to lose ground moving to the nForce3 250Gb, but by only a negligible amount in dx-08 and less than 1.5% in light-06.
NVIDIA’s driver optimizations shine in viewperf’s proe-02 test, where the GeForce FX 5950 Ultra is over 32% faster on nForce3 than on the K8T800. Again, the Radeon 9800 XT’s performance drops as we move from the K8T800 to nForce3 250Gb, but by less than 2%.
Driver optimizations don’t help the GeForce FX at all in viewperf’s ugs-03 test. They don’t hurt the 9800 XT’s performance at all, either.
Quake III Arena
The nForce3 250Gb is faster with both graphics cards in Quake III Arena, but the GeForce FX’s 35% jump in performance is much more impressive than the 9800 XT’s 10% performance increase.
Wolfenstein: Enemy Territory
The trend continues in Wolfenstein: Enemy Territory, which is based on the Quake III engine. Both the Radeon and GeForce FX show improved performance moving from the K8T800 to the nForce3 250Gb, but again, the GeForce FX comes out ahead.
Unreal Tournament 2003
NVIDIA’s optimizations don’t help the GeForce FX at all in Unreal Tournament 2003, but they don’t hurt the 9800 XT, either.
Serious Sam SE
In Serious Sam SE, the nForce3 gives both cards a performance boost. The 5950 Ultra’s performance improves more than the 9800 XT’s, but not by much.
Comanche 4
The 9800 XT’s performance is unchanged between the K8T800 and nForce3 250Gb in Comanche 4, but the GeForce FX gains just over 1% from the platform switch.
3DMark03
Platform optimizations don’t help the GeForce FX post a higher overall score in 3DMark03, but the 5950 Ultra does gain a few frames per second in the benchmark’s CPU tests. Overall, the 9800 XT is a little faster on the nForce3 in the 3DMark03 CPU tests, but the Radeon’s performance deltas are much smaller than the GeForce FX’s.
Our testing shows that NVIDIA has successfully optimized its nForce3 and GeForce FX platforms to work better together than apart, at least in the case of the nForce3 250Gb and GeForce FX 5950 Ultra. Although the performance benefits associated with an all-NVIDIA setup are certainly not widespread, they’re quite impressive in a couple of the viewperf suite’s workstation graphics tests and in a handful of games. Our testing shows that the nForce3 250Gb can improve the GeForce FX 5950 Ultra’s performance by up to 35% over a K8T800, under the right conditions.
The impressive performance gains offered by NVIDIA’s platform optimizations don’t significantly hinder the performance of ATI’s Radeon 9800 XT. In our testing, the 9800 XT was often faster on the nForce3 250Gb than on the K8T800, and in the few cases where NVIDIA’s optimizations appear to slow the 9800 XT, the performance difference is never greater than 2%. Not bad, all things considered.
So NVIDIA’s platform optimizations work—in some cases quite well—but they also raise some interesting issues. Should NVIDIA work with ATI to ensure that Radeons can benefit from the same nForce3 optimizations that GeForce FX cards enjoy? Could NVIDIA’s efforts to optimize the GeForce FX for nForce3 harm GeForce performance with other core logic chipsets down the road? Thus far, I see no reason to rule NVIDIA’s nForce3 optimizations as either unfair or detrimental to GeForce performance on other platforms. Platform-specific optimizations may come across as a little nefarious, but NVIDIA certainly isn’t alone in that regard. After all, you don’t see ATI extending its SURROUNDVIEW multimonitor technology beyond the Radeon 9100 IGP and Radeon graphics cards.