NVIDIA GeForce 3-Way SLI and Radeon Tri-Fire Review @ [H]

Status
Not open for further replies.
FrgMstr

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
55,602
NVIDIA GeForce 3-Way SLI and Radeon Tri-Fire Review - We've seen what a Radeon HD 6990 can do when paired with a Radeon HD 6970 for "Tri-Fire" performance. Now it is time to find out what three NVIDIA GeForce GTX 580 video cards in 3-Way SLI game like in comparison. We will look at A2A performance comparisons and discuss which setup offers the best gameplay experience.
 
:eek: So well condensed article :)

All if it is a-a, for the people who insist on only seeing those....

Excellent review, Mr. Bennet & Brent!!
 
This definitely was not expected. I am an AMD/ATI guy, always have been, and I expected Nvidia to crush in a 3-way Sli vs. 3-way Crossfire performance comparison. Great review even if some of the Team Green Members will come in here ranting about something.
 
Wow that is shocking actually. Do you think the memory on the 580s is holding them back? Or is it simply a matter of poor scaling? I'm a fan of both camps and know that traditionally Nvidia has much better scaling with SLI than xfire, so this is very interesting to see the role reversal.
 
Last edited:
:eek: Damn... I really did not expect it to turn out like that at all. AMD's back in a big way this generation.

Imagine the bang for the buck in a 3x unlocked 6950 configuration.
 
Dayum!! Nvidia has work to do. Being that none of the scenarios were vram limited I'm actually shocked that AMD still managed to lay the smack down on them. I mean even if you used the 3GB versions that would have made 0 difference. Damn, AMD is even $500 cheaper in this review. This is tempting me to go 2011 platform and tri fire.
 
NVIDIA really needs to start shipping video cards with adequate memory.
 
NV should have disabled SLI for GTX 580 1.5GB then, since they are obsolete in high rez / multi-monitor setups.
 
Revdarian said:
From Vega's thread, yes, the memory holds them back a lot.

So you would need 3GB versions in order to make them shine, which i haven't checked, but am pretty sure to be more expensive even.

http://hardforum.com/showthread.php?t=1573598
Click to expand...

they are a 100-150 dollars more depending on what GTX 580 1.5GB card you compare it to.

as per amazon the palit GTX 580 3GB is 628 dollars x 3 = 1884 dollars not counting possible taxes or shipping.

so is it really worth the 884 extra dollars? doubt it.
 
Originally posted by Brent in This Article

We made sure we were not running into VRAM capacity limitations at these settings, so performance is being correctly compared and we are not hitting any VRAM walls at these tested settings.
Click to expand...

Vram is not going to help the fps at all as it isn't a factor at the settings used here.
 
GeForce GTX 580 3-Way SLI peaked at a whopping 1016 system Watts during our gaming! That is a lot of power folks, and is approaching on the limits of this PSU at 1200 Watts
Click to expand...

Do you really think that is the case?

Power supplies are rated based on output not input (as measured from the wall).

If the power supply was putting out 1200w, assuming it's maintaining at least 80% efficiency, that would be almost 1500w from the wall. Wouldn't that be more along the lines of the limits of the PSU?

With my Antec TruePower Quattro 1000w I regularly pull over 1000w during games like BC2 and the max I've seen it at is 1173 when running MSI Kombustor (same as furmark). But even at 1173w from the wall, assuming at least 80% efficiency, that was only ~938watts output.
 
bobzdar said:
RTFA, they made sure to stay under VRAM limited res/AA. It won't help them at all.
Click to expand...

EXACTLY! The vram would just sit there and do nothing.

If the game at those settings is using 8x AA exactly 1.422MB of Vram (Hypothetical) it's not like having more than 1.5GB of vram is going to help anything. That is just what the game is using. If Hardocp went further and saturated the memory of the card to the point where fps tanked e.g. 16X aa in the same game using 1.910MB of Vram usage then the extra vram from the 3GB version would help.

This was a thorough spanking due to

Better 6000 series multicard scaling
Lower power consumption from the 6000 series over Fermi 2.0
$500 or so less money spent.
Lowest performance comparison possible on Cayman Trifire

3 6970 would have thoroughly spanked the 3 GTX 580s at the same settings and so would 6970/6990 at the AUSUM mode settings.

Hell I'm betting 3 6950s stock would match GTX 580 in this comparison and save you $750 or more. If you unlock the 3 6950s as some forum members have done then you will once again spank 3x GTX 580.
 
Last edited:
Excellent new's for all involved with the red team, "go red team!" I guess the green team can continue to boast whatever is left, and what is that?
 
HOLY COW! NVidia got trashed!!!

After my 5870 fiasco, I swore I would never get another AMD card. Guess I need to reconsider.
 
Alpert said:
Excellent new's for all involved with the red team, "go red team!" I guess the green team can continue to boast whatever is left, and what is that?
Click to expand...

I think only Single Screen single GPU card performance crown. That is really where most people game. Also perhaps Single Screen Multi GPU could be a good scenario for them.
 
Wow... I'm totally floored! Did not expect that at ALL. AMD really has pulled out a new level of performance with multi-card setups. They really reversed their issues with 58xx scaling with the 69xx!

Lord_Exodia said:
Vram is not going to help the fps at all as it isn't a factor at the settings used here.
Click to expand...

While that is true for the most part, I do think there are a couple of those scenarios where the 3gb VRAM would make a bit of a difference- there were a couple places where they mentioned a possible CPU bottleneck that would be removed possibly with higher settings if there was more VRAM.
"We think we are actually running into possible CPU bottlenecks with three of these video cards here. Unfortunately, we cannot push the game to higher AA settings like 8X MSAA or 16X AA because the NVIDIA GTX 580 video cards have hit their VRAM wall at anything higher than 4X AA."

