F5F Stay Refreshed Power Users Overclocking Different GPU loads in sli

Different GPU loads in sli

Different GPU loads in sli

C
ClumsySky
Senior Member
526
09-05-2016, 02:27 AM
#1
Hi everyone.
I’m currently working on overclocking two MSI 970 gaming cards in SLI. I’ve been experiencing crashes even at 80+ core clock, though 70+ seems okay. It might just be bad luck, but it feels much lower than others I know. I checked gpuz and everything seems to sync properly, operating at similar speeds. There’s a 25% difference between the two cards. Could this imbalance be causing the instability?

I opened the NVIDIA control panel and set Physix to auto-select. I chose one card as dedicated, and the loads balanced between both.

I’m running tests again—this process is slow and sometimes frustrating. It’s hard to say for sure, but imbalanced loads might be the reason for the crashes. Would this mean my OC is unstable? Any feedback would be appreciated.
C
ClumsySky
09-05-2016, 02:27 AM #1

Hi everyone.
I’m currently working on overclocking two MSI 970 gaming cards in SLI. I’ve been experiencing crashes even at 80+ core clock, though 70+ seems okay. It might just be bad luck, but it feels much lower than others I know. I checked gpuz and everything seems to sync properly, operating at similar speeds. There’s a 25% difference between the two cards. Could this imbalance be causing the instability?

I opened the NVIDIA control panel and set Physix to auto-select. I chose one card as dedicated, and the loads balanced between both.

I’m running tests again—this process is slow and sometimes frustrating. It’s hard to say for sure, but imbalanced loads might be the reason for the crashes. Would this mean my OC is unstable? Any feedback would be appreciated.

M
Minetoblend
Member
110
09-05-2016, 03:50 AM
#2
The issue arises in SLI because the card nearest to the CPU tends to overheat more. This happens since SLI isn't flawless; the main card usually carries more load than the secondary. Another factor is that the top card sits between the CPU and the bottom card.

To address this, ensure your case's airflow is adequate. Then identify the top overclocker and test each one separately in the machine. Place the overclocked card at the top—this is where it will consistently run hotter. Even though it might overheat more, its larger overclocking headroom should help balance things out.
M
Minetoblend
09-05-2016, 03:50 AM #2

The issue arises in SLI because the card nearest to the CPU tends to overheat more. This happens since SLI isn't flawless; the main card usually carries more load than the secondary. Another factor is that the top card sits between the CPU and the bottom card.

To address this, ensure your case's airflow is adequate. Then identify the top overclocker and test each one separately in the machine. Place the overclocked card at the top—this is where it will consistently run hotter. Even though it might overheat more, its larger overclocking headroom should help balance things out.

X
Xekarion
Junior Member
12
09-05-2016, 05:13 AM
#3
The issue lies in SLI because the card nearest to the CPU tends to overheat more. This happens since SLI isn't flawless; the main card usually carries more load than the secondary one. Another factor is that the top card sits between the CPU and the bottom card.

You should follow these steps: First, ensure your case's airflow is sufficient. Then, identify which card performs best with overclocking. Test each card separately in the machine. Place the one that overclocks better at the top—this card will consistently run hotter. This approach lets the stronger card handle more heat, even though it may be the one with less headroom.

Thanks for your advice. I plan to dive deeper after a bit more testing.
FYI, my temperatures stay under 70°C. The hottest card is still that one. Using it as a physical overclock in the NVIDIA control panel balanced GPU loads. Previously, loads were about a 25-30% difference. It seems stable now—thank you! Over an hour, not just 20 minutes like before. Do you think uneven load distribution might have been the main problem? I’ll keep testing with Heaven and Prime95 for confirmation.
Appreciate your guidance. Cheers
X
Xekarion
09-05-2016, 05:13 AM #3

The issue lies in SLI because the card nearest to the CPU tends to overheat more. This happens since SLI isn't flawless; the main card usually carries more load than the secondary one. Another factor is that the top card sits between the CPU and the bottom card.

You should follow these steps: First, ensure your case's airflow is sufficient. Then, identify which card performs best with overclocking. Test each card separately in the machine. Place the one that overclocks better at the top—this card will consistently run hotter. This approach lets the stronger card handle more heat, even though it may be the one with less headroom.

Thanks for your advice. I plan to dive deeper after a bit more testing.
FYI, my temperatures stay under 70°C. The hottest card is still that one. Using it as a physical overclock in the NVIDIA control panel balanced GPU loads. Previously, loads were about a 25-30% difference. It seems stable now—thank you! Over an hour, not just 20 minutes like before. Do you think uneven load distribution might have been the main problem? I’ll keep testing with Heaven and Prime95 for confirmation.
Appreciate your guidance. Cheers

M
Minmonter
Member
56
09-18-2016, 10:48 AM
#4
Physx in Heaven? I've heard it's not really supported. I just tried it on my machine, and the card I set aside for Physx was at 0% usage. Unfortunately, having an unbalanced load is unavoidable with SLI – that's why you should assign the weaker cards the lighter workload to maximize your overall performance.
M
Minmonter
09-18-2016, 10:48 AM #4

Physx in Heaven? I've heard it's not really supported. I just tried it on my machine, and the card I set aside for Physx was at 0% usage. Unfortunately, having an unbalanced load is unavoidable with SLI – that's why you should assign the weaker cards the lighter workload to maximize your overall performance.