• Happy holidays, folks! Thanks to each and every one of you for being part of the Tom's Hardware community!

Need help with SLI Temps.....

s1lent

Distinguished
Apr 11, 2009
160
0
18,680
Hi all,

Having some trouble with SLI TEMPS!!!

I just upraded to SLI with 2-Zotac Geforce 980 ti Amp! Extreme 6tb (# ZT-90505-10p....fastest out of the box 980ti available right now).


The problem I am having is with temps......when in idle, my top gpu is at 38 to 43 degrees, as for my bottom gpu it runs at 61 to 63 (which is about 20 degrees hotter). Now in in full load, my top card (53 to 63 degrees) the bottom gpu (80 to 84) they can range anywhere from 20 to 30 degrees hotter than the other. I do see in other posts that 10 degrees hotter is ok but 20 to 30?? Is this normal??

The cards are huge!!! They each take up about 3 spots on my Mobo, they are basically sitting right next to each other with no room to breath. Also, they sit about an inch away from the PSU.

Cooling....I have Cooler Master HAF 932 case with 3 huge case fans and a Krakenx61 cooling my cpu. I also run a portalable A/C in my office at around 68 to 71 degrees.

Also, i ran a few test...

#1-Unigine Valley Benchmark
FPS:120.4
Score: 5038
Min FPS:39.7
Max FPS: 155.9
Graphics: 1607mhz
Memory: 3610
temps Top card 53/63 bottom card 80/84

#2--Ran some BF4 and GTA V and they both ran fantastic!

#3--ASIC quality checker
GPU #1 68.9% and GPU #2 68.5%!! I was told anything under 70% I should RMA. Not sure if this is something to be worried about.


The 20 to 30 degree range between the two cards is my concern anyone have any suggestions?

Thank you in advance.
 
Solution
Both cards work fine in both slots individually. So you know it's not the cards

Put both cards in and make sure the SLI bridge is seated correctly. Be sure to activate SLI under 3D settings and that, assuming you only have one monitor connected, set "Single Display Performance mode" under Nvidia control panel's 'Manage 3D settings->Multi-display/mixed-GPU acceleration" and under "Power Management Mode" set it to "Prefer Maximum Performance".

Turn Anitialiasing transparency "on" under global settings in Nvidia control panel. Then set AA as high as you can in BF4 and use HWmonitor to determine the CPU and GPU utilization while you play multi-player BF4. Also, check out the temps. Does the CPU hit 100% utilization on any...
how do you have your PSU mounted in the case. Is the fan pointing up or down? Normally the lower card is cooler then the top since the top card is right next to the second card, the second card will block airflow to it.


Try removing the side panel and see if your temps improve, also make sure the fan for the PSU is pointed down to the bottom of the case.
 
The PSU fan is pointing down and pushing air out the bottom and back. I forgot to mention I have tried with and without the side panel not much of a noticeable change.
 


Try swapping cards, so put the top one in the lower slot and the bottom one it the top and see if the heat follows the card or if it's just air movement issues in the bottom of the case.
 
First thing... I have a non-contact infrared thermometer and when I read the temps of each card the "top card is hotter than bottom".

I have swapped the cards.

Zotac utility Firestorm shows @ idle

GPU#1 at 36c
Core clock at 135mhz
Mem clock 405mhz

GPU#2 at 63c
Core clock at 1013mhz
Mem clock at 3610mhz

Opened Valley benchmark. I haven't even started the benchmark and GPU #1 43 degrees and GPU #2 71 degrees!

My HAF 932 case fans..

1-200mm pulling air in the bottom front
1-200mm fan from the side panel. (side panel off right now)
1-120 or 140mm fan at the back
2-140mm with radiator at top pushing air up and out (Kraken x61)

Not understanding why it's so hot and why the core/memory clock are down sampling!!

Any suggestions? thanks for your help
 
I had two two-slot cards in SLI and three two-slot cards in crossfire at given points in time. The biggest difference I've seen between the hottest (top) and coldest (bottom) was 10C. 20C is a little much.

Even though the temps are within tolerances for that GPU (92C max), I would be a little concerned with the bottom one(?) being that much hotter?

Is it possible that you have different fan profiles assigned to each in the Firestorm software? I run Precision X and if you have multiple cards, you can give them each their own profile. Also, that core clock reported in your second post seems odd for GPU1, but it may idle low as the GPU2 should as well.

I could see either of the cards getting into the 70s or 80s under load, but neither should idle in the 60s unless you have a conservative fan profile for that one.

