Our own Lynx had the chance to see NVIDIA SLI first-hand at Fragapalooza. Read his brief article and watch the video comparing a computer running the Unreal 3 engine with one GPU compared to 2 GPU’s.
I’m a bit late to get this up as I saw the presentation on Friday night but, I guess better late than never. I was at Fragapalooza, Canada’s biggest gaming event and nVidia was there fresh from their debut European trip in Paris of the SLI technology and ready to make the N. American debut of their SLI technology.Â
The crate containing the SLI machines arrived on Wednesday afternoon while volunteers and staff were busy setting up Fragapalooza. I happened to help off-load the crate and was quite eager to see the beasts in action. The two machines were kept “under-wraps” overnight and were then set-up first thing on Thursday morning.
In their presentation to the attendees at Fragapalooza, they went over the technology and the fact that the SLI naming doesn’t have anything to do with Vodoo’s own SLI; it only shares the name “SLI.” Voodoo’s SLI tech stood for “Scan Line Interleave” and each GPU was responsible for drawing every other line on the screen. The NVIDIA technology is completely different and stands for “Scalable Link Interface,” and essentially one GPU is responsible for drawing either the top or bottom half of the screen, but not quite because with the algorithms in place to make sure each GPU shares an identical workload, it can end up with one GPU drawing the upper 1/3 of the screen consisting of a heavier graphics load while the other GPU draws the bottom 2/3 of the screen with a lower graphics load.
This is what NVIDIA themselves say about SLI:
- Bus bandwidth and system communication: NVIDIA SLI technology fully exploits the increased bandwidth provided by the latest bus architecture—PCI Express—when communicating between the GPUs and the CPU. Providing 60x the bandwidth of PCI, the PCI Express interface eliminates the bottlenecks caused by previous bus architectures and allows for maximum system performance in a multi-GPU configuration.
- GPU communication: Featuring an intelligent communication protocol embedded in the GPU, and a high-speed digital interface, NVIDIA SLI-based GPUs can easily communicate with one another without the overhead associated with a bus-only implementation. In addition, unique software algorithms efficiently share the workload to deliver unbelievable performance.
- Image Quality: Advanced compositing, rendering, and scanout technology delivers uncompromised image quality, eliminating the “flashing” effect found in previous solutions. NVIDIA SLI multi-GPU technology also supports both analog and digital output.
The cool thing with NVIDIA’s SLI is of course the fact that it is transparent to games and applications, which means that any game/application should benefit from it.
You can read more about SLI at NVIDIA on this page: http://www.nvidia.com/page/sli.html.
The machines they brought contained mucho power in the form of the reference designed PCI-EX Motherboard (Dual CPU capable but only using a single CPU), 6800 Ultra GPU’s and 1GB Corsair XMS RAM.
They were running the demo using the new UT3 engine and running two SLI machines side-by-side, one running with only one GPU enabled and the other running full SLI dual 6800U GPU’s. I tried to get them to let me do a “stat fps” in the console so I could view the FPS, but of course they wouldn’t let me. NVIDIAÂ is claiming a 1.8X increase over running a single GPU.
The display all set up and awaiting participants. Chris talking to participants about the SLI technology
I made a small video of the UT engine demo to show the difference in speed between the single 6800U GPU and the dual/SLI 6800U. You can download it by clicking on the link below. It’s 1.95MB in size and runs about 30 seconds long.
 AVI Video comparing single 6800U and SLI/dual 6800U
(if you cannot open the AVI directly, right click on the link and save the file to the HD before watching it)
What do you think about NVIDIA’s new SLI-technology? Come and discuss this on our forums. |
Â