Mutli-card temperature differentials

[Ferzerp]

What kind of temperature differentials are others getting with dual video cards? I expect the top one to be hotter, but I don't know by how much. I know my lower card is a better card (ironic because it's the one with lower stock clocks), but I don't know if that's contributing to the differential.

Right now I see a difference of 10-12C between the top and bottom card when both are loaded. If that's normal, I won't touch it, if it isn't, I may swap the two cards and see if that equalizes the temps a little as I know the lower card runs slightly cooler in general when in single card configs.

 

[bigsnyder]

Depends on your setup, but a 10-12C difference does not sound unusual. Tell us more about your setup to give a better idea.

 

[Ferzerp]

Depends on your setup, but a 10-12C difference does not sound unusual. Tell us more about your setup to give a better idea.

It's just 2 reference GTX 680's sandwiched on top of one another in a Corsair 600t case (so no cable mess, it's all under the motherboard except the 4 pcie power cables), cpu is cooled with an H50, dumping heat inside the case. The 200mm case fans blow in the front, and out the top. PSU airflow is isolated from the rest of the system and exhausts out the bottom of the case. There's a ceton carblecard tuner right under the lower card. Nothing is above the upper card.

edit: top card is minorly OC'ed +25, +200. bottom card is much more aggressively OC'ed +130, +300

 

[notty22]

Yes that's actually good cooling performance.
That x58 board does not have a empty slot between the 2 dual slot cards. So what you are getting is fine. Even With a slot in between, sometimes the top card still runs about 10c hotter.

 

[hyrule4927]

It's just 2 reference GTX 680's sandwiched on top of one another in a Corsair 600t case (so no cable mess, it's all under the motherboard except the 4 pcie power cables), cpu is cooled with an H50, dumping heat inside the case. The 200mm case fans blow in the front, and out the top. PSU airflow is isolated from the rest of the system and exhausts out the bottom of the case. There's a ceton carblecard tuner right under the lower card. Nothing is above the upper card.

edit: top card is minorly OC'ed +25, +200. bottom card is much more aggressively OC'ed +130, +300

I hope I misinterpreted this . . . the fans on your H50 are blowing INTO the case?

 

[Ferzerp]

Of course they're blowing in. It's the corsair suggested configuration anyway. (See: http://www.corsair.com/media/cms/manual/H50_Supplemental_Mounting_Info_-_Forum_PDF.pdf and http://www.corsair.com/media/cms/manual/h50_qsg.pdf)

You have two options with that cooler. Use warm in case air over the radiator, or use cooler ambient air and dump the heat in to the case. For optimal processor cooling, yes, that's the way you set it up. Especially with the 200mm exhaust fan right above it anyway. There's going to be little opportunity for mixing with the air in the lower end of the case anyway because the flow is in the lower front, out the top.

 

[aaksheytalwar]

Afaik aren't we supposed to pull out the air?

 

[chimaxi83]

I had an H70 pulling into the case too, and performance was fine for that cooler. Do you have any fans near the bottom/front of case pulling air in too? That might help feed the cards slightly cooler air. I'm not sure what your case offers.

 

[Ferzerp]

Yeah, the 200mm fan in the front bottom is an intake fan. (see pic above)

 

[Ferzerp]

Swapping the cards around reduced the peak temp of the hottest card by about 2C and reduced the difference to 8-10C instead of 10-12C. The new bottom card runs a couple of degrees higher in the bottom slot than the other did (as I expected), but the new top card deals with the reduced air flow better and despite having a hotter card below it than the other did when it was on top, is slightly cooler.

It also consistently clocks slightly higher, so that's the one I would want to use in non-sli supporting games anyway. When I initially set it up, I just assumed the "superclocked" card would be a better card, but my vanilla card is superior in every way. But that's just the luck of the draw.

 

[psolord]

What kind of temperature differentials are others getting with dual video cards? I expect the top one to be hotter, but I don't know by how much. I know my lower card is a better card (ironic because it's the one with lower stock clocks), but I don't know if that's contributing to the differential. I always suspected this would be the case, that's why I always by radial cards for dual gpu, but I'm glad real life testing proved my suspicions correct.

Right now I see a difference of 10-12C between the top and bottom card when both are loaded. If that's normal, I won't touch it, if it isn't, I may swap the two cards and see if that equalizes the temps a little as I know the lower card runs slightly cooler in general when in single card configs.

