Triple-GPU Scaling: AMD CrossFire Vs. Nvidia SLI

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
Status
Not open for further replies.
An EVGA 570 is priced @ 350$ on newegg
A Sapphire 6950 2GB is priced @ 277$ on newegg
that's a 156$ less if going for xfire setup, still having a better performance, impressive indeed.
 
Why are you comparing two gpu`s that are not even in the same price range. Accroding to newegg the 6950 is ~240$ and the 570 is ~350$. Multiply that by three or two and you get a lot of price difference.Favoring nvidia are we?
 
Hope NVidia wakes up, seems their former driver superiority in multi-gpu environments have made them lazy! AMD sure have speed up the pace in multi-gpu scaling compared to what their crossfire were a few years ago (often did more harm than good!).
 
This confuses me. So in retrospect, I should choose to couple the less expensive 6950 than the more expensive 570 to get better performance in higher resolutions?
Is sli really that much worse or are these scenarios just happen to favour amd? Sure if they would have included more titles, we would have had a more accurate result. However, the fact that this result is derived from a comparison between 6950 and 570 rather than 6970 and 570, kind of scares me (for the lack of a better word).
 
I hope that the next die shrinks allow for the Sandy Bridge-like power consumption and heat characteristics in GPU's. The 500 series nVidia and the 6000 AMD cards are all pretty good, but not to the point where noise isn't an issue anymore. The 570 may be much quieter than the 470, but I wouldn't guess two or three of them are anywhere near quiet. I ditched my 470 for a 460 "Cyclone" just because of noise. Since then, I'm on my 3rd CPU and second motherboard, but I won't upgrade GPUs until I can find something just about silent. My sanity is worth it.
 
Seriously? With triple crossfire 6970s at 1080p the first 10 minutes of AVP maxed out with a 5.2 GHz i7 2600k is 240 FPS. Crysis scaling? 80-100 with CPU limit sneaking in. It takes any third party with similar setups a handful of minutes to verify. The rest of toms credibility swirls down the drain.
 
I love the complaints on here. The one's complaining about it "not being fair". First of all, "fair" is a relative term. If you wanted to go out and test a Honda civic against a Lambo Diablo, you could. What's unfair about it? Performance/$ is the big end, is it not? the 6950's showed better value and performance/$, which scaled even more so with each card added. I don't see the issue here. Xfire stomped on SLI in most instances, requiring less money, less power, and a little less heat. Life's not fair, but, this article is about as close as you can get, IMO.
 
I really do wish you would have tested Eyefinity/Surround... why would anyone bother with 2 or 3 cards unless they were running 3 displays! I'm waiting to upgrade to SB and Surround but I've been looking for a proper review of surround with 570 SLI. Its a pity cause this article could have encompassed that as well, but now it seems unlikely that you will do another article with these exact same cards just with multiple monitors..

Oh well, other than that, good review
 
The article suggests the AMD cards are much better value for money.

Pity you didn't compare apples with apples ... the 6970 was the weapon of choice there.

Still ... you showed that a cheaper card (6950) in CF holds up very well against the better NV (single).

I think the point you were trying to make by this comparison came across well ... so no criticism from me in that regard ... as you spelled this out in the article.

An honest review.

:)
 
[citation][nom]obarthelemy[/nom]and nobody cares... the %age of multiple-GPU users (and gamers) is what ... 0.01% and 0.1%, respectively ?[/citation]
Wow - how smug, self-important and very assuming. I am very disappointed.

First, 0.01% to 0.1%, or whatever, clearly isn't "nobody," even if it is a small fraction. Though I would believe the number is a bit higher than these. The people reading and writing the comments obviously care ENOUGH. Also this is not the first or ONLY review or comparison of an ultra-high-end multi-card setup, not at Tom's or anywhere else, which have been going on since people have had the option to put them in their machines (now going on 7 years not including the old 3dfx versions from the 90's)

AMD and NVidia seem to find it important ENOUGH to pay engineers and programmers to continually add driver support and Intel/AMD seem to find a reason to add the ability (and increasing number of pci-express lanes) to motherboard chipsets time and again, which is no trivial task.

The fact that most people do not spend $700+ for such a setup does not dull their fascination in it, and at the very least the continuing development of this technology has made it less exotic, more affordable, and a reasonable option to those willing to try it as lower-end graphics boards continue to support it with continually improved scaling and subsequent value over time.

As if the weight of these facts, which are well known to virtually all Tom's readers, however inexperienced, is not enough to flatten you into feelings of foolishness, at the very least this article demonstrates that a three-way setup of this type is far more economical at this point than most of us would have expected.

