Cooling crossfired 6990's

[tylr0drdn]

im building a new rig and im wanting to crossfire two radeon 6990's on a asus p9x79 deluxe mobo. while doin research i found that the 6990's run hot when crossfired so i was going to use the accelero twin turbo 6990 (x2) to solve the problem but i dont know if they will fit...is anyone running anything similar that can help or know whether or not it will fit? thanks in advance.

 

[simon12]

The board has a 2 card spacing so it should fit fine. I would suggest that you look at crossfire 7970s as quad crossfire scaling is not that great so it may perform as well or better. Also it won't be that long until AMD release a dual 7970 anyway which should smash the 6990 performance.

 

[alrobichaud]

I am using stock cooling with no issues at all. With a 2 slot space for airflow to the blower you won't have any issues either. The 6990 runs hot when you switch the BIOS to position 2 which increases the voltage and runs it at stock 6970 speed. In this position, the card draws 450W at full load which is why it runs hot. I found that with a 6990 and a 6970 and a two slot spacing between them did not run any hotter than the 85 degrees that any benchmark review says the 6990 runs under full load by itself. I was leary about adding another 6970 in between the two cards with stock cooling but after watching a couple of you tube videos about quadfire 6970 i went ahead and tried it. My cards do not get any hotter than a single card would on it's own. I do have 2 sidemount 120mm fans blowing directly on the cards and I have some rubber spacers inbetween the cards to allow for more airflow. I also have a front mount exhaust directly in front of the exhaust of the 6990. I say you will not have any issues with stock cooling. If amd designed the card to run at 85 degrees do you really care if it runs any cooler? I found the scaling in quadfire to be quite well in bf3. I play with 5760x1080 ultra no AA. After some testing, I came up with these results: 6990 = 40FPS, 6990 and 6970 = 57 fps, 6990 6970 6970 = 77fps. sythetic benchmarks like 3dmark11 show no difference between trifire 6990+6970 and quad 6990+6970+6970. I am going to try heaven 2.5 later and can post my results if you like.

 

[alrobichaud]

I apologize for not answering your question in my previous post. I was about to leave work when I thought I would leave my 2cents. I have no experience using aftermarket cooling on these cards. The bottom line is the only time the 6990 will run hotter than what AMD designed it to is when you overclcock it. Overclcocking a single 6990 shows some improvement in benchmarks and in game fps but as soon as you crossfire it there is no gain at all to overclock. Running at stock speeds and doing some heavy gaming, you will find your temps occasionally rise above 85 but the blower turns up to compensate and then immediatley goes back to running quiter. I barely notice the sound from all 3 of my cards over the sound of my case fans. Reviewers like to exaggerate when it comes to things like noise. Finally, scaling crossfire 6990's does not scale to well at 1920x1080. From my own experience as well as benchmark reviews, the gain can be as little as 50-60 %. On the other hand, if you look at how well they scale running high res like 5760x1080 you will find they scale very well. finding benchmarks for crossfire 6990 at 5760x1080 res are nearly impossible on the inet but i ran the heaven 2.5 benchmark at 5760x1080 and came up with these results:

6990 = score 1018 FPS 40.4
6990+6970 = score 1581 FPS 62.8
6990+6970+6970 = score 2083 FPS 82.7

 

[tylr0drdn]

I apologize for not answering your question in my previous post. I was about to leave work when I thought I would leave my 2cents. I have no experience using aftermarket cooling on these cards. The bottom line is the only time the 6990 will run hotter than what AMD designed it to is when you overclcock it. Overclcocking a single 6990 shows some improvement in benchmarks and in game fps but as soon as you crossfire it there is no gain at all to overclock. Running at stock speeds and doing some heavy gaming, you will find your temps occasionally rise above 85 but the blower turns up to compensate and then immediatley goes back to running quiter. I barely notice the sound from all 3 of my cards over the sound of my case fans. Reviewers like to exaggerate when it comes to things like noise. Finally, scaling crossfire 6990's does not scale to well at 1920x1080. From my own experience as well as benchmark reviews, the gain can be as little as 50-60 %. On the other hand, if you look at how well they scale running high res like 5760x1080 you will find they scale very well. finding benchmarks for crossfire 6990 at 5760x1080 res are nearly impossible on the inet but i ran the heaven 2.5 benchmark at 5760x1080 and came up with these results:

6990 = score 1018 FPS 40.4
6990+6970 = score 1581 FPS 62.8
6990+6970+6970 = score 2083 FPS 82.7

so i would be alot better off (if i was worried about framerate) going for the quad set up (6990,6970,6970)? what if i just went to 7970 quadfired would that help the situation even more? and furthermore would a seasonic x1250 power all of it along with four case fans and a coolermaster v10?

