SLI vs. CrossFire, Part 1 – mid-range multi-GPU scaling & value
Conclusion
This has been quite an enjoyable one month hand’s on experience with these video cards evaluating these SLI and CrossFire video card combinations. It did get cut short because of our motherboard malfunction but there is plenty of information regarding multi-GPU scaling in this introduction to our CrossFire versus SLI series. There are going to be at least two or three more parts to this particular series. Eventually we want to set up testing of “micro stutter” as observed in multi-GPU setups which is easy to see for yourself but very difficult to demonstrate in an evaluation.
Micro Stutter and Greater than 100% Scaling
Micro stutter bothers some gamers so much so that they cannot ignore it and they return to using a single powerful card while others never notice it at all and will swear that it does not exist. Micro stutter does exist as it is inherent to Alternate Frame Rendering (AFR), but drivers minimize its effects so that generally the performance gained by the second card outweighs any disadvantages. If you would like to weigh on our future testing methods or to specifically discuss micro stutter testing methods, feel free to comment below, ask questions, make requests or have a detailed discussion in our ABT forum.
We have also noted CrossFire and SLI scaling greater than 100% occasionally. Often this can be attributed to “benchmark noise”, the way averages are rounded off, or the nature of Alternate Frame Rendering (AFR) where multi-GPU scaling may be slightly better than 100% due to the frames that are read ahead by the CPU.
If frames for a multi-GPU system are only pre-rendered as for a single card there would be less than optimal scaling. The AFR dual-GPU system likely has to render ahead further than a single-GPU system does because there are now two GPUs to feed. In this scenario, the framerate may possibly have higher than 100% scaling during some GPU-bound situations.
In other words, a GPU can render only one frame at a time. If there were no pre-rendering at all, the CPU has to wait before constructing a new frame, while pre-rendering allows the CPU to keep on preparing frames while the GPU is busy. Since the prepared frames are stored in an offline buffer, the GPU ideally never has to wait for the CPU to construct them. Now since the pre-render value is likely higher for a multi-GPU system than it is for a single GPU system, it is possible to achieve greater than 100% scaling. The multi-GPU system’s higher pre-render values makes it probable that the CPU can continue working on frames instead of stopping for a single video card with a lower pre-render value. Of course, this would cause more input lag but that is beyond the scope of this discussion.
Other factors can also contribute to greater than 100% scaling which include “benchmarking noise” – the margin of error built into benching is generally accepted as 1-3% – and also driver bugs that might hold back a single GPU but which don’t show up in multi-GPU situations. Usually you will not encounter this; ideally scaling approaches 100% if your SLI or CrossFire drivers are working perfectly.
VALUE
Let’s again compare value by looking at e-tail and street pricing of our cards. Cards denoted with an asterisk(*) are generally more discounted and pricing is approximate and what we found as street pricing; especially the HD 5870 and GTX 480 are both EoL’d and pricing can vary greatly. The GTX 480 is not particularly appealing unless it is severely discounted as the GTX 570 is a better all-around value.
However, this fact is to be noted, if you are looking for bang for buck now with a single video card, the flagship video cards are generally quite expensive with a diminishing set of performance value for price, and as we have seen, often a flagship card is not the fastest solution when compared to our relatively low-budget cards in SLI or CrossFire.
The GTS 450 is a budget mainstream gaming card but it doesn’t translate to a great performance gaming value when SLI’d if it is compared to single more powerful cards like the HD 6950 or the GTX 560 Ti which can be purchased for about the same price. The only reason a person might choose this route is because they already have a GTS 450 and desire better performance. On the other hand, the GTX 460 is much faster than the GTS 450 as a single card for gaming, and when it is paired up with a second GTX 460 in SLI, it becomes a performance monster – often it trades blows with the GTX 580.
