MultiGPU Update: Does 3-way Make Sense?
by Derek Wilson on February 25, 2009 2:45 PM EST- Posted in
- GPUs
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 285 | 16 |
| NVIDIA GeForce GTX 280 | 18 |
| NVIDIA GeForce GTX 260 | 19 |
| NVIDIA GeForce 9800 GTX+ | 19 |
| ATI Radeon HD 4870 1GB | 18 |
| ATI Radeon HD 4870 512MB | 16 |
| ATI Radeon HD 4850 | 18 |
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 285 | 0 |
| NVIDIA GeForce GTX 280 | 0 |
| NVIDIA GeForce GTX 260 | 0 |
| NVIDIA GeForce 9800 GTX+ | 1 |
| ATI Radeon HD 4870 1GB | 0 |
| ATI Radeon HD 4870 512MB | 2 |
| ATI Radeon HD 4850 | 3 |
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 285 | 11 |
| NVIDIA GeForce GTX 280 | 12 |
| NVIDIA GeForce GTX 260 | 14 |
| NVIDIA GeForce 9800 GTX+ | 12 |
| ATI Radeon HD 4870 1GB | 13 |
| ATI Radeon HD 4870 512MB | 8 |
| ATI Radeon HD 4850 | 14 |
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 285 | 6 |
| NVIDIA GeForce GTX 280 | 5 |
| NVIDIA GeForce GTX 260 | 7 |
| NVIDIA GeForce 9800 GTX+ | 7 |
| ATI Radeon HD 4870 1GB | 5 |
| ATI Radeon HD 4870 512MB | 9 |
| ATI Radeon HD 4850 | 4 |
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.
Facebook
Twitter
RSS
46 Comments
View All Comments
magnusr - Saturday, February 28, 2009 - link
Got my second 4850. Huge fps upgrade in 1920x1080. But noise during gaming has gone up. Idle just a bit.It would have been nice if you added noise numbers. Since some would probably prefer getting a fast single gpu like nvidia 285 instead of 2x4850/4870 due to possibly lower noise levels.
taltamir - Saturday, February 28, 2009 - link
the following statement is made:It's very difficult to really collect high quality quantitative data that shows microstutter, as the only way to really get a good idea of what's going on is to analyze raw frame data on a per frame basis (which you can't get with FRAPS)
This is incorrect, FRAPS can capture those, its in the settings, it gives exact MS "time" of each frame, and it gives them in order, in a csv file. You can create a function showing the difference between each pair, giving you the ms per frame, and you can analyze that data.
please check this thread in anandtech:
http://forums.anandtech.com/messageview.aspx?catid...">http://forums.anandtech.com/messageview...=31&...
kpgoebel - Friday, February 27, 2009 - link
I just finished reading both pieces - great stuff. the analysis was dense but your pace kept it moving.I use three 20" samsungs for my home workstation that doubles as a flight sim trainer since I took up flying. i added two 880GT's to a Q6600 quad core system, in hopes that it could handle triple 1600X1200 screens for accurate flight control response. it's only proven reliable with the textures turned down to medium or low.. which is a bummer because i've invested in some top-notch sceneries.
any chance you guys will offer some analysis for multiple monitor gaming setups?
orionmgomg - Thursday, February 26, 2009 - link
Thank you for this great article!So much information, and very helpful to get an idea of the where to spend money on a build and where the price/value margins lie.
I cant wait for the Quad shoot out!
PS: Does anyone know if the ASUS Rampage II Extreme is the only X58 Mobo available right now that fully supports X3 PCIe 2.0 x16?
The Other ASUS boards say they support: PCIe 2.0 x16/PCIe 2.0 x16/ or PCIe 2.0 x16/PCIe 2.0 x8/PCIe 2.0 x8...
Contemplating 3 Way SLI...