 

[tylr0drdn]

no worries about the answer un-answered i go off all the time and forget what i was even taslking about lol

 

[alrobichaud]

I am the king of tangents. It can be a real problem sometimes. Go with crossfire 6990's. The only reason I went with 6990 and two 6970's was the price. The only difference between on 6990 and two 6970's is the latter has a stock gpu speed of 880MHz vs 830MHz and a memory speed of 1375MHz vs 1250MHz. The architecture of the cards are the same. The 6990 was scaled back because it would draw 450W if it ran at 6970 speeds. You can easily clock the 6990 at 6970 speeds but you need to be able to provide 450W to a single pci-e slot. The slot is 75W and each 8 pin power connector is 150w. My motherboard has two molex connections which add additional power to the pci-e slots so running 450W is no problem. Two cards are easier than one for sure. The 7970 is supposed to be roughly 30% faster than the 6970 so quad 7970 will definatley destroy quad 6970 but will cost roughly $2200 for four cards. Crossfire 6990's running at stock speeds will crush just about anything you throw at them and what I have found is running any faster than stock 6990 speeds added no performance at all to this setup whether I tried a benchmark or a game. The only thing I can't say is whether 4 x 6970 is any better than 2 x 6990. You are really splitting hairs at this point. I can tell you that you will need lot's of room in your case for four of these cards. In my cooler master stacker 830, When I put a card in slot 4 I had to cut away a small aluminium bracket so the card would fit and it is a large case.

Again, the easiest setup is crossfire 6990's.

 

[alrobichaud]

And, yes a 1250 W supply will do just fine. AMD says you need 1000W for crossfire 6990 and recommends 1200W for stability which I think means if you are going to overclock. I have a crappy 1200W supply powering, 7x 120mmm case fans, a water pump, a 130W processor which I have overclocked so who knows what it draws and 875W worth of video cards and I have no issues.

 

[wolfram23]

What resolution are you running, OP?

Quadfire scaling really doesn't scale nicely, even at high res you're paying a lot for a few frames. Triple cards tends to be a great point because it almost entirely reduces micro stutter to zero and you still get decent scaling.

A pair of 7970s would be super fast, 3 of them would absolutely crush 2 6990s or 4 6970s.

It really depends on what you want to run, though, and at what resolution.

 

[alrobichaud]

What resolution are you running, OP?

Quadfire scaling really doesn't scale nicely, even at high res you're paying a lot for a few frames. Triple cards tends to be a great point because it almost entirely reduces micro stutter to zero and you still get decent scaling.

A pair of 7970s would be super fast, 3 of them would absolutely crush 2 6990s or 4 6970s.

It really depends on what you want to run, though, and at what resolution.

Not that it really matters but all of the benchmark reviews that I have read including the one on Tom's spec the 7970 to offer 30% more performance than the 6970. So that would put 3 of them on par with quad 6970's, not absolutley crushing them. Also, maybe earlier generations 'don't scale well even at high res' but my own experience with bf3 and using the heaven 2.5 benchmark have these cards scaling very well at high res as per my results that I posted earlier in this thread.


Heaven 2.5


6990 = FPS 40.4
6990+6970 = FPS 62.8
6990+6970+6970 = FPS 82.7


I did some lengthy benchmarking with bf3 using fraps and came up with pretty much the same results in terms of scaling at 5760x1080. I do agree with 'It really depends on what you want to run, though, and at what resolution.' My scaling was not as great running a single 1920x1080 monitor but this setup really isn't good for that. I am guessing if you are even considering crossfire 6990's it isn't going to be to play wow on a single 1920x1080 monitor.

 

[wolfram23]

What settings were you using? I'd be interested if you reran the test to match the results here: http://www.vortez.net/news_story/tri_fire_7970s_pack_a_hell_of_a_punch.html

They also show extremely high performance with 2 of them http://www.techpowerup.com/reviews/ASUS/HD_7970_CrossFire/1.html

 

[diec4st]

go for 7970's...

 

[tylr0drdn]

the resolution i will be running is 3840x1080..i dont quite have enough room for three monitors. and mainly games and movies..ok games. and further thinking about price.. quad 7970's is out i didnt think when i posted four..three possibly, but i would feel guilty each time i looked at the screen knowing i paid more for graphics than i did my car :lol: and overclocking is definatley a possibilty which is why im allowing for so much cooling if that helps the answer at all.

 

[esrever]

2 7970s oc'ed would be more powerful than 2 6990s considering the crossfire diminished returns passed 2 cards really stack up.

Also would be a lot cheaper and run less hot with less power. I don't see going with 2 6990 being a good choice.

 

[alrobichaud]

What settings were you using? I'd be interested if you reran the test to match the results here: http://www.vortez.net/news_story/tri_fire_7970s_pack_a_hell_of_a_punch.html

http://www.vortez.net/index.php?ct=news&action=file&id=1940

They also show extremely high performance with 2 of them http://www.techpowerup.com/reviews/ASUS/HD_7970_CrossFire/1.html

I had a look and ran some more benchmarks using the same graphic settings but I am limited to widescreen resolutions so I used 1920x1080 to compare to the 1600x1050 results with a 6990, 6990+6970 and 6990 +6970+6970. As soon as I figure out how to upload a pic I will upload the results. Any tips on how to do this. I tried photo bucket but the results are .html and it will not let my upload that. BTW, crossfire 7970 really do not come close to crossfire 6990. Also, if you are overclocking then yes get aftermarket cooling but based on my results as well as all the reviews that I have read it will only add heat and noise with really no gain in performance.

