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

Status
Not open for further replies.
My three unlocked 6950s cost me under 750; they run all day long at 940 / 1400 (aka higher than stock 6970)..... I didn't expect to see the results as they turned out, but man oh man; spending seven hundred and change never felt so damn good.
 
Interesting comments. I like the point about CPU limited testing scenario. I don't THINK it is valid, but it is something worth testing. I will build Brent a new nicely overclocked (4.5GHz) Sandy Bridge box this evening, get it stress tested, and send it out to him for Monday delivery. Then next week we will run the tests again. I think if nothing else, it will satisfy some so the questions. We have been wrong before, and if we are, we surely want that corrected. I have a nice P8P67 WS Revolution board here that will give us some nice wide PCIe pipes as well that needs to be put to good use. Hopefully after that, we will have some solid answers that the conspiracy theorists cannot shoot holes in.

And to address some of the other comments...this is REAL WORLD GAMEPLAY done in a Apples 2 Apples testing scenario. I think some people do not understand that either. I will post this at OCN to make it clear as well.

I posted this over at OCN as well.
 
BE SURE TO BUY THREE 6970'S TOO SO WE CAN TEST 3 CARDS VS 3 CARDS. us OCN users are all about fair play.
 
Why are you directing your post @ me ppelo Could you reread my post maybe and the one/two before? I said that is what I think the OP was trying to say, not that I agreed with it. Don't call me a fanboy when I only interpreted something to make it clearer to others as to what the op said.

And Kyle could you do 4.8ghz just to make sure everyone can sleep good at night? Because you know people will come crying that 4.5 is too low ;)
 
hey, fuck it. test 5.3 ghz. anything else is no enthusiast enough. Nvidia card get twice the frame rate if you OC the CPU donchya know.

shit, just try random settings and multiple runs until AMD loses something.
 
Cant wait to se 2600k in real action. Maybe I am wrong but generaly I think
that Nvidia high end GPUs need more CPU power. BF2 was running much better with my ex 470 sli setup and 2600k @ 4.6, even at x8 pcie, than ex i7 950 @ 4.0.
Good work!
 
Kyle, Pandoras box has now been opened. Boy, talk about it not being CHEAP or EASY to run a web site. Not that I would ever propose to tell you how or what to test but if you have 'em, 3 x 6970's, 6990, 590, 580's etc etc. You might wanna dust them off. I could see a mob mentality popping up asking you to bench every conceivable combination with those cards. Boy, the fun just doesn't stop does it? :)
 
SixFootDuo said:
Kyle, Pandoras box has now been opened. Boy, talk about it not being CHEAP or EASY to run a web site. Not that I would ever propose to tell you how or what to test but if you have 'em, 3 x 6970's, 6990, 590, 580's etc etc. You might wanna dust them off. I could see a mob mentality popping up asking you to bench every conceivable combination with those cards. Boy, the fun just doesn't stop does it? :)
Click to expand...

Honestly, we have some "gaps" in our schedule right now and it has been on my To Do list to get Brent a new system for 4 months or so. So the time is right and we will get it done. :) Just finally got the right mobo for it. The P8P67 WS Revolution......needed more PCIe lanes to do it "right." Just been waiting for the right board and it just got returned from Morry finishing up his portion of the review. :)
 
Seriously, why do you care what the children over at OCN think? That forum is full of toms hardware rejects anyways.

Have you read some of the threads on that forum?
 
BababooeyHTJ said:
Seriously, why do you care what the children over at OCN think? That forum is full of toms hardware rejects anyways.

Have you read some of the threads on that forum?
Click to expand...

Let's not go there.......

And let me say, we are all in this together. Forum loyalties are surely part of the industry, but different strokes for different folks.
 
If I could just send you that. i7 2600K at 5.4 1.49v, running 24/7 at 5.1 on a Maximus IV.

With 6990+6970+6970. I'm sure it could take 580 Quad-SLI head-on no problems. :)

5667664356_78f5e61d34_b.jpg
 
AMD has what a one year jump on Nvidia with eyefinity? On top of that crossfire scaling this generation is very good. I don't see why people are surprised that AMD tends to perform better in tri-fire with eyefinity. I'm sure that the extra vram doesn't hurt either. It wouldn't make sense to test $1500 in video cards on a single display at this point in time imo.
 
Last edited:
Wow, great article!

When I first started reading this, I figured that the ATI combination was going to get pummeled after reading that you guys chose to use the stock bios settings on the 6990. The results of this are shocking to say the least. I don't have an Eyfinity 3 setup, but one can get the same or better performance than a tri-580 setup for $500 less and 200 watts less? Wow!

