Who Scales

As we mentioned, we'll be looking at scaling from both 1 to 3 and 2 to 3 GPUs. This gives us a few more metrics than last time to look at both overall and on a per game basis. From the overall standpoint, we'll first look at scaling from 1 to 3 GPUs. We'll again look at general success as >33.3% scaling and complete failure will be <5% scaling. This will give us information on how many titles seem to be only CPU limited and how many are of zero or negative value as compared to a single card.

Before we get to the numbers, it is important to note that all of this data is out of 21 tests for AMD cards (like the previous article) but out of 20 for NVIDIA hardware. We had an issue with FRAPS running at 2560x1600 with 3-way NVIDIA solutions. We do want to be clear that the game ran fine, and this seems to be a high res high memory usage issue in Race Driver GRID when FRAPS is combined with 3-way and higher SLI. Let's make it clear that this isn't an issue with the game or the hardware per se, but an issue in combination with FRAPS. 3-way SLI runs really well on Race Driver GRID: we just can't tell you how well unless and until this issue is resolved. We leave the game in this article because there is AMD data at 2560x1600 and because there's still usable data at 1680x1050.

First up in our look at who scales is general success (>33.3% scaling) in moving from 1 to 3 GPUs:

NVIDIA GeForce GTX 28516
NVIDIA GeForce GTX 28018
NVIDIA GeForce GTX 26019
NVIDIA GeForce 9800 GTX+19
ATI Radeon HD 4870 1GB18
ATI Radeon HD 4870 512MB16
ATI Radeon HD 485018

The cards that sees the least success in moving from single GPU 3-way are the 4870 512MB and the GTX 285. With the 4870 512MB, this is a combination of failures and CPU/system limited situations while the GTX 285 is purely CPU/system limited here. 16 out of 21 tests isn't hugely different than the 18 or 19 out of 21 (20 for NVIDIA cards), but we do see less "success" in general as compared to our two card situation. And keep in mind that this is 33.3% out of a possible performance improvement of 200%. We are being less strict and seeing less success.

Now lets look at complete failure of scaling from 1 to 3 GPUs. This is based on scaling of <5% and ends up catching the cases of negative scaling. While we did this on a per game basis for 2-way scaling, this time we are looking at the results out of the total number of tests (out of 21 tests for AMD cards and out of 20 tests for NVIDIA)

NVIDIA GeForce GTX 2850
NVIDIA GeForce GTX 2800
NVIDIA GeForce GTX 2600
NVIDIA GeForce 9800 GTX+1
ATI Radeon HD 4870 1GB0
ATI Radeon HD 4870 512MB2
ATI Radeon HD 48503

Again, this is generally failure (negative scaling) at 2560x1600 and is generally an issue we can attribute to 512MB of RAM not being enough at high resolution. There are some differences here as compared to 2-way scaling, but generally this isn't that many cases of abject failure to contend with. We'd still love to see AMD and NVIDIA implement something that caught multiGPU failure and reverted to running on a single card in those cases rather than producing a negative experience. But since we can manually disable both SLI and CrossFire, this isn't a deal breaker (it's just an annoyance).

When we look at scaling up from 2 to 3 cards, things get a bit more dim. Since the maximum scaling percent is 50%, we decided to lower our bar for calling a configuration "successful" by reducing our threshold to 10% (which is very generous at only 1/5th of the theoretical maximum). Our results show that much reduced improvement when moving from two to three cards. Here's the data on the number of "success" (>10% improvement) we saw when going from 2 to 3 cards:

NVIDIA GeForce GTX 28511
NVIDIA GeForce GTX 28012
NVIDIA GeForce GTX 26014
NVIDIA GeForce 9800 GTX+12
ATI Radeon HD 4870 1GB13
ATI Radeon HD 4870 512MB8
ATI Radeon HD 485014

With the best success rate coming in at 14 out of 20 with the GeForce GTX 260 3-way SLI, and our threshold for success so low, 3-way isn't looking so great out of the gate. Let's take a look at failure to round that out. We'll consider failure to be <2.5% scaling.