 

[alrobichaud]

What settings were you using? I'd be interested if you reran the test to match the results here: http://www.vortez.net/news_story/tri_fire_7970s_pack_a_hell_of_a_punch.html

http://www.vortez.net/index.php?ct=news&action=file&id=1940

They also show extremely high performance with 2 of them http://www.techpowerup.com/reviews/ASUS/HD_7970_CrossFire/1.html

If all goes well with the link to photobucket, you should see 5 pics.


http://i1175.photobucket.com/albums/r637/alrobichaud/trifire1920x1080.jpg

http://i1175.photobucket.com/albums/r637/alrobichaud/quadfire1920x1080.jpg

http://i1175.photobucket.com/albums/r637/alrobichaud/crossfire5760x1080.jpg

http://i1175.photobucket.com/albums/r637/alrobichaud/trifire5760x1080.jpg

http://i1175.photobucket.com/albums/r637/alrobichaud/Quadfire5760x1080.jpg

 

[alrobichaud]

The first two links show my results with the same graphic quality but at a higher resolution. The following 3 show how well the scaling is at 5760x1080. Going from crossfire to trifire to quadfire gave 40fps, 62 fps and 82 fps respectively. This is why I keep arguing when someone says that these cards do not scale well especially at high res. I agree going from 116FPS to 140FPS at 1920x1080 is not good but who is running quad 6970 at that resolution. Since I am unable to use any resolution other than widescreen formats, the best I could do is 1920x1080 to compare to 1680x1050. It is really an unfair comparison but it does show that crossfire 7970 will not outperform crossfire 6990 even at low resolutions. As you can see by the less than stellar scaling at 1920x1080, doing any comparison at such low resolutions is not a good comparison at all. So unless we can find some benchmarks at 5760x1080 with the 7970's it is a matter of opinion at this point.

 

[esrever]

benchmarks like uningen don't show real world performance and you'd be hard pressed to get any real gain in games with quad fire.

thats pretty close considering its 4 gpu vs 2 gpu, $1400 vs $1100

 

[jk47]

You seem to like spending money.

I'd recommend ASUS ROG II - in quad xfire.

 

[alrobichaud]

benchmarks like uningen don't show real world performance and you'd be hard pressed to get any real gain in games with quad fire.

http://tpucdn.com/reviews/ASUS/HD_7970_CrossFire/images/heaven_1920_1200.gif

http://i1175.photobucket.com/albums/r637/alrobichaud/quadfire1920x1080.jpg

thats pretty close considering its 4 gpu vs 2 gpu, $1400 vs $1100

http://www.guru3d.com/imageview.php?image=35479

I agree with you, it is pretty close. We can all agree the 7970 is much better than the 6970 but It would be nice to see some benchmarks, either synthetic or in game, that use high resolutions such as eyefinitey. I admit that the scaling through 4gpu's is not great at 1920x1080 and if that is all it they would get used for then it's a bit of a shame. I won't get into the details but bf3 is a good example of quad gpu scaling at 5760x1080. I have spent many hours running samples of the same small part of a level for the same timeframe using fraps and came up with results that mimic what heaven 2.5 showed at eyefinitey resolutions. I have yet to find a benchmark online that has done similar testing. I have also yet to see anything posted by any members here that show similar testing. In regards to 3dmark11, my results are roughly the same as the graph you provided but the reason 3dmark11 sucks is as follows:

6990 = p9135
6990+6970 = p12888
6990+6970+6970 = p12987

This a good example of a benchmark that does not like quadfire. If you have a look at the results for trifire vs quadfire 6970's you will find that the highest numbers are the same for each instance on futuremarks website. Regardless of benchmarks quadfire 6970's at 5760x1080 resolution in ultra playing bf3 gives me roughly 80FPS without AA and 60FPS with 4xAA. That is what is important. If anyone comes across any eyefinity benchmarks using bf3 and 7970's in quad or tri, I would like to see the results if you could msg me pls.

 

[esrever]

http://www.hardwareheaven.com/reviews/1362/pg4/amd-radeon-hd-7970-crossfire-performance-review-battlefield-3.html

 

[esrever]

http://www.hardwareheaven.com/reviews/1376/pg1/sapphire-radeon-7970-quad-crossfire-first-look-performance-quad-graphics-cards.html

 

[tylr0drdn]

so if im understanding the data correctly two 7970s would be almost the same as running two 6990s on multiple displays but the two 7970s would perform better on a single display if i was to hit up a lan party?

 

[tylr0drdn]

granted its not all for me.. i appreciate the time you guys are putting into this very interesting lessons to be learned. without me throwing away money on experimenting lol.

 

[alrobichaud]

so if im understanding the data correctly two 7970s would be almost the same as running two 6990s on multiple displays but the two 7970s would perform better on a single display if i was to hit up a lan party?

I think the jury is still out on this one. It is easy to believe a chart from a benchmark and then you find something like this and say whaaa?

Found that in one of the forum discussions on hardwareheaven talking about poor 6990 performance in some of the bechmark testing done.