Since Kyle will be sending out a highly overclocked SB rig for a retest, I eagerly await what the results might be.
 
There is no CPU limitation affecting this evaluation in any way. People keep bringing it up like it's something that needs to be examined, but it's just not an issue. 7MP is approaching two 2560 x 1600 monitors. Would an i7 @ 3.6GHz HT bottleneck a 2560 monitor? No, it wouldn't, and it definitely wouldn't bottleneck almost two of them. This setup's weakest link is its GPU, not its CPU. I stand by this post, and will reference it when the re-test proves my claim.
 
what a smackdown. i can't wait till i can get a cheap 6990 to pair with one of my 6950's unlocked. the other 6950 unlocked i have will go into the living room pc and provide ample oomph for my 1920x1080 hdtv ;)
 
Great review!

I feel some people have just had bad experiences with AMD/ATI and they just refuse to be open minded about another company other than Nvidia.

Even with all this evidence they'll still be like "It's okay I'll take the GTX 580" ... just like that video about the HTC vs iPhone ...

AMD FTW!
 
Nice review. I was thinking of swapping my 2 x HD6970's out with 2 x GTX580's, but that's not gonna happen now. Thank you for the review guys.
 
Torture testing now. 4.8GHz. Hopefully it will make a solid 12 hours or so. Using a Corsair H60 cooler.
 
Seems irresponsible to not test at 2560x1600. Neither setup is really fast enough for triple monitor as evidenced by the garbage average framerates throughout the test. Every time the framerate drops below 60 it's a huge distraction, and most of the averages here are below 60. The vast majority of even the hardcore multi-gpu segment is still using only a single monitor.
 
Since you are using a Sandy Bridge platform for multi-monitor testing then the availability / limitation of PCI-e lanes might be a minor factor. What are the odds that after the test somebody will complain about the x8/x4/x4 of the tri-sli versus the x8/x8 on the 6990+6970 trifire if AMD wins again?
 
Running 5x1 portrait under quad 6970 with sandy @ 5.25 GHz. What may be unbeknownst to OCN devotees is crossfire scaling blows through ceilings with high clocks, just as well as NVIDIA. I see 99% usage on 3x1 portriat, landscape, and 11 megapixel 5x1. And that's with a fourth command buffer to feed. My prediction: we will see proportional gains to both gpu configs, with the exact same conclusion ( + sli choking above 2 or 4 MSAA).
 
anonmoniker said:
Seems irresponsible to not test at 2560x1600. Neither setup is really fast enough for triple monitor as evidenced by the garbage average framerates throughout the test. Every time the framerate drops below 60 it's a huge distraction, and most of the averages here are below 60. The vast majority of even the hardcore multi-gpu segment is still using only a single monitor.
Click to expand...
As an owner of an SLI 580 system, I can assure you that TWO 580s is enough to drive most games at 5760x1200, let alone three. 2560x1600 is simply not an issue for top-end multi GPU setups, it would be a waste.
 
As some have pointed out here the 590 and 580 cant be SLI'd...... oh poop I guess I didnt know that and it would be nice if nVidia could deliver. Such nice cards and so limited.

Oh well another win for AMD.

**Yeah im fanboying, so what!
 
Rizen said:
As an owner of an SLI 580 system, I can assure you that TWO 580s is enough to drive most games at 5760x1200, let alone three. 2560x1600 is simply not an issue for top-end multi GPU setups, it would be a waste.
Click to expand...

Why would anyone want to SLI $1K in cards for that resolution anyways (2560x)?

Ans: because they can! But you are right in what you say :)
 
anonmoniker said:
Seems irresponsible to not test at 2560x1600. Neither setup is really fast enough for triple monitor as evidenced by the garbage average framerates throughout the test. Every time the framerate drops below 60 it's a huge distraction, and most of the averages here are below 60. The vast majority of even the hardcore multi-gpu segment is still using only a single monitor.
Click to expand...


the difference is they are running single player testing which means you CAN get away with lower frame rates. were as on the multiplayer side 60fps is important. but all of us hardcore gamers know we drop the quality settings for multiplayer to make sure we get a solid 60fps so it doesn't matter. even for me in single player i'll max out as much as i can no matter what the frame rate is in single player as long as its playable. but in multiplayer AA/AF, shadows(when possible) always get turned off when i'm playing online.