I maintain a Youtube channel which is oriented on benchmarking PC stuff and through my testing it has occured that radial blower HSFs generally keep about the same temperature on dual card configurations.

To give a couple of examples, my 5850 crossfire system, was showing about same temps for the video cards (3C difference) while running Battlefield 3.
http://www.youtube.com/watch?feature=player_detailpage&v=ysZs5Tb4uEo#t=884s

Same goes for my 570 SLI system, with even less difference.
http://www.youtube.com/watch?feature=player_detailpage&v=cWIIwQxlWb8#t=834s

Contrary to these results, I found that when using axial type HSFs, which exhaust hot air inside the case, the temperature difference between cards is huge.

Testing a friend's GTX 460 SLI system, I saw a whopping 21C delta between the two cards.
http://www.youtube.com/watch?feature=player_detailpage&v=6-zife_qFUs#t=889s

All systems have similar cooling properties, with a side panel fan blowing additional cooling air on both cards.

So what surprises me here, is that your system also has radial HSF cards but still exhibits a considerable temperature delta.

I strongly believe this is due to your very limited breathing space between each card. Both my 5850 and 570 systems have a wee bit more space in between. Asus P5Q Deluxe for the first system and MSI Big Bang Trinergy on the second, just so you can compare the distance of the cards with your mobo.

 

[Ferzerp]

I strongly believe this is due to your very limited breathing space between each card.

Oh, I am sure that that's the reason. There is just no other way to do it though. I guess I could get a really long bridge, but then I'd be using x16/x8 instead of x16/x16

 

[Ferzerp]

I got the difference under extreme load (furmark) to be only 5C now. I installed two 90mm fans in the side of the case. (http://www.corsair.com/pc-cases/gra...e-series-600t-mesh-mid-tower-gaming-case.html). It adds a tiny bit of noise, but it's white noise that I don't notice. It shaved 3C off the max temp.

 

[Hauk]

Hi OP, yea temp variance between cards is affected by card spacing, case cooling, and GPU load (scaling) in particular applications. I have that case and have a space between cards which helps. I ignored Corsair's recommendation and am exausting with my H80 fans. No issues, it only helped. I run SSDs only, and put them in the upper drive cage and removed the bottom cages. I found in winter I was okay with clear side window installed. As room temps rose though, my card fans were spinning up so I installed the mesh panel and two 120mm fans at 1,200 in the bottom slots blowing in. That helped tremendously, warm air escapes through the open mesh above them.

In this case and others I've had, I've seen that with optimal case cooling, variance between cards was usually less than 5c. 680 SLI.. good stuff. That may be my next config!

 

[imaheadcase]

I'm literally face palming myself reading this thread. I can't believe people worry about a few Celsius in temp.

Its about as insane as people lapping CPU because one core was a few C difference.

 

[Ferzerp]

I'm literally face palming myself reading this thread. I can't believe people worry about a few Celsius in temp.

Its about as insane as people lapping CPU because one core was a few C difference.

You must not be familiar with what happens to a GPU when it gets warm.

Here is a hint. $@#%%$#%

Oh, sorry, you couldn't hear me over the jet that is a GPU fan when too warm.


edit: The point is, when they're sandwiched that closely together, the bottom card heats the top, and the top can't get much air. It's all too easy to end up with the bottom card in the high 70's and the top card is almost 90C which makes the fans blow full blast and makes it uncomfortably loud. You can either lower the temps for both cards, or try to get some of the heat away and make the differential less. So far, with minor changes, I've made the temp differential go from 10-12C down to 5-6C making my completely, utterly, never going to reach this in games temp max drop from 90 down to 84. This is a pretty big change and a pretty big noise savings. In games, I've gone from having noise in GPU intensive games, to having almost no extra noise.

 

[Hauk]

I'm literally face palming myself reading this thread. I can't believe people worry about a few Celsius in temp.

Its about as insane as people lapping CPU because one core was a few C difference.

There are many aspects an "enthusiast" considers important when designing and running a rig. Performance, cost, size, noise levels, power consumption, temps, are but just a few.

What is this site becomming when someone "facepalms" another, someone simply wishing to discuss whatever enthusiast variable they consider important at the time? This is half hobby you know. Many hobbies if put under a microscope make little sense..