SLI vs. CrossFire, Part 2 – High-end multi-GPU scaling
The evaluation that we are presenting here today compares the performance of the very fastest AMD and Nvidia video cards in SLI or CrossFire. This is Part Two of a new series that continues where our “SLI vs. CrossFire, Part 1 – mid-range multi-GPU scaling & value,” left off. In fact, that article is a continuation of an older series where we benchmarked the Core i7-920 versus the Q9550s and Phenom II CPUs in “Core i7 vs. Penryn vs. Phenom II with HD 4870-X2 & TriFire,” which we published in January, 2010.
This second part of our evaluation only deals with CrossFire scaling versus SLI scaling on Core i7-920 as later parts of this series will expand to cover over- and under-clocked multi-GPU configurations. Besides overclocked video cards, we will include additional CPU platforms, including dual- versus quad-core and hopefully will include Bulldozer FX series “Zambezi” performance and later on, compare to Intel’s Ivy Bridge CPU platform.
One thing to note is that all of our video cards were run at their stock speeds. For our next Part 3 evaluation, we will attempt to overclock our HD 6990 so that it approaches the stock speed of the HD 6970. This will give us a very good idea of further scaling of CrossFire-X3. And of course, we will over- and under-clock our CPUs to see if there is a “bottleneck”.
We are actually expanding on our SLI versus CrossFire series with three synthetic benches and 22 relatively new games – including 12 DX11 games, 5 DX10 games, 4 DX9 games and 1 OpenGL game. We are using the most powerful graphics cards available as we are now benchmarking with the two fastest single-GPU graphics cards from each vendor – the GTX 580 and the HD 6970.
We also test with the two fastest dual-GPU cards; the HD 6990 and the GTX 590. We also include HD 6990 plus HD 6970 results for Tri-Fire X-3; unfortunately, the GTX 590 will not SLI with the GTX 580 for Tri-SLI as they are not supported together by Nvidia’s drivers. We also test HD 6970 CrossFire and GTX 580 SLI which perform faster than the stock HD 6990 or the GTX 590 respectively.
After over a week of benchmarking runs – literally hundreds of individual benchmark runs which we have gathered, analyzed and charted for you – we now have some more solid evidence about multi-GPU scaling. We also want to see if perhaps our Core i7-920 at 3.8GHz (3.97GHz with Turbo) is beginning to “bottleneck” our graphics – especially at 1920×1200 and below.
We are continuing to test at two of the most demanding wide-screen resolutions, 1920×1200 and 2560×1600 and we will also look at the Super-widescreen resolution of 5760×1080 when available using in-game settings. We generally use 2x/4xAA or 8xAA plus 16xAF and with maximum (ultra) DX11/10.1/10/9c details whenever it is available. For all of our testing in this review, we are benching with Catalyst 11.8 and GeForce 280.26 drivers; each one a very solid and stable WHQL driver set. GeForce 280.26 is still current at the time of publication and Catalyst 11.9 was released yesterday (expect a Performance Analysis this weekend).
What about “micro-stutter” and how well do games generally scale by adding a second video card – or even a third? Please continue on to the next page for the complete hardware and software setup of our testing platform. We shall see what happens when high performance single-GPU video cards meets multi-GPU video cards in PC gaming for the Autumn of 2011.
Thanks for this article. It was a very interesting read
I’m really looking forward to the next parts in the series. Overclocking and potential CPU-bottlenecking in Single-Card VS SLI is something I’ve been wondering about for a long while. And also microstuttering, I’ve never experienced it myself, but it scares me enough to make me cautions of buying another GTX 570 to SLI.
And oh, is it possible to get Battlefield 3 Beta added in your test-games?
Thanks for the feedback. I would not be afraid of getting a second GTX 570 for SLI. Nvidia (and AMD) work to minimize micro stuttering in the drivers and it is something that you can generally further alleviate by backing down on settings if you notice it.
I plan to add BF3 to my regular benching suite after it is released. The beta is only going to be valid for less than a month.
Thank you very much for this! It isn’t easy finding benchmark results with newer drivers. Not for a quad-SLI or quadfire setup that is. Cheers!
Err, strike the quad-SLI and quadfire part, but all the same – it is nice to find more up to date benchmarks!