For $500, if the performance choice comes down to a GTX 580 or a pair of GTX 560 Ti’s in SLI, the 560s take it overwhelmingly in most cases. There will be some games that do not scale at all in which case, you are playing with a single GTX 560 Ti. Generally SLI and CrossFire scale well in most newer games and you will usually find issues in the oldest and the very newest games. The new and popular games generally get multi-GPU issues solved very quickly, but not always. And of course, unless your motherboard is set up for a third video card, most of us will be stopped in our upgrade path with two cards; you will not be able to add a third GTX 560 Ti or HD 5870, but you can always add a second flagship card later on. And of course, as we have seen, pricing does not always drop on discontinued cards like the GTX 480 which still commands a high price considering the GTX 570 is generally faster in the new games.
Hopefully, the benefits and disadvantages of CrossFire or SLI will become even clearer as we continue on in our investigation of SLI versus CrossFire. They both scale very well in newer games and in most of our 29 games that we use as our regular benchmark suite. We see some great values in our line-up from both AMD and Nvidia – it is a great time to upgrade!
We will not give any final judgments now but will withhold them until after we are done with this series. We know that we have a new powerful flagship video card coming from AMD shortly and we hope to give you a hands on full evaluation as soon as it is available. Of course, it will fit in very well with this series. We are in the middle of one of the most interesting times in the recent history of video cards as AMD and Nvidia are locked into a battle to win your hearts and money. So far, the consumer is winning!
Stay tuned, there is a lot coming from us at ABT including a Diamond USB 2.0 7.1 sound card evaluation, and a Diamond Multimedia 1080p USB to HDMI (PC to TV audio and video) converter review this week – and of course, the next part in our ongoing series – CrossFire vs. SLI.
Mark Poppin
ABT Senior Editor
Please join us in our Forums
Become a Fan on Facebook
Follow us on Twitter
For the latest updates from ABT, please join our RSS News Feed
Join our Distributed Computing teams
- Folding@Home – Team AlienBabelTech – 164304
- SETI@Home – Team AlienBabelTech – 138705
- World Community Grid – Team AlienBabelTech
I’m not 100% certain, but to analyze microstuttering, place a check in the box next to “Frametimes” in Fraps. Then when you press the hotkey, it will create a log file with a timestamp when each single frame was outputted. Only a few seconds is enough to make the log file really, really long. Then take a portion out of the log file and make a chart out of it, that measures the time between each timestamp, to see if the frames are consistent with each other in similar intervals, or if every other frame is too close to the other one.
If a game runs at say, 45fps with your SLI or CF setup, but feels more like 23-30fps, then definitely analyze this with FRAPS.
Great review so far.
How do the numbers change, if at all, if Split Frame Rendering is used instead of Alternate Frame Rendering?
The last time I used SLI was with my Voodoo2 3000s. It was a gigantic waste of $200, in 1996 dollars.
If SFR eliminates micro-stutter without too much of a performance penalty I might have to try SLI again.
why don’t they add BF:BC2?
and also 6950 n 6970 crossfire?
Concerning the microstutter, frames time (using that fraps option) is supposed to fluctuate more erratically on crossfire/sli than what it would be on a single card. I think instead of testing a moving scene, it would make more sense to test it on a completely still scene for a few seconds and see how they compare in the excel output file. You don’t want a moving scene because then you won’t be able to differentiate between the erracticness you would get from a moving scene and the erraticness you would get from microstutter.
Another interest option would be to downclock a sli/crossfire setup to a point where it matches the average framerate of the single card. This way you could could see if the multi-gpu setup looks choppier than a single card despite having the same average frame rate.
Excellent work! At the end, simple recommendations would have been nice. =)
Please include Civilization 5 if possible the next time you benchmark.
It is an important game which will test the tesselation feature and its scaling ability in multi-gpu configurations.
Civilization 5 has been added to my benching suite along with DiRT 3 and Total War, Shogun 2.
You’ve done a great job of benchmarking gaming performance, but including charts with FPS vs $$, and $$ vs wattage would be much more useful.
The wattage (both idle and load) figures can be especially important, as some of these cards can easily draw more juice than all but the most powerful (and expensive) power supplies can provide — and that definitely factors into the cost analysis.