Zorro3740 - Thursday, February 26, 2009 - link
For triple and quad cross to really be worthwhile a high refresh rate LCD can be a great way to benefit from the huge headroom in performance given from multi GPU setups. I have several PCs with triple crossfire Radeon 4850s on 60 hertz displays. A game like Devil May Cry 4 comes to mind. If you can sustain that ultra high 120 frame limit of these new displays the value of multi GPU will suddenly become that much more apparent. I sure could easily with my Phenom 2 920 or I7 920. With max settings at 1080p you get as much as 200 frames or more. It just moves so smooth at frame rates higher than 60. I imagine it would be similar to the 99 hertz refresh rate of my 24" CRT at 1080p. I stopped using that display because of its dimness but it gave me a damn good idea of what a real fast 120hz LCD might do. How about 120 FPS with 24 sample FSAA in Devil May Cry 4? I'm sure there are many instances where 120 hertz will make a huge difference for Multi GPU enthusiasts.I highly recommend that somebody test their multi GPU rig with a 120hz LCD HDTV and tell us about how smooth the action is and how clear it is with higher levels of FSAA. It is a shame that 120hz is not here in the PC monitor world. I have the HP LP3065 that does 2560 * 1600 @ 60hz. I would trade that in for a 120hz 1080p or 1200p LCD model in a flash if I could if it had a high quality backlight. I HATE that the world is becoming a TN world when it comes to monitors. We need 120 hz 8 bit panels and we need to just get the hell away from shoddy 6 bit monitors with light bleed among other unacceptable faults.
Anyhow I really enjoyed the article and confirmed to me that for the games I play the 4850 is a good multi GPU choice. I just would have liked to see more of an emphasis on future developments like 120hz LCDs and higher levels of FSAA in particular at "lower" resolutions like 1920 * 1080 and 1920 * 1200 with 512 meg setups and 1 gig setups. Multi GPU 4850s are way more affordable than 30" monitors and some people will definitely "settle" for nice 24" 8 bit panels like the Westinghouse L2410NM that rock at 1920 * 1200.
far327 - Thursday, February 26, 2009 - link
Would like to others comments including authors on the differences between some of the numbers of the charts on Anandtech and TheGuruof3d http://www.guru3d.com/article/core-i7-multigpu-sli...">http://www.guru3d.com/article/core-i7-m...li-cross...I think an overclocked CPU would of been a much more well rounded approach to this article. Even thought the performance gains aren't huge, let's face it, most people hardcore enough to game with 3 GTX 280/285's is probably overclocking there CPU as well.
chizow - Thursday, February 26, 2009 - link
Haven't really dug into the review yet Derek, but meant to suggest this in the last article. Is it possible to do something about the text in the bar graphs? Its just overly cumbersome trying to make sense with all the extraneous text in there. Some suggestions:1) Perhaps you're required to put Vendor + Brand for each config, but can we get an icon for Nvidia or ATI instead of ATI Radeon or Nvidia GeForce for every single line item?
2) Color code the relevant data for each line item to differentiate. For example:
Nvidia GeForce [green]GTX 285 3-way[/green] SLI
ATI Radeon [red]HD 4870 3-way[/red] CrossFire
Would go a loooooong way into making those graphs easier to parse and digest. Back to reading, thanks.
OblivionLord - Thursday, February 26, 2009 - link
What about the thought of 3way 100% scaling with Lucid Hydra?Why not make mention that if that becomes a reality, then you will infact get what you pay for.
DerekWilson - Thursday, February 26, 2009 - link
I love the idea of Lucid and their Hydra technology ... but it isn't here yet. I enjoy writing articles about what their technology could mean and how beneficial and amazing and awesome it would be ... but it's not here yet and we haven't had a hands on chance to see for ourselves that it works as advertised.until it's out, SLI and CrossFire are the best we've got and the focus of our articles.
maybe we'll run into Lucid again at GDC and write up more cool stuff about Hydra though ...
Denithor - Thursday, February 26, 2009 - link
Guru3D did an article at the i7 launch showing how well i7 scales with multiGPU setups compared to C2D & C2Q. Night & day difference, to tell the truth, i7 just pushes the multiGPU rigs much better than the older architecture can.My question is this: how does PhII compare to i7 on the multiGPU front? I'd love to see some benching done at high resolution with 2/3/4 card setups on the two CPU architectures to see how they look.