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
DerekWilson - Monday, February 23, 2009 - link
It really is a great looking game for an MMO. It's not the most played MMO around, but it is definitely the easiest to test. There is an area near the beginning where the player is alone in the environment and it's always the same time of day and all that stuff ... It takes out some of the factors that make getting consistent data out of other MMOs incredibly difficult.I've never had any real "issues" with it or with the results either. It's been very consistent as well. It does add value, and it's clear that games can be coded in a way that looks really good and perform like this one, so we feel it's important to getting a better feeling for what's out there and what's possible.
IKeelU - Monday, February 23, 2009 - link
Not really a big deal, but could you cut out the offhand game review comments when introducing benchmarks? I.e.: "Crysis Warhead, while not the best game around..." It feels out of place in a hardware analysis.SiliconDoc - Wednesday, March 18, 2009 - link
And Derek disses Far Cry 2 and Oblivioin where nvidia slaughters ati - then derek praises Bioshock where ati has an edge.Derek CAN'T HELP HIMSELF.
SiliconDoc - Wednesday, March 18, 2009 - link
Oh yes, and below don't forget the age of conan that favors the ati card - Derek can't stop drooling all over the place.Then come to COD, where nvidia once again slaughters - red blood everywhere - Derek says "do we really need another war game~" or the like.
Derek is red fan central and cannot stop himself.
The0ne - Monday, February 23, 2009 - link
This game is poorly programmed in the first place, does it deserve to even be included in the benchmark tests? Yes, it has the programming necessary to for the test but they're poorly programmed.IKeelU - Monday, February 23, 2009 - link
The fact that CryEngine 2 is taxing on today's hardware (and that Crytek will no doubt use derivatives of it in future games), makes it very useful in benchmarks. I hope reviewers keep using it. But by all means, feel free to disassemble Crytek's binaries and point out their code's weaknesses.Yeah, I thought not.
poohbear - Wednesday, February 25, 2009 - link
what do u mean they shouldnt include crysis warhead??? its the seminal game to see how graphics performance is to get an idea of how a particular video card will perfrom in the future. Cryengine2 is the most advanced graphics engine on the market. If a video card can provide 30 fps on a cryengine @ your resolution, then its good to last u for atleast 2 years.Razorbladehaze - Wednesday, February 25, 2009 - link
Yeah.... NO.I totally disagree with it being the most advanced. It is a decent game engine especially for benchmarking, but....
In all reality the STALKER Clear Sky revamped xray engine is far and away more advanced and superior in almost every way. It is about the same or better in regards to taxing the system (low frame rates does not necessarily translate to the game is taxing the system.). Being that these are also used in similar FPS titles they would make a interesting comparrison.
I would really like to see Anand include or swap a clear sky bench (there is a premade one available), for the Crysis or Crysis warhead. Either way no big deal many other sites post results with a CS bench that view all the time.
DerekWilson - Monday, February 23, 2009 - link
i'll take care of it.Stillglade - Monday, February 23, 2009 - link
I would love to see more info about the 4850 X2 1GB version. For over $50 cheaper, is the 1GB memory enough to compete? Is it worth paying 24% more for the 2GB version that you reviewed here?