AMD CrossFire Vs. Nvidia SLI Scaling Analysis

[eyefinity]

Not even 1% of people use 2560x1600 so who cares.

 

[eddieroolz]

A very predicatable result. Nvidia has been good with SLI performance improvements. This is where AMD lags quite a bit behind.

 

[kikireeki]

The bottom line is:
A second high-end card is a waste of money!

 

[rhino13]

I think you're seeing the memory size difference more than anything as you go to the 2560x1680 resolutions.
Could you also run this with crossfire 5280s and see if you have a difference?

 

[scook9]

I still feel pretty good about my CF 5870s since I got them for GTX 470 prices

I was considering some GTX 470s but did not feel the need to have nvidia that bad this go around (and I was coming off dual GTX 275s)

I can say though that 5870 CF on a mATX board gets just as hot as the fermis lol (93c on each card in FurMark - reference coolers)

 

[pcman911]

The only other thing I wish we could have seen here was the midrange cards in CF/SLI as well. A pair of 5770 and the GTS 450. The 5770 1GB GDDR5 at around $125ea/$250 pair and the GTS 450 $110ea/$220 pair.

 

[wolfram23]

Nice review. I hope to see another one comparing triple monitor config performance!

 

[Onus]

I know you won't be able to use the same mobo, but I'd REALLY like to see this done with an AMD CPU. After AMD/SLI did so poorly in a recent article (sorry, I don't remember which one!), I'm very curious to know if there's a chipset issue on the 980a boards that strongly disfavors using AMD if you want a pair of nVidia cards.

 

[tom thumb]

And I must confess I barely sleep thinking of what golden deity AMD will let me waste my money on in just a few days.....

Me too.

... I want it now dammit! I can't wait 5 days.

 

[K2N hater]

There could be an article about 4-way SLI and hacked drivers. Wonder how much powerful 4x260 or 4x285 would be compared to the newer chips.

 

[masterasia]

How about my GTX 260 SLI? Do I need to upgrade yet?

 

[scrumworks]

[citation][nom]ares1214[/nom]Nvidia did well, but thats to be expected, whats more impressive is how much better ATi has gotten! They use to be mediocre with this, and i was really hoping they would overome it with HD6OOO, but it looks like they dont have to improve much.[/citation]

Yeah. It would show up even better if Tom would use a bit fresher drivers than 10.7 betas.

 

[scrumworks]

[citation][nom]eyefinity[/nom]Why do this 5 days before the new 6 series is released?Will you be updating the results then?[/citation]

Nvidia asked them to do it.

 

[ohseus]

[citation][nom]cangelini[/nom]I'd like to hear some discussion on the best way to reliably test WoW, actually.[/citation]

I am not sure why TwoBoxer got down voted for asking a very valid question.

Testing WoW is very difficult to benchmark as there are so many factors influencing it's performance and these factors are inconsistent. Lag, zone or if thew guy next to you is casting an effects heavy AOE spell can all affect frame rates. none of these are factors a tester would have control over. Additionally with Cataclysm the graphics requirements take a bit of a bump up. The most important time for good frame rates tend to be raids where there are many players and spell effects while response time and awareness is very important.

My suggestion for finding the right WoW/Cataclysm card. Determine the resolution that you intend to Wow on. Find a card that gives "good" or "great" performance "in most games" according to the most recent "Graphics Card For the Money" article. WoW tends to be less video intensive than other games so "good" for a more graphics intense title should do great for WoW. A card that reads "great" for your resolution will likely allow you you play with the graphics on "ultra" Remember with Cataclysm comes some DX11 features. If that is an important factor for you playing than be sure that the card you chose supports it. (I wouldn't suggest a card that does not support DX11 personally)

 

[RazberyBandit]

Chris (or Don), why are you still using Catalyst 10.7 Beta drivers when there have been 2 official releases (Cat 10.8 & 10.9) since?

I can't help wondering if you began this comparison roughly two months ago, or in an effort to save time, re-used results from a test (or tests) done at that time instead of testing all the components again. It would make sense to use the same driver(s) from past tests in order to maintain a sense of fairness, but doing so doesn't reflect a present-day scenario. Instead, it's indicative of results users would have seen 2-3 months ago.

Every time I read the driver notes from both camps, card performance is getting tweaked and improved upon in at least one game, and usually several. To use anything except the most recent driver versions in a head-to-head seems... dishonest?

Also worth noting, there's a brand new Forceware driver package (260.89 WHQL) as well. Released today, I know you couldn't use them. However, the accompanying notes indicate multiple performance improvements for two of the three nVIDIA cards tested. These drivers could easily change the SLI-scaling results of these cards because many of the improvements noted for them are specific to games & resolutions used in this comparison.

 

[JamesSneed]

Dito on the update of this when the AMD 6xxx seris comes out. Im expecting AMD to make some progress with Crosfire scaling but in either case it would be nice to know right off the bat.

 

[WheelsOfConfusion]

Hey Cleeve, what's up with the drivers being two versions (+ beta) behind for both companies? I didn't think these benchmarks took that long.

 

[RazberyBandit]

One other thing. I'm growing tired of "simulated" results. By that I mean the usage of a 5870+5850 and the ASUS Ares to "essentially" become an intended resulting configuration. I know you have the best of intentions, but it's simply not the exact same thing. (Same thing goes with simulated CPUs via core-disabling, overclocking/underclocking.)

If you can't test the actual configuration, stop right there and don't go any further until you can. Simulated results are just that - simulated - not the real deal.

 

[bystander]

I was quite shocked to see the scaling was almost identical, as long as you didn't go to the extreme resolutions, which is likely impacted by VRAM. It would be nice to see the results with the 2GB Radeon cards.

I also was disapointed to not see benchmarks with AA enabled for Crysis and Starcraft 2.

 

[billiardicus]

I don't like the way you put the graphs together. I'm interested in seeing how much gain you get for SLI vs just the single card. I wish you would have arranged the graphs to make this information easy to see. The last graphs are good though. It shows you where to spend your money if you want 2560x1680...but I don't have a 30" so....

 

[Guest]

Since things are being ranked in price order, a brief aside (or graph) noting the prices would have been welcome. Those of us not currently shopping don't track them on a daily basis!

 

[damasvara]

[citation][nom]Twoboxer[/nom]12 million people with WoW subscriptions, new expansion due 12/7/10, lot's of vid cards being bought for that purpose . . . I know its not easy, the format may have to be different, but can you guys consider including WoW in the benching runs?[/citation]
[citation][nom]ohseus[/nom]I am not sure why TwoBoxer got down voted for asking a very valid question.[/citation]
All I can say is, ignorance knows no limit... Unless the question is some stupid "can it play Zynga's Poker?", there's no need to look down on it. It was a very reasonable question.

Shallow minded people... Thumb me down if that makes you happy

 

[IM0001]

As for the SC2 benchmark. I would recommend a retest with the latest ATI 10.10 Beta drivers. I got a huge performance bump with my single 5870 from the 50fps high usual without AA at the start of a level. To about 100fps+ at the start of a level without AA. With AA it starts at about 70FPS and remains playable the whole game. (1680X1050 res)

 

[IM0001]

Also in my test SC2 is at Ultra settings. Just to be clear. (And 4X AA when using AA)

 

[lashton]

here we go another nvidia article yawn, is it me or do these get boring