Still- the fact remains that the AMD trifire is by far the best value.
 
Flyinfinni said:
While that is true for the most part, I do think there are a couple of those scenarios where the 3gb VRAM would make a bit of a difference- there were a couple places where they mentioned a possible CPU bottleneck that would be removed possibly with higher settings if there was more VRAM.
"We think we are actually running into possible CPU bottlenecks with three of these video cards here. Unfortunately, we cannot push the game to higher AA settings like 8X MSAA or 16X AA because the NVIDIA GTX 580 video cards have hit their VRAM wall at anything higher than 4X AA."

Still- the fact remains that the AMD trifire is by far the best value.
Click to expand...

Hmm, yes possibly and yet they would have to raise the settings to do that which would lower the overall fps on both setups but I guess as long as it's playable. So anyone comfortable gaming at lower fps yet higher settings, than those shown in this review in those scenarios can consider a 3gb GTX 580.
 
I will say this, and I know it's not a case for every machine I see on the net, but I visit a lot of forums and see pictures of a lot of new builds and to be honest, it's always an ATI card that I am seeing in these new builds most of the time. The 6970 is very very popular.
 
Nice article.

BUT....let me offer this piece.

I recently bought two HD 6950s and flashed them to HD 6970, shaders and overclocked to HD 6970 speeds.

I had GTX 580s in SLi.

In my comparison of one game, Dead Space 2 (basically that's all I've had time for)....in surround and EyeFinity at 5760 x 1200 ......on the exact same system.....same settings...latest drivers and profiles.......2 cards, SLi vs CFX:

nvidia: 80-90 FPS
amd: 50-60 FPS

2 cents.:D
 
magoo said:
Nice article.

BUT....let me offer this piece.

I recently bought two HD 6950s and flashed them to HD 6970, shaders and overclocked to HD 6970 speeds.

I had GTX 580s in SLi.

In my comparison of one game, Dead Space 2 (basically that's all I've had time for)....in surround and EyeFinity at 5760 x 1200 ......on the exact same system.....same settings...latest drivers and profiles.......2 cards, SLi vs CFX:

nvidia: 80-90 FPS
amd: 50-60 FPS

2 cents.:D
Click to expand...

your doing it wrong ;)
 
GotNoRice said:
Do you really think that is the case?

Power supplies are rated based on output not input (as measured from the wall).
Click to expand...

We like to see PSU usage at a about 75% max. So it depends on how you look at it. Still you are arguing over semantics.
 
@ Magoo

Interesting, I have a 6990 and againt my other box with 590's in SLI, all my games run smoother, 5 - 10 frames faster. The 6990 is a one card solution. The other box with the 590's has 2 cards, forced me to buy a bigger power supply and the heat generated out of that box is pretty warm. On top of, the performance is less that the 6990. The 590 GTX's also cost me over $1,000 dollars as where the 6990 was under $700.

Really wish I would had just bought the 6990 with 4gigs of ram that was cheaper, faster, more memory, used less power and was a one card solution.

I will try and be smarter next time.
 
:eek: WOW! :eek:

That is a huge, and unexpected, difference in raw performance.

Just... wow...

Just floored at the difference... $500... 200watts...


...oh, and Great review! ;)
 
@ Jeremyshaw, I was being sarcastic / condescending toward Magoo. Hell no I don't have any 580's. Those are for college kids with credit cards that think a few extra frames are worth the extra $150 - $200 dollars per card.

I'm very happy with my MSI 6970 2gig 920 / 1430

Yup, thanks for the review guys!
 
Last edited:
Finally the competition is swinging back the other way. Perhaps now Nvidia can go back to the drawing board and rethink some of their designs.
 
Can I just ask one thing:

Why did you use a 3.6Ghz i7-920?

You could have used a 990x, or even a 2600k. Something above 4.2Ghz..

There's a discussion at OCN about this and I took the time to ask.
 
SixFootDuo said:
@ Jeremyshaw, I was being sarcastic / condescending toward Magoo. Hell no I don't have any 580's. Those are for college kids with credit cards that think a few extra frames are worth the extra $150 - $200 dollars per card.

I'm very happy with my MSI 6970 2gig 920 / 1430

Yup, thanks for the review guys!
Click to expand...

ah, I already have/had the HD6950, HD6990, GTX570, GTX580, and lots of older cards, lol.
 
Yeap sorry for not saying it entirely, i meant for the 3 games that seemed CPU limited.

But hell no, they are so not worth the pricetag!

Forgot to thank for the article, always great to see proper comparisons (and by this i mean actual gameplay ones with eyefinity, not the 3 vs 3 disregarding the huge difference in pricetag)
 
This is a fantastic review and really shows the turnaround for AMD with the 6000 series (i'd consider the 5000 series a stepping stone to the performance we're seeing now).

I just recently bought my first AMD card and probably won't be looking back. The power consumption numbers alone are absolutely staggering, take into account the massive price difference and now the performance as well? I'm turning red. I'm turning red all over.

Nvidia has been paying far too much attention to their tegra's, because they've been getting their asses handed to them recently. Even with single monitor setups and single card solutions you're seeing AMD providing a broader and cheaper options than nvidia. But the market for single card, single monitor solutions at ~$4-500 isn't exactly a large niche in the market.
 
Status
Not open for further replies.
Back
Top