Slave GTX 980m running way too hot in p870dm

I have a p870dm with dual GTX 980m's in SLI. I've been battling cooling issues for a while with this laptop and can't seem to figure out what the cause is.

With fans at 100% the Master gets to about 75C in games such as the Witcher 3. The slave however runs MUCH hotter. Usually about 10-15C hotter than the master.

Also, whenever the slave reaches 86C the laptop hard crashes. It doesn't throttle, just hard crashes immediately when the slave hits 86C. It can be at 85C all day and not have an issue, but as soon as it hits 86C, it hard crashes. What is going on here? Aren't the cards supposed to start throttling at 92C or 93C with prema's vbios?

Both cards are running prema's vbios v2 and I've repasted multiple times with IC Diamond.

Additionally, I've tried undervolting with nvinspector, but if the voltage is reduced below stock, the settings won't stick.

 

similar issue here:
Clevo p870dm-G

Seems like one GPU runs hotter than the others in SLI setups due to the not so great heatsink setup on these laptops

 

What's strange to me is that the laptop will hard shut down when the slave hits 86C. Doesn't thermal throttle. Just hard shut down. I tried disabling SLI, and smothering the Master's vent to test if it had the same issue (since its never actually gotten to 86C in games). The Master hit 92C and started throttling like it should. So why is the slave behaving differently in this sense?

Also, I've already removed the plastic sheets and foil blocking some of the vents on the back cover.

 

can you try the Prema VBIOS please

GSYNC VBIOS for 980M

NON-GSYNC VBIOS for 980M

 

I'm actually already using prema's v2 vbios on both gpu's.

 

Maybe @Prema can comment why isn't the throttling kicking in on the 2nd GPU to prevent the hard lockup

 

You should try also grizzly paste (easier and better, i earn 5°c from ic7).
Do you apply generally a nuts or you paste the full core? Last one is better. If you have time without good result, you can check also if there is a real full compression between bga and heatsink. Then you can mod the heatsink to give more compression and also installing more copper pads to increase surface to dissipate heats faster.

 

I actually got way worse results with Thermal Grizzly Kryonaut because it's not as thick as IC Diamond so it doesn't make very good contact. see: http://forum.notebookreview.com/threads/clevo-overclockers-lounge.788975/page-303#post-10270401

 

Try switching the master and slave cards around in the machine. If the master is so much cooler than that should solve the problem since the master won't get as hot anymore.

With that said it sounds like something is defective with that card. Have you tried flashing back to stock and seeing if the behavior continues, switching drivers, etc?

 

The system should never hard shutdown unless the ACAdapter trips.
GPU hard shutdown value is 94c sustained for more than 30s.

Slave is all ways hotter on P870 because of the smaller FAN.
Do as Ethrem and the others suggested and swap and repaste with ICD+ grade thermal paste.

With my Mod you could also set a lower thermal limit in Inspector to introduce throttle.

 

Also the VBIOS does not support undervolting IIRC, you can try modding it to set a lower stock voltage then move it up from there.

 

Maxwell doesn't like to be undervolted. It can make the cards incredibly unstable especially when switching power states.

 

So I actually took a look at this machine in person. Swapping the Master card to the Slave position resulted in the same problem, which eliminates a bad card. There's definitely something wrong with the cooling.

So we tried lapping the heatsink. It was very, very warped. This fixed the hard shut down at 86C, so my guess is that only a portion of the die was reaching a critical temp and triggering the shutdown due to the warped heatsink.

However this did not fix the overheating. Now the slave reaches 92C in a matter of minutes and thermal throttles. The fan works fine and there's plenty of airflow under the laptop and through the vents. My guess is maybe the gas in the heatpipes leaked out. So OP is ordering a new heatsink and will report back.

 

I would make the same assessment in that case. Good luck with the replacement.