Triple-GPU Scaling: AMD CrossFire Vs. Nvidia SLI

[ubercake]

[citation][nom]lemlo[/nom]No my stance remains the same. They are trying to say the cheaper cards are better with their carefully crafted set of guidlines when in fact the general performance will, in my personal opinion, show otherwise. It's not the first time I've seen it and Tom's isn't the only site who does it. I think most do to some extent and I think it's rather misleading.[/citation]
I concur.

 

[ngoy]

Two GPUs are great, but are three that much better? When it comes to multi-card scaling, can AMD finally beat Nvidia? Who really needs this much performance? We loaded a super-fast system in single-, dual-, and triple-GPU configurations to find out.

Triple-GPU Scaling: AMD CrossFire Vs. Nvidia SLI : Read more

Although you mention a Nvidia Physx bias in the 3dmark benches, I would be more interested in how this benchmark set works with all GPU physics disabled. Nowhere in the article do you mention if you modified the physics settings or what affect they would have on the gaming benchmarks. Without knowing where the processing load is occuring, it makes this benchmark set arbitrary at best.

Shango

 

[Terrace]

What motherboard is used for the Crossfire setup? the ASUS P8P67 WS Revolution listed on the specs page is for SLI only isnt it?

 

[lemlo]

No, that motherboard can do xfire and sli.

 

[hixbot]

How about an article devoted to the still persistent issue of microstuttering?

 

[casperstouch]

Is it just me, or is this the type of rig that we want to see in a TOMS Hardware Giveaway?

 

[shineon2010]

This is a great setup considering you can get two 6950's off amazon for under $500

 

[Crashman]

[citation][nom]lemlo[/nom]No my stance remains the same. They are trying to say the cheaper cards are better with their carefully crafted set of guidlines when in fact the general performance will, in my personal opinion, show otherwise. It's not the first time I've seen it and Tom's isn't the only site who does it. I think most do to some extent and I think it's rather misleading.[/citation]Here's the thing: Tom's didn't "do it". Tom's went into the article expecting SLI to scale better and, happy day ATI fans that CrossFire scaled better. Had this been a conspiracy the editor would have replaced the benchmark that didn't respond to 3-way CrossFire...but please don't let the facts get in the way of your opinion.
Having stated the facts, you definately will find different scaling results at different CPU clocks using different platforms. ATI seems to be more CPU-reliant and more PCIe-restricted than Nvidia, according to numerous tests on this very site.[citation][nom]ngoy[/nom]Although you mention a Nvidia Physx bias in the 3dmark benches, I would be more interested in how this benchmark set works with all GPU physics disabled. Nowhere in the article do you mention if you modified the physics settings or what affect they would have on the gaming benchmarks. Without knowing where the processing load is occuring, it makes this benchmark set arbitrary at best.Shango[/citation]3DMark 11 gets rid of the PhysX bias found in previous versions, and was unmodified. Since most of the 3DMark data you'll find on the web is also from unmodified tests, the results are only as arbitrary as the test itself.

 

[10hellfire01]

[citation][nom]amk09[/nom]i wish i could afford that[/citation]
Same here. Wish I could get a tri-Gtx580 or tri-6970 setup with water cooling...

 

[littleleo]

I find these test puzzling, customers keep bring the AMDs back and asking for nVidia complaining about the artifacts on the screen and poor drivers. Not just local customers but customer all over the world. Yet you guys seem to have no driver issues at all. Is AMD providing special drivers for you guys for these test? I have them DL the latest drivers yet they all complain and want to switch to nVidia, why is that?

 

[cburke82]

I find these test puzzling, customers keep bring the AMDs back and asking for nVidia complaining about the artifacts on the screen and poor drivers. Not just local customers but customer all over the world. Yet you guys seem to have no driver issues at all. Is AMD providing special drivers for you guys for these test? I have them DL the latest drivers yet they all complain and want to switch to nVidia, why is that?

Im guessing by the reply that you work at some sort of computer/electronics store? I recently purchased a XFX HD 6870 Diamond 6950 and XFX 6950 (ought the 6870 wanted more swaped for the diamond didnt unlock stable so I purchased the XFX lol). All 3 cards worked great so Im guessing that there is no need for special drivers. I think there is a possibility that AMD does ship there cards closer to max clock speeds however. I say this because all 3 cards where perfectly fine but if I used MSI Afterburner to OC them INSTANT artifacting. So if you get alot of customers that are into OCing there cards being as the three cards I just bought all had that issue to some extent im sure that is not coincidental. But im sure Tom's most likly does most tests at stock settings so therefor no issues. My unlocked 6950 worked fine in all games however I did get a few little green spots wile playing Crysis so I went to stock mem speeds.

 

[Guest]

I would like to see Crossfire Scaling on an AMD 6-core platform vs. a SB platform.

 

[Guest]

You're comparing two different value cards. You should compare the 6970 with the GTX570 and not the 100€ cheaper 6950...
Shame...

 

[Crashman]

[citation][nom]bgmordred[/nom]You're comparing two different value cards. You should compare the 6970 with the GTX570 and not the 100€ cheaper 6950...Shame...[/citation]This has been covered in the article and like, six times in this thread. Nvidia doesn't make a cheaper card (in current models) that supports 3-way SLI. This is "cheapest vs cheapest" for current models, though I'd be happy to retest if you would simply send the 6970's.