Continue to post canards of the type, and I shall suggest that you be excommunicated from the PC-enthusiast world.
 
[citation][nom]MichaelTurbo[/nom]Seriously? With triple crossfire 6970s at 1080p the first 10 minutes of AVP maxed out with a 5.2 GHz i7 2600k is 240 FPS. Crysis scaling? 80-100 with CPU limit sneaking in. It takes any third party with similar setups a handful of minutes to verify. The rest of toms credibility swirls down the drain.[/citation]

First, it would be nice to see when and where this was done. Not that it is implausible, but 'I knew a guy whose brother worked with a guy who said he saw..." is the epitome of weasel-wording.

Considering the previous 6990 review with 'borrowed' quad-crossfire data had Crysis pinned at just under 70 FPS @ 2560X1600, the numbers here seem reasonable enough.

Also, this testbed at Tom's, a very high-end Sandy Bridge machine moderately overclocked is accessible to the many enthusiasts willing to spend a lot of money, and represents a fair goalpost for keeping CPU limitation to a minimum at very high resolutions. 5.2GHz is much harder to obtain and maintain without extreme cooling methods and while it is ideal to measure the effect with maximal CPU performance possible the author circumvents this by staying in the higher resolution range which he clearly identifies.

1080P? Well, that doesn't strike me as 'high,' though it is adequate to be sure. The incentive is based on the 2560X1600 and potentially surround-gaming setups with similar pixel counts, where multi-gpu rigs actually become necessary.

The purpose of the article is to determine the possible improvements of AMD's triple-card scaling compared to NVidia's well established implementation, using boards suited for the job. The choice of 6950 may not have been optimal, since 6970 is more comparable dollar-for-dollar, but by no means does it diminish the significance of the results.

For the most part, folks like Chris and Thomas are under sufficient scrutiny and are likely compelled to put together an honest review when they do, even if parts of it are arguably imperfect in some people's opinion.

If the credibility of the authors at Tom's were apparently as much a function of the worldview and opinions of the readers like yourself, based on your criteria here for why this article lacks 'credibility,' nobody would believe anything anywhere from any hardware review sites at all.

IMO there are a few holes in the 'sink' you have for interpreting the results of this write-up, if that is what you really believe.
 
I think what this article shows, more then anything, is AMD's gain in scaling for the latest generation. Not so much that Nvidia is bad, just how much better crossfire has come along then what it was. (surpassing Nvidia's scaling now??).

I have no doubt that if this article included an Evergreen GPU set-up, you would see that crossfire was not so good at scaling. That is a major thing for AMD to acomplish from one generation to the next. Bravo AMD.

I agree with others. The lowest level cards from this current generation, from either company, that can do triple card set-ups are the GTX 570 and the HD 6950. Because dont forget the HD 6850/6870 only allowed two cards in crossfire...just like the GTX 560. So this article shows the least your spending for tri-card setup from each company. I think that is a good test.

I have a HD 6950 right now and for a single card it is pretty darn good. I think I may add a second when BattleField 3 comes out. But, for me.. another thing this article showed, was a third card.. from either company.. just doesnt make sense when you look at the $$/performance ratio.
 
Seriously? With triple crossfire 6970s at 1080p the first 10 minutes of AVP maxed out with a 5.2 GHz i7 2600k is 240 FPS. Crysis scaling? 80-100 with CPU limit sneaking in. It takes any third party with similar setups a handful of minutes to verify. The rest of toms credibility swirls down the drain.
Click to expand...




But, 1080p is the best resolution my console can do, and we both know consoles are at the cutting edge, so that has to be the epitome of gaming.

/sarcasm
 
This makes me wonder if the gains in CrossFire scaling are due to a hardware change in the new 6000 series chips or just improved drivers. Because I know the 5870s in my machine do not scale nearly as well as the 6900 parts 🙁 (and Tom's proved as much comparing SLI and CF for the GTX 400 vs Radeon HD 5800 GPUs)
 
Man, it's too bad you did this review before the 11.4 Catalyst drivers were released... there's some very substantial performance upgrades included with it.

Very interesting review though. A lot of people bash ATi/AMD drivers, especially their crossfire performance, but it seems that all the trash talking had no real proof to back it up.
 
The 6950 is not the cheapest 3-way crossfire configuration. There are some 5770s with two connectors to allow up to 4-way crossfire, and the same goes for 6850s, and 6870s. I suppose if you're strictly talking about reference designs though then yes the 6950 would be the lowest option considering the 5800 series is being phased out.
 
We can draw a simple conclusion: Use the best suited card for the game you love to play and always use the latest drivers.
 
Status
Not open for further replies.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom