MultiGPU Update: Two-GPU Options in Depth
by Derek Wilson on February 23, 2009 7:30 AM EST- Posted in
- GPUs
Who Scales: How Much?
To calculate this scaling data, we simply looked at percent performance improvement of two cards over one. With perfect scaling we would see 100%, while no improvement is 0% and a negative performance improvement means that the multiGPU solution actually produced worse numbers than the single card. There's a lot of data here, so we'll break it down a bit before we present it all.
It is possible to see more than 100% scaling in some tests for different reasons. Fluctuations in benchmark performance can contribute to just over 100% situations, and some times optimizations to enable better multiGPU performance can cut some work out enabling higher performance than would otherwise have been possible. In one of the cases we test today we have a situation where single GPU performance is limited at some framerate while multiple GPUs aren't hindered by the same limit. This artificially inflates the scaling percent.
When looking at games that scale overall, we end up seeing both Radeon HD 4870 configurations (512MB and 1GB) performing worse than we expected. Granted, the 4870 1GB looks better if we only take 2560x1600 into account, but even then the Radeon HD 4850, GeForce GTX 260 and GTX 280 beat out the 4870 1GB in terms of average performance improvement (when performance improves). When we add in CPU limited cases, the 4870 cards look even worse. Consistently, most of the ways we attempted to analyze the magnitude of performance improvement (averages, geometric means, per game, across games where call cards scaled, etc.), the Radeon HD 4850 and GeForce GTX 260 (and sometimes the GTX 280) did pretty well, while the Radeon HD 4870 cards came in pretty low on the list with the 1GB often looking worse because it hit harder CPU limits at lower resolutions.
Hitting CPU or system limits does speak more to value than desirability from a performance standpoint, but it's still important to look at all the cases. Configurations with lower baseline single GPU performance will have more headroom to scale, but these might not always scale enough to be playable even if they scale well. So it's important to take both value and absolute performance data into account when looking at scaling.
We've put all this data on our benchmark pages with the performance data to make it easier to see in context. There just isn't one good way to aggregate the data or we would talk about it here. Depending on the type of analysis we try to do, we could present it in ways that favor AMD and NVIDIA and since there really isn't a "correct" way to do it we've decided to just present the data per game and leave it at that.
Facebook
Twitter
RSS
95 Comments
View All Comments
Hauk - Monday, February 23, 2009 - link
To you grammer police... get a life will ya?!?Who gives a rats ass! It's the data!
Your smug comments are of ZERO value here. You want to critique, go to a scholarly forum and do so.
Your whining is more of a distraction! How's that for gramaticly correct?
Slappi - Tuesday, February 24, 2009 - link
It should be grammar not grammer.
SiliconDoc - Wednesday, March 18, 2009 - link
Grammatically was also spelled incorrectly.lol
The0ne - Monday, February 23, 2009 - link
"In general, more than one GPU isn't that necessary for 1920x1200 with the highest quality settings,..."I see many computer setups with 22" LCDs and lower that have high end graphic cards. It just doesn't make sense to have a high end card when you're not utilizing the whole potential. Might as well save some money up front and if you do need more power, for higher resolutions later, you can always purchase an upgrade at a lower cost. Heck, most of the time there will be new models out :)
Then again, I have a qaud-core CPU that I don't utilize too but... :D
7Enigma - Monday, February 23, 2009 - link
Everyone's situation is unique. In my case I just built a nice C2D system (OC'd to 3.8GHz with a lot of breathing room up top). I have a 4870 512meg that is definitely overkill with my massive 19" LCD (1280X1024). But within the year I plan on giving my dad or wife my 19" and going to a 22-24". Using your logic I should have purchased a 4850 (or even 4830) since I don't NEED the power. But I did plan ahead to future proof my system for when I can benefit from the 4870.I think many people also don't upgrade their systems near as frequently as some of the enthusiasts do. So we spend a bit more than we would need to at that particular time to futureproof a year or two ahead.
Different strokes and all that...
strikeback03 - Monday, February 23, 2009 - link
The other side of the coin is that most likely for similar money, you could have bought something now that more closely matches your needs, and a 4870 in a year once it has been replaced by a new card if it still meets your needs.7Enigma - Tuesday, February 24, 2009 - link
Of course. Or I could spend $60 now, another $60 in 3 months, and you see the point. It's all dependant on your actual need, your perceived need, and your desire to not have to upgrade frequently.I think the 4870 is one of those cards like the ATI 9800pro that has a perfect combination of price and performance to be a very good performer for the long haul (similarly to how the 8800GTS was probably the best part from a price/performance/longevity standpoint if you were to buy it the day it first came out).
Also important is looking at both companies and seeing what they are releasing in the next 3-6 months for your/my particular price range. Everything coming out seems to be focused either on the super high end, or the low end. I don't see any significant mid-range pieces coming out in the next 3-6 months that would have made me regret my purchase. If it was late summer or fall and I knew the next round of cards were coming out I *may* have opted for a 9600GT or other lower-midrange card to hold over until the next big thing but as it stands I'll get easily a year out of my card before I even feel the need to upgrade.
Frankly the difference between 70fps and 100fps at the resolutions I would be playing (my upgrade would be either to a 22 or 24") is pretty moot.
armandbr - Monday, February 23, 2009 - link
http://www.xbitlabs.com/articles/video/display/rad...">http://www.xbitlabs.com/articles/video/display/rad...here you go
Denithor - Monday, February 23, 2009 - link
Second paragraph, closing comments:Fourth paragraph, closing comments:
Please remove the apostrophe from the first sentence (where it should read its) and instead move it to the second (which should be we're).
Otherwise excellent article. This is the kind of work I remember from years past that originally brought me to the site.
One thing - would it be too difficult to create a performance/watt chart based on a composite performance score for each single/pair of cards?
I do think you really pushed the 4850X2 a bit too much. The 9800GTX+ provides about the same level of performance (better in some cases, worse in others) and the SLI version manages to kick the crap out of the GTX 280/285 nearly across the board (with the exception of a couple of 2560x1600 memory-constricted cases) at a lower price point. That's actually in my mind one of the best performance values available today.
SiliconDoc - Wednesday, March 18, 2009 - link
Forget about Derek removing the apostrophe, how about removing the raging red fanboy ati drooling ?When the GTX260 SLI scores the 20 games runs of 21, and the 4850 DOESN'T, Derek is sure to mention not the GTX260, and on the very same page blab the 4850 sapphire "ran every test"...
This is just another red raging fanboy blab - so screw the apostrophe !
Nvidiai DISSED 'em because they can see the articles Derek posts here bleeding red all over the place.
DUH.