Strix 980 Ti in SLI. One runs 13 degree hotter than the other, why?

P

pclausen

Gawd
Joined
Jan 30, 2008
Messages
697
I got a pair of Asus Strix GTX 980 Ti OC GAMING cards.

The first one was purchased new from NewEgg a week ago and has a ASIC of 75.0% The 2nd was purchased from a member here a few ago. It has a ASIC of 73.6%, so pretty close.

So without any overclocking other than what Asus did, I'm seeing significant temp differences. At the start of the test they are both about the same as seen here:

heavenbench01.JPG


But a few minutes into the test, the 2nd card is running at 83 degrees where the existing one is at 70 degrees.

heavenbench02.JPG


An ASIC difference of 1.4% could not account for such a temp difference, could it?

I plan to put these on water (EK blocks), but I'm concerned that 2nd card won't OC well at all given the stop temps I'm seeing.

The 3 fans on the hot card appear to be spinning at the same rate of the fans on my existing one.

And it's not a case air flow issue either and they are in an open chassis. The one running hot is on the left.

sli01.JPG


Any ideas?
 
Last edited:
Is it possible for you to stand the open air case up vertically and see if that makes a difference? It's normal to see one card hotter than another, but not by this much. Another guess would be that the hotter card has a poor thermal compound application between the core and the heatsink. With Maxwell ASIC doesn't really matter, and it certainly wouldn't affect the core temperature in such a drastic way.
 
I stood the chassis up vertically (actually the default configuration), and it didn't make a difference. 70 vs. 83 still.

build-10.JPG


At idle the hot card is at 59C and the other one is at 52C.

I ran the heaven bench again (window mode), and this time captured the sensors from GPU-Z (running 2 instances at the same time.)

temp2.PNG


It is interesting to note that the GPU load is 76% on the cool one and 99% on the hot one. This remained true for the entire test. So maybe this benchmark isn't very good at splitting the load between the GPUs?
 
The difference in utilization does seem odd, but I haven't used SLI in awhile. I would definitely try different benchmarks and games to see if the issue is happening consistently over a range of software.
 
I ran 3dMark and captured the Max values from the sensors in GPU-Z. They were as follows:

temp4.PNG


So on this benchmark, both GPUs are equally loaded, but one still runs 10 degrees hotter than the other. It is interesting to note that the cool running one has a higher VDDC voltage (1.2000 V vs. 1.1870 V on the hot one).

The actual score I got was:

firestrike01.PNG


Which seems about right for stock clock speeds.
 
Same thing happened with my Asus DirectCU II 780s...the primary (top spot) always ran 10-20C hotter than the secondary (bottom slot).

This was with my current Z77 Gene MoBo where the two GPUs were damn near touching. I always wondered if spacing them apart by one more slot width would help, like with my Z77 Sabertooth...but I never tried it.
 
Try switching the two cards. If the card closest to your CPU still runs hotter, you know it isn't a hardware problem.
 
One card draws in air that is warmed by the rear of the other card.
Top cards always run hotter.
The heat from the rear of 980ti cards is substantial too.

I fitted an AC Xtreme IV cooler to my ti which has, what I thought were, over the top finned heatsinks on the back.
But they still get incredibly hot such that I fitted 2 fans to help cool them.
Even then they are still very hot.

That and it looks like the top card runs a bit hotter anyway probably explains the difference.
 
Actually, the card closest to the CPU is the one that runs cool. I'll swap the cards around just to make sure, or maybe just bench one card at a time to make sure they have similar temps.

Once my EK blocks gets here and they are on water, that will hopefully eliminate the issue completely.
 
My top card runs 8-11c difference as well but i don't let them go higher than 80c (set a fan curve so i don't get throttled)
 
You guys were right, the top card is the one that gets hotter. Anyway, I tested them one at a time, and the results are almost identical:

The one that was running hot in SLI is on the right side in the below comparisons.

Right after completing Heaven. Both cards max out at 73 degrees. Same as the bottom card in SLI.

sidebyside02.PNG


Right after completing Heaven with fan at max. "Hot" card got to 55 where the "cool" card only reached 53.

sidebyside03.PNG


Heaven score of each card. Pretty much identical.

sidebyside04.PNG


So they are very close. That pretty much confirms that the top card gets heated by the lower card. This also happens, to pretty much the same extent, when the mobo is horizontal.

Once I get my waterblocks installed, this issue should go away completely I would think.

Appreciate the tip on setting fan curves, need to look into that until I'm on water.
 
Last edited:
Yeah Nenu hit the nail on the head.. that and the top card being the main frame buffer for the SLI pair usually gets a little more load anyways.
 
So with SLI should we swap the cards around every 6 months or would that make any difference?
 
Lol, yeah, those are cooling an empty radiator for now. :)

Still waiting on all my other water parts (mobo and strix water blocks, pump, hoses, etc). I'm swapping out the Maximus VIII HERO for the EXTREME as well. I need to be able to run my 10Gig NIC along side the SLI and the 950 Pro.
 
This happens with Crossfire too, the "master" card is always hotter. When I had 280x Crossfire the top card was always about 10 degrees hotter at load.
 
Both my cards run the exact same temps - but that is because they are both water blocked which tends to balance and hold the temps low. As others have stated, the SLI primary card is pushed a bit harder as it handles a bit more of the house keeping chores in SLI and therefore tends to produce a bit more heat. Air cooling isn't as effective and you'll typically see the master card running hotter. Also, If you have one card with a dramatically higher ASIC, I'd recommend running it as your primary in that it will handle the heavier load a bit better by running slightly cooler.
 
It has to do with the top card pulling in the hot air off the back of the bottom card. This is typical of sli unless under water, or another semi-solution is using blower style cards. Nothing to be too concerned about
 
You must log in or register to reply here.
Back
Top