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
7Enigma - Thursday, February 26, 2009 - link
http://www.anandtech.com/mb/showdoc.aspx?i=3506&am...">http://www.anandtech.com/mb/showdoc.aspx?i=3506&am...Just a quick review (would have preferred to see the upper resolutions tested....wait did I just write that?), but shows that with CF you have a pretty competetive war. I would guess going tri-gpu (if the game supports it well) the i7 would pull ahead by quite a bit, especially when OC'd, but that remains to be seen.
Mr Roboto - Thursday, February 26, 2009 - link
BORING! Not to sound like a total dick but seriously do we need 2 articles (with probably a third to come discussing quad GPU) to find out what a total waste multi GPU is? I mean until NV and ATI invest some serious research into hardware based SLI\XFire I'd never even consider buying two GPU's. The current software based multi GPU setups don't offer enough of a reward. Having to use a software based profile means you'll always be waiting for them to create one for the latest games, while some simply won't get one at all.Software in general is lagging so far behind current hardware that there is practically ZERO use for two GPU's let alone 3-4. Someone please name 10 games that can successfully use more than 2 cores at a time. OK now name 10 software applications that do the same (not counting video encoding apps). It's pathetic how bad it is. Look at all the XBox360 ports built on 5 year old hardware that are coming out in rapid succession and you have a pretty glib picture.
Still waiting for some Lucid Hydra benchmarks showing 100% linear scaling. If it actually materializes on the desktop then I would consider multi GPU.
7Enigma - Thursday, February 26, 2009 - link
I find these articles well done and beneficial. The stereotype that tri/quad/sometimes even dual is a waste is just that a sterotype. For games that are coded well for multi-gpu and when you are on a large display I'm actually a bit impressed with the multi-gpu stats. A year ago you would likely not see even near the improvement from a single card. It's a testament to the drivers of the gpu manufacturers and as importantly the game programmers.What this also shows is games that are poorly suited for multi-gpu (either due to poor coding/drivers or bottlenecked at another point). Sure most of us don't have the 30" display that warrants the tri and upcoming quad articles, but for those that do it is not a waste, and look further down the rabbit hole to see where we will be in another 2 years or so with a single-gpu card.
We have already pretty much approached a saturation point for mainstream graphics gaming (<$300) under 24" (1920X1200). That has to be dangerous for ATI/Nvidia because we are now becoming limited by our display size as opposed to the historical trend of ever demanding games keeping us at pretty much the same resolution year after year (with slow progress from 800X600 to 1024X768 to the now most common 1280X1024, to the quickly becoming gaming standard 22", and finally within a year or two to 24" gaming goodness).
What I imagine you will see in the next year or two is heavy reliance on physics calculations on the gpu to burden them greatly bringing us back to the near always present gpu-limited status. If game makers fail to implement some serious gpu demands above and beyond what is being used today pretty soon we won't need to upgrade video cards near as frequently. And for the consumer that is actually a really good thing...
Pakman333 - Thursday, February 26, 2009 - link
"BORING! Not to sound like a total dick"You do, and you probably are.
If you don't care, then why read it? No one is forcing you to.
poohbear - Thursday, February 26, 2009 - link
props for posting such a quick follow up! sometimes hardware sites promise follow-ups & never deliver.;)Jynx980 - Thursday, February 26, 2009 - link
Nice job on the resolution options for the graphs! This should alleviate frustration on both ends. I hope it's the standard from now on.rcr - Thursday, February 26, 2009 - link
I read some times ago, that the GTX285 will draw less power than a GTX280. But the graph of power consumption is here also made wih 3xGTX280OC.ilkhan - Thursday, February 26, 2009 - link
As a reader I hate going back and forth between pages. Id love to see the price for each option on the right hand side of the performance graphs. Would make it super easy to compare the value of each solution, rather than the seperate graphs you are using here.And on the plus side for 4 way scaling, at least theres only 4 cards to bench (9800GX2, 4850X2, 4870X2, and the GTX295). Maybe a couple more for memory options, but a lot less than the 3-way here.
LiLgerman - Wednesday, February 25, 2009 - link
i currently have a 4870, 512mb TOP from asus and i was wondering if i can pair that with a 1gb model. Will i have to down clock my 4870 TOP to the speed of the 1gb model if i do crossfire?Thanks
DerekWilson - Thursday, February 26, 2009 - link
AMD is good with asynchronous clocking (one can be faster than the other) ... but i believe memory sizes must match -- i know you could still put them in crossfire, but i /think/ that 512MB of the memory on the 1GB card will be disabled to make them work together.this is just what i remember off the top of my head though.