NVIDIA GeForce GTX 2856
NVIDIA GeForce GTX 2805
NVIDIA GeForce GTX 2607
NVIDIA GeForce 9800 GTX+7
ATI Radeon HD 4870 1GB5
ATI Radeon HD 4870 512MB9
ATI Radeon HD 48504

These numbers show that the majority of the time that cards don't scale up well from 2 to 3 GPUs, they either make no statistical difference or they degrade performance. This is in contrast to scaling from 1 to 2 GPUs. So despite the fact that 3 GPUs can offer good improvement over 1 GPU, it doesn't seem that 3 GPUs consistently offers good improvement over 2 GPUs.

But this is the high level overview. Let's take a look at each game test to get a better idea of what's going on. First we'll recap prices and the test setup and then we'll get to the analysis.

Index Prices, Stutter, and The Test
Comments Locked

46 Comments

View All Comments

  • Snarks - Wednesday, February 25, 2009 - link

    hmm, i find my self questioning these articles more and more..

    but anyway carry on.
  • DerekWilson - Wednesday, February 25, 2009 - link

    what's the question ... seriously, any criticism is helpful. this is the first time we've really done a series like this, and it's a complicated situation with lots of data and lots of analysis ... there's no one way to look at it, and all the feedback i get will help me down the road.

    i don't see the need for this type of article or series very frequently, but we'll have to do it every once in a while just in case something changes. knowing what you guys think is important and what you guys want to read about is key to us getting things done right.
  • Flyboy27 - Wednesday, February 25, 2009 - link

    Sell you an extra card that you don't really need.
  • Flyboy27 - Wednesday, February 25, 2009 - link

    oh yeah... and a more expensive motherboard, power supply, and case.
  • Burrich - Wednesday, February 25, 2009 - link

    Would the recently release Catalyst 9.2 drivers improve any compatibility or fps issues? Their release date was 2/20.
  • 7Enigma - Thursday, February 26, 2009 - link

    Check out xbitlabs' review of the 9.2 drivers. If you have a 4870 X2 then yes it appears to be a nice upgrade for several games with minimal losses in the games it doesn't benefit. But if you are sporting a single 4870 1gig it actually degrades performance more than it improves!

    On the flipside they claim stability is better with the 9.2's so it depends on what you want/need. If you are comfortable with the framerates in the games you currently play then jump on the 9.2's for stability reasons. If you are on the edge of playable performance I would stick with the previous drivers...

    http://www.xbitlabs.com/articles/video/display/cat...">http://www.xbitlabs.com/articles/video/display/cat...
  • DerekWilson - Thursday, February 26, 2009 - link

    That article compares 9.2 to 9.1 ... the 8.12 hotfix would show similar performance improvements over the 9.1 drivers. 9.2 does benefit more games, but these are games that have been more recently released than the ones we tested.

    if they compared the 8.12 hotfix to 9.2, we would expect to see more parity, especially with the games we tested in this article.
  • DerekWilson - Wednesday, February 25, 2009 - link

    The recently released 9.2 catalyst drivers are basically the 8.12 hotfix drivers with some additions to support performance and scaling in recently released titles. So not really.
  • smartalco - Wednesday, February 25, 2009 - link

    I don't like that you use 0 for those that score under 25 FPS, specifically because that is under 25 at the res/settings you use. If a card scores 24 FPS at 1680*1050 with maxed settings, what that really tells you is that if you were to drop to half the AA, or turn down some other setting, is that you could still have a perfectly playable game. It seems to me, that giving them a value rating of 0 is acting like everyone has to play on max settings, and if it doesn't meet the standard, its useless.

    IDK, just me talking, I'm going to be happy with my 4850 for quite some time.
    Still an excellent article.
  • DerekWilson - Wednesday, February 25, 2009 - link

    i've actually got the graphs without the 0 scores in the article front to back -- just commented out at the moment ... i wasn't sure which one to go with until the last minute, and i thought about putting both in (but that wouldbe really redundant for games that no card had trouble with)

    i could do some more complex web programming, but i'm not a web developer and i hate javascript ...

    thanks for the feedback. i'll be taking it into account in the final article on 4-way.

    also, if you wanna see the value numbers for the single and dual cards that scored less than 25 fps, you can still look at the first article and see them.

Log in

Don't have an account? Sign up now