 

[pandemonium_ctp]

I'm glad to see that tri-scaling is as good as bi-scaling for AMD. I recently just completely a table based off of Anandtech's information about GPU efficiency with the latest cards out on the market and my findings also coincide with those here: that the 6950 2GB rules!

(To be posted as a follow-up to my thread regarding costs to run a computer: http://www.tomshardware.com/forum/forum2.php?config=tomshardwareus.inc&cat=31&post=297910&page=1&p=1&sondage=0&owntopic=1&trash=0&trash_post=0&print=0&numreponse=0&quote_only=0&new=0&nojs=0)

 

[ubercake]

[citation][nom]littleleo[/nom]I find these test puzzling, customers keep bring the AMDs back and asking for nVidia complaining about the artifacts on the screen and poor drivers. Not just local customers but customer all over the world. Yet you guys seem to have no driver issues at all. Is AMD providing special drivers for you guys for these test? I have them DL the latest drivers yet they all complain and want to switch to nVidia, why is that?[/citation]
Totally my experience with AMD crossfire (single GPU AMD is great; more than one GPU is flawed). Bad drivers turned me to Nvidia. High FPS don't mean squat when they aren't good FPS.

 

[Guest]

hmmm. I see a problem with the selection of the hardwares. Sorry:
GeForce Graphics Asus ENGTX570/2DI/1280MD5: 742 MHz GF110 GPU, 1280 MB GDDR5-3800 ----> LOWER CLOCK/HALF THE MEMORY

Radeon Graphics XFX HD-695A-CNFC: 800 MHz Cayman GPU, 2 GB GDDR5-5000
----> SLIGHTLY HIGHER CLOCK/ DOUBLE THE MEMORY

couldn't that be enough to explain why Nvidia suffers only in the highest specs and still beats Radeon in the other ones?

I wonder if one should not pay attention to that even with the argument that the price range was the reason for the choices.

 

[Guest]

So the AMD cards see a bigger gain. Kind of like how cold water heats faster when put on to boil.

 

[Crashman]

[citation][nom]AugustRedd[/nom]So the AMD cards see a bigger gain. Kind of like how cold water heats faster when put on to boil.[/citation]No it doesn't. 99° water will reach 100° faster than will 15° water.

 

[Crashman]

[citation][nom]aventura_alex[/nom]hmmm. I see a problem with the selection of the hardwares. Sorry:GeForce Graphics Asus ENGTX570/2DI/1280MD5: 742 MHz GF110 GPU, 1280 MB GDDR5-3800 ----> LOWER CLOCK/HALF THE MEMORYRadeon Graphics XFX HD-695A-CNFC: 800 MHz Cayman GPU, 2 GB GDDR5-5000 ----> SLIGHTLY HIGHER CLOCK/ DOUBLE THE MEMORYcouldn't that be enough to explain why Nvidia suffers only in the highest specs and still beats Radeon in the other ones?I wonder if one should not pay attention to that even with the argument that the price range was the reason for the choices.[/citation]That might explain some of the results. The only problem is that the card with more memory, that scaled better and offered better overall performance in 2-way and 3-way configurations, also costs less. It's like AMD is paying you to take the extra RAM.

 

[amstech]

The reviewer makes a statement in his conclusion about Radeon's supremecy with scaling and performance.
It looks good and all, but with all the driver issues, countless bugs/glitches and problems AMD has, they would need to stomp Nvidia handidly to make up for thier lost ground with reliability/stability.

The 570's/580 scale very well themselves, have superior tessellation and PhsyX right out of the box. I haven't had any issues with games like I did with my 5870 and CrossfireX 5770 setup.

 

[cburke82]

The reviewer makes a statement in his conclusion about Radeon's supremecy with scaling and performance.
It looks good and all, but with all the driver issues, countless bugs/glitches and problems AMD has, they would need to stomp Nvidia handidly to make up for thier lost ground with reliability/stability.

The 570's/580 scale very well themselves, have superior tessellation and PhsyX right out of the box. I haven't had any issues with games like I did with my 5870 and CrossfireX 5770 setup.

I dont know much about the 5XXX series cards I have only been gaming on a PC for 2 months now so admittedly im a noob lol. However in my amazing ability to not be happy with what I have I ended up using 3 AMD cards in a month lucky for me frys allows returns lol. So in the first month I used a XFX 6870 a Diamond 6950 and a XFX 6950. All three installed quick and easy with no issues and played Crysis just fine. So maybe the fact that ATI is no more and AMD took over there will be less issues? All I know is reading your post one would think using a AMD card is a guaranteed nightmare with drivers and stability wile I have seen the exact opposite. I think the fact that someone who has never built a PC before chose AMD cards and was able to very easily get things up and running means something.

 

[Guest]

The AMD crossfire setup is still limited to 2 cards like for the alien vs predator game, we see that at lower resolutions, so we shouldn't take too much attention to the 3 way-xfire result at a 2560*1600 resolution like for the alien vs predator

 

[cburke82]

The AMD crossfire setup is still limited to 2 cards like for the alien vs predator game, we see that at lower resolutions, so we shouldn't take too much attention to the 3 way-xfire result at a 2560*1600 resolution like for the alien vs predator

Based on the results of this test the ammount on improvement you get in these games with Nvidia or AMD cards is no good enough to justify a 3rd $300 card IMO lol The second card makes a huge diff and the 3rd well not so much

 

[Guest]

Just Curious why it's not the 6970 you are comparing being it's cheaper than the 570 now?