GeekTech: nVidia's Powerful, Complicated Quad SLI

Because of the overhead of running four GPUs, we expected the Quad SLI to run a tad slower than a comparably equipped system with a single card. And let's face it, nobody is buying a system like this to type Microsoft Word documents and surf the Web. Still, the performance results we recorded were clearly out of whack. To see whether the Quad SLI was responsible for the system's mediocre overall numbers, we removed one of the dual-GPU boards, disabled one of the two GPUs on the remaining board, and reran our tests. The system's WorldBench 5 score rocketed to 136.

We shared our testing information with Aeoncraft and (more specifically) nVidia, to help us figure out what was going on. It turned out that our high-end system needed some fine-tuning.

After much back-and-forth with nVidia, we determined that Aeoncraft had shipped us a PC with the wrong BIOS, some incorrect settings, and a slightly dated video driver. The driver part is forgivable--these things rev all the time--but the BIOS issues suggest sloppy craftsmanship, especially on a system this pricey.

After installing the proper BIOS, we reran our basic WorldBench 5 tests, and the system scored a much more reasonable 134--not tops, but pretty good. The system did a lot better on our basic desktop graphics tests, too.

Of course, where a system like this should really shine is with games. To test this sucker at the resolutions nVidia recommends, we had to reconfigure a couple of our standard graphics tests. But even at high resolutions, the system posted mixed results at typical antialiasing and anisotropic filtering settings.

For example, in our FarCry test (with antialiasing at 4X and anisotropic filtering at 8X), the Quad setup registered 85 fps at 1920 by 1200 resolution and 48 fps at 2560 by 1600 resolution, whereas a Dual SLI setup (with two standard 7900 GTX boards) posted 94 fps and 59 fps, respectively. A single 7900 GTX board hit speeds of 65 fps and 33 fps on the same tests.

Our Doom 3 tests showed similar numbers for both the Quad and Dual SLI systems at the same settings: the Quad SLI posted 105 fps and 51 fps, the Dual SLI hit 103 fps and 56 fps, and the single card notched 65 fps and 30 fps.

Not until we moved out of the game setting and cranked up the SLI-only antialiasing and anisotropic filtering settings did the Quad SLI finally show some world-class muscle.

In our Doom 3 test running at 1920 by 1200 resolution with SLI settings at 8X antialiasing and 16X anisotropic filtering, the Quad SLI system posted 86 fps versus 64 fps for the Dual SLI system. And at 2560 by 1600 resolution, the Quad hit 48 fps while the Dual posted 29 fps.

In our Far Cry test using the same settings, the Quad SLI achieved 84 fps versus 49 fps for the Dual SLI at 1920 by 1200 resolution, and it rolled out 46 fps against 28 fps for the Dual SLI at 2560 by 1600 resolution.

Does anyone need to run a game at such high antialiasing and anisotropic filtering settings? I don't think so. In fact while I believe that these technologies contribute to better-looking game play, I'm unconvinced that such super-high settings enhance what the average person perceives on screen. But do some people want to run them that high anyway? Sure they do.

  • 1
  • |
  • 2
  • |
  • 3
Join the Discussion
blog comments powered by Disqus
You Might Also Like...