Triple-GPU Scaling: AMD CrossFire Vs. Nvidia SLI

[Crashman]

This is a good article in concept, but the problem is I could write the same article ..... that is someone pick the desired, (ATI or nVidia as winners) and one can easily "prove" the desired conclusion. Then I can do it again and get the opposite conclusion, simply by handpicking the games in the test.

My issues with this article are twofold:

...... 4 games ???????

....... for consistency, would have liked to see consistency in the test suite so as to compare with past THG articles. Choosing a new set makes results less relevant and perhaps even suspect as to why those 4 games.

5 games, picked for popularity and easily-reproduced results. And there's something fishy about Guru3d's numbers, but ATI cards seem to be more "CPU-bottlenecked" than Nvidia's, so it's probably their lame old CPU getting in the way.

That did give me a great idea for a new article though: CrossFire, SLI and CPU Scaling

 

[K2N hater]

4xAA proves all the hardware tested is very powerful. Even the single cards at the highest resoluton +AA perform beautifully in comparison with the numbers without AA as if their true processing capabilities were never put into test.

There seems to be some CPU-related bottlenecks which may be related to the game engine. With Crysis the hardware is not really under stress on lower resolutions but it's clear nVidia performs much better. At the opposite side lies JC2, which seems also to be bottlenecked but AMD outperforms nVidia.

 

[K2N hater]

Please excuse the grammar mistakes.

 

[alidan]

i have a question.

why is it that the 2 highest resolutions are 1920x1080 16:9 and 2560x1600 16:10 (i believe) with the lowest end being the equivalent of a 4:3 (i believe)

if you are putting a graphics card into a computer i believe that for the two highest, a 16:10 monitor should be use, as in the 1920x1200 instead of 1920x1080. when you are playing a game in the 1920xXXXX resolution, its safe to say you got the monitor because its good, and i believe most people who have ever used a 16:10 prefer it over a 16:9

i know me personally bought the 16:10 specifically for games, also got my little brother one too, i cant be the only person, and if people are looking at new cards, or even the higher end of gaming, that they would at least have a 16:10, as its the next closest resolution to 2560x1600.

im a bit tired, and i don't know if i'm making a great point... ill end it there, but why is it that 1080 was chosen over 1200 for the majority of benchmarks i see.

 

[sinblesser]

TOO BAD THAT NVIDIA SLI CAN'T EVEN RUN DRAGON AGE 2 BECAUSE IT'S SO POORLY CODED BULLSHIT, LAME, fail. lol ^^

 

[kingnoobe]

You people have some serious issues. They used a cheaper AMD card that won this battle. What are you complaining about neither amd or nvidia fanbois have the right to whine in this one.

 

[undead_assault]

hmmm, so in this "unfair" review, we get some cheaper cards are tested against more expensive cards.
But in the end, AMD is cheaper, better scaling, less heat, less power. It is NVIDIA complete defeat. I appreciate what AMD achieved with that superiority, but I'm really sorry that wasn't the case in 6990, we only get more performance.
I'm hoping they will release more efficient cards onward

 

[Crashman]

[citation][nom]alidan[/nom]i have a question. why is it that the 2 highest resolutions are 1920x1080 16:9 and 2560x1600 16:10 (i believe) with the lowest end being the equivalent of a 4:3 (i believe)if you are putting a graphics card into a computer i believe that for the two highest, a 16:10 monitor should be use, as in the 1920x1200 instead of 1920x1080. when you are playing a game in the 1920xXXXX resolution, its safe to say you got the monitor because its good, and i believe most people who have ever used a 16:10 prefer it over a 16:9i know me personally bought the 16:10 specifically for games, also got my little brother one too, i cant be the only person, and if people are looking at new cards, or even the higher end of gaming, that they would at least have a 16:10, as its the next closest resolution to 2560x1600.im a bit tired, and i don't know if i'm making a great point... ill end it there, but why is it that 1080 was chosen over 1200 for the majority of benchmarks i see.[/citation]Because it's getting increasingly harder to find 1920x1200 monitors, so that more "current buyers" have 1920x1080.

 

[yizusc]

why the hell toms tests radeon cards on Catalyst 11.1 ? I wish 11.4 preview was the drivers being used.

 

[mattmock]

Does anyone know much about microstutter on current gen SLI/Xfire systems? I was reading some posts about the problem (on older cards) on the HardOCP forum. Apparently sometimes in multi-gpu systems the frames aren't evenly spaced in time and this causes the game to appear more more jerky than a single GPU system at the same overall framerate. Does anyone with a current gen multi GPU system have this problem?

 

[theusual]

The reason to have the 3 Nvidia cards would be for 3D. I know AMD supports 3D on HDMI 1.4 but I haven't seen any monitors with this available and don't believe the driver/game support for it is there either. I think 3D is more useful than triple monitors, as I don't feel like carting 3 monitors to LAN parties, but that's jut me.