Teenyman45 said:
Since you are using a Sandy Bridge platform for multi-monitor testing then the availability / limitation of PCI-e lanes might be a minor factor. What are the odds that after the test somebody will complain about the x8/x4/x4 of the tri-sli versus the x8/x8 on the 6990+6970 trifire if AMD wins again?
Click to expand...

then direct them to the reviews done showing the lack of bottlenecking between 16/16 8/8 8/4.. its only at 4/4 where you start to see a small issue with multi-screen resolutions and even then its maybe 1-2 fps. but they cried for a sandy bridge setup so they better deal with it.
 
Ya' you don't need mega cards for 2560x. [H]ard showed GTX 460 1Gb SLi @ 810 on the core pushing Metro maxed at that res.

 
Last edited:
sirmonkey1985 said:
the difference is they are running single player testing which means you CAN get away with lower frame rates. were as on the multiplayer side 60fps is important. but all of us hardcore gamers know we drop the quality settings for multiplayer to make sure we get a solid 60fps so it doesn't matter. even for me in single player i'll max out as much as i can no matter what the frame rate is in single player as long as its playable. but in multiplayer AA/AF, shadows(when possible) always get turned off when i'm playing online.



then direct them to the reviews done showing the lack of bottlenecking between 16/16 8/8 8/4.. its only at 4/4 where you start to see a small issue with multi-screen resolutions and even then its maybe 1-2 fps. but they cried for a sandy bridge setup so they better deal with it.
Click to expand...

they ran X58 which has 16/16 - 16/16 ? 36 pci-e lanes, so you can run 580's in
16-8-8 ? config.


as for the fps statement, a single 6950 570 would suffice, competetive people never run max effects, AA however is wanted.
 
That tests were done in Eyefinity a Amd feature.

Lets go again round two ding ding ding

6990/6970 vs 2x Gtx 580s in 3d gaming
 
Spooony said:
That tests were done in Eyefinity a Amd feature.

Lets go again round two ding ding ding

6990/6970 vs 2x Gtx 580s in 3d gaming
Click to expand...


nice spam attempt.. it was done in eyefinity AND Nvidia 2D surround.. same crap.. and it would be round 3, get your spam right.
 
sirmonkey1985 said:
nice spam attempt.. it was done in eyefinity AND Nvidia 2D surround.. same crap.. and it would be round 3, get your spam right.
Click to expand...

spam attempt? Think you got me mix up with your refugee letters your busy with

Btw do you own 3 gpus? Have you ever experience 3x gpus in real life with with real world gaming? Or are you just amplifying everything your read what happened with 3 or 4 games. Some games love Amd cards some games love nvidia cards. Im saying this coz I'm experiencing it.

How's this for spam
battlefield68.jpg


When you own 3x of these fuckers
3-way-sli-install.jpg

one day and experiencing awful gameplay in everything or whatever you throw at it then come back to me talking about spam. I will leave your ego trip so you can carry on with whatever you were pretending to do
 
Last edited:
What are you trying to say spoony? Your setup gives you awful gameplay?
 
spoony is that really your rig? i see a picture copyright logo in the bottom right ;)
 
Thanks for this excellent review. I liked this jab against other, lesser review sites: :D
Brent Justice said:
In order to test this, we of course tested at no less than a triple-display setup at 5760x1200, anything less would be irresponsible on our part to show you meaningful performance.
Click to expand...
Of course there will be always people who try to find something to criticize in your articles.

6970 CF vs. 580 SLI? omg the SLI setup is so much more expensive
6990+6970 Tri-Fire vs. 580 SLI? omg 3 GPUs against 2 GPUs is unfair
6990+6970 2GB Tri-Fire vs. 580 1.5GB Tri-SLI? omg you are VRAM limited

But I think your reviews accurately reflect the choices that consumers have in this market. A 580 3GB SLI review would be nice, but those cards are hard to find and almost as expensive as a 6990. Add to it 30 W for the extra memory per card and you'll become PSU limited ;)
 
When a community gets as large as [H], it would be prudent for the management/admins to consider the audience when hearing complaints about reviews before deciding to dedicate time and resources to humoring those complaints.

Considering the scaling trends highlighted in every single 580 SLI/Tri-SLI review, the numbers in this review matched up very close. I think it came out closer to 65% than 70 or 80, but it's close. That's within the limits of what I expected going from two to three 580's.

So while you may have had a seemingly (as in, they were loud) large portion of the readership complaining that Nvidia didn't get a fair chance... you should have told them that's Nvidia's fault, not [H]'s. $500 is a lot of money. That's two 128GB SATA 3 SSD's in RAID-0 or a small NAS setup... or a whole number of other things. It's impossible to ignore.

I wonder what Nvidia's GTX 580 sales numbers look like...
 
Status
Not open for further replies.
Back
Top