Can you install HWmonitor and get the temps from that? I'm thinking maybe there's some issue with your infrared thermometer or possibly how you may be using it?
 
Hw monitor shows

GPU1
volts- 1.018v
temp 67c max
fans-- 686rpm
Fans pwm 25%
graphics clock @595
memory clock @ 3610
Utilizations--Gpu power 20.93%

GPU2
volts-0.868 v
temp-30c max
fans-- 1575rpm
Fans pwm 56%
graphics clock @ 405mhz
memory clock @ 405mhz (not a typo)
Utilizations--Gpu power 10%

I also, ran each card by themselves and they are about the same.

Heaven-Dx11/2560x1440/8xAA fullscreen/shaders-high/textures-high/filter-trilinear/occlusion enabled/refraction-enabled/Volumetric-enabled
FPS: 111.9
Scores: 2818
Min FPS: 48.9
Max FPS: 149.2

Valley-Dx11/2560x1440/8xAA fullscreen/ultra quality
FPS-61.6
Score-2576
min fps-31.7
Max fps-118.1

Not sure if this is what you needed! Thanks for your help.

 
So now the top one is the hotter one which is pretty standard. I would definitely be alarmed by the fan running too slowly on the top one (GPU1). That seems wrong. And because GPU1 is running slowly, GPU2 will follow. Make sure the power connections are connected properly.

Remove GPU2 and try to run GPU1 by itself.

Is a fan not spinning on GPU1?
Does it run fine by itself?

Do the same for GPU2 removing GPU1 and putting GPU2 in the GPU1 slot.
 
Ubercake,

I ran both cards by themselves. They both ran fine, fans spin fine on both. Specs below

HW monitor

GPU1 #34
volts- 1.037v
temp 40c idle
fans-- 1687rpm
Fans pwm 61%
Gpu power 20.83%
graphics clock @595
memory clock @ 3610


GPU2 #35
volts-1.037v
temp-42c idle
fans-- 1682rpm
Fans pwm 60%
Gpu power 20.57%
graphics clock @ 595mhz
memory clock @ 3610mhz



 
I have had the Psu for more than 4yrs!! My motherboard is a Asus Maximus vi hero.

So I ran both cards separately in second slot and they both work fine. About same numbers as yesterday, Valley and Hwmonitor.


Just came across this article about this GPU. If you scroll to the "power consumption" part, Barsho90 recommends a 950w PSu when running SLI with these cards.....see link below.

http://www.overclock.net/t/1563088/guru3d-zotac-geforce-gtx-980-ti-amp-extreme-review

Is my power supply not enough??? What do you think?

Thanks
 
Both cards work fine in both slots individually. So you know it's not the cards

Put both cards in and make sure the SLI bridge is seated correctly. Be sure to activate SLI under 3D settings and that, assuming you only have one monitor connected, set "Single Display Performance mode" under Nvidia control panel's 'Manage 3D settings->Multi-display/mixed-GPU acceleration" and under "Power Management Mode" set it to "Prefer Maximum Performance".

Turn Anitialiasing transparency "on" under global settings in Nvidia control panel. Then set AA as high as you can in BF4 and use HWmonitor to determine the CPU and GPU utilization while you play multi-player BF4. Also, check out the temps. Does the CPU hit 100% utilization on any cores/threads?

What does it show for GPU utilization for each of the cards? Temps?
 
Solution


Ok done all you requested.

Anitialiasing transparency "on" 8x super sample (should this be set them same as in game at 4x AA)
BF4 AA in game set at as high as it could go----4x

Was on BF4 mutliplayer, used HW monitor for results. Please see photos for results.

CPU

14jpzpk.jpg




GPU's

2yud3ig.jpg




Since the changes you had me make, the temps difference between the 2 cards is still about 20 degrees. The down sample is no longer happening. I have since adjusted the fans settings thru EVGA Precision, now about about 10 to 17 degrees difference!

thanks


 
I bet it looks amazing with those two cards cranking. GPU utilization looks great!

So your temps for both cards are still well within tolerance (92C max) for that GPU. I think your temperature difference from one card to the next is greater than the norm because you have three-slot cards. After all of this, I'd say your cards are working fine.

Because of the fan design, the heat is exhausting directly into the case. If your side fan is currently sucking air in, try turning it around to blow the air out and see if that makes a difference. It may make things a little cooler (or a little hotter). Whichever yields better temps, keep that.
 



Ubercake,

I turned the side panel fan around to push air out, but it's still around the same temps. The Gpu's are working fantastic and it does look freakin awesome!!

Thank you for all your help. You da man!