 

[Crashman]

[citation][nom]YiZuSc[/nom]why the hell toms tests radeon cards on Catalyst 11.1 ? I wish 11.4 preview was the drivers being used.[/citation]You're not going to like the answer: Testing began in February for a series of articles that focus on multi-GPU scaling and PCIe performance, and this is the first part.

 

[brendonmc]

3 AMD cards. Certainly would be too loud for Chris!

 

[Guest]

Once again... a great idea done wrong. The selection of games was pathetic and poorly chosen and I'm not sure why 1680x1050 was even included. Sure, the results that the 6950 has amazing 2-way CF scaling is completely spot on, but more games should have been used to test 3-way scaling. Don't really get how it's fair to use FurMark at all when you state that the 6950 was idle with that test... FurMark is known to use more power draw than most games and nVidia doesn't make full use of the GPUs in JC 2.

 

[Crashman]

[citation][nom]draemn[/nom]Don't really get how it's fair to use FurMark at all when you state that the 6950 was idle with that test... FurMark is known to use more power draw than most games and nVidia doesn't make full use of the GPUs in JC 2.[/citation]It's fair, right and just to make observations, since a real game was used for the "real" comparison. Now, looking at the power scaling between 1, 2, and 3 cards, it looks like JC2 was the perfect game for that real-world power comparison.

 

[lemlo]

This article disgusts me. Ask anyone who runs a multi monitor setup with load utilities in the 2nd screen which programs REALLY hammer the resources and which ones REALLY do not. So happens that of the ones tested here nvidia does walk away from amd's offering by quite a bit. On the others... 4x AA seriously? Who in there right mind is going to run 2 or more 570's at 1920x1080 and not run 8xAA if not higher. Even at your 2560x1600 res, common tom's turn the settings up and see who takes the show! Try that on a 2x or 3x 6950 setup and watch the framerates plummet.

 

[lemlo]

Exactly the point I'm trying to make. The xfire cards are scaling a lot better but I'm really tired of seeing the "careful" test scenarios. I have an 875k machine running at 4.2, comparable to a 4ghz HT sandy bridge, with dual 470's and I run higher settings. Most of my games I actually max out on that machine at 1920x1080.

 

[Crashman]

[citation][nom]Lemlo[/nom]Exactly the point I'm trying to make. The xfire cards are scaling a lot better but I'm really tired of seeing the "careful" test scenarios. I have an 875k machine running at 4.2, comparable to a 4ghz HT sandy bridge, with dual 470's and I run higher settings. Most of my games I actually max out on that machine at 1920x1080. Someone can delete the other post I cited the wrong poster.[/citation]You should delete the second post because it refers to an original poster who was at least partly proven wrong, ironically by the post you originally quoted. Here's a link to where you can delete it:
http://www.tomshardware.com/forum/2865-56-triple-scaling-crossfire-nvidia

 

[Crashman]

[citation][nom]kingnoobe[/nom]You people have some serious issues. They used a cheaper AMD card that won this battle. What are you complaining about neither amd or nvidia fanbois have the right to whine in this one.[/citation]Because that's what fanatics do?

 

[lemlo]

[citation][nom]Crashman[/nom]You should delete the second post because it refers to an original poster who was at least partly proven wrong, ironically by the post you originally quoted. Here's a link to where you can delete it:http://www.tomshardware.com/forum/ [...] ire-nvidia[/citation]

ty sir

 

[ubercake]

I liked the high framerates when I ran 3 cards in crossfire too. What I didn't like was the driver issues (flickering video, GPUs staying pegged after gaming, HDMI/DP audio drivers failing to install in later driver releases, etc...) from game to game and release to release. The lack of driver support for the crossfire customer is a huge turnoff. It's not so fun looking at your high numbers in FRAPS when the video flickers at different inopportune moments during BFBC2 or Medal of Honor.

Single AMD card setups work great. Every one of mine gives me no issue. It's when you start adding cards that you begin to see the lack of AMD's driver support for crossfire.

This is a non-issue with SLI. There's a reason Nvidia dominates the market despite all the good AMD press in the article.

 

[lemlo]

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.

 

[lemlo]

Because that's what fanatics do?

the term is enthusiasts!

 

[lemlo]

Here's a fun fact, does anyone know that in certain scenarios lesser nvidia cards perform better than higher rated nvidia cards on some games with low settings? I haven't tested it with amd cards yet but I'll bet it's the case there as well.

Done ranting, I just saw the avg scaling performance grid and it sent my Horse$%(@ alarm active. I don't normally post on forums anyhow. I typically just like to read and let everyone else argue it out.

PS one last edit: Way to go on the scaling AMD you've really come along way! Keep going!

 

[jfem]

Why 6950 when the direct competitor of gtx 570 in both price and performance is the 6970?