[SOLVED] My OC is too high after thermal paste and pad replacement - GTX 1060 SSC 6gb

[BrownStar]

I have an EVGA GTX 1060 SSC 6gb that I’ve had overclocked to a stable +125 core and +450 mem for months now. Under load in Heaven benchmark it would average about 72 degrees before new paste and pads. I used Artic MX-4 and Thermal Grizzly Minus Pad 8 (1.5mm thick, cant seem to post pictures right now).

I was mind blown to see that it now stays at a consistent 60 degrees while running Heaven benchmark! Seems like witchcraft to me.
HOWEVER, now my original OC of +125/+450 isn’t stable in the exact same benchmark!?!

Any thoughts???
All system and graphics settings stayed the same. My only thought is maybe the thermal pads are too thick covering the VRAM modules and possibly hindering efficient heat transfer causing them to heat up and the benchmark crashes?

 

[TechyInAZ]

Wow that's crazy! If anything the lower temps should give you even more stability with that overclock LOL!

But seriously, make sure the thermal pads are the exact same size as the original ones, anything thinner will give you less thermal dissipation, and any pads that are too thick will put pressure on the VRAM and PCB, which can lead to micro stutter and possible instability. Because the thermal pads can cause the thermal pads to bend in weird directions.

If the thermal pads are the same thickness, you could have possibly overtightened the mounting screws for the card, if you really cranked down on the card that is.

 

[Deleted member 362816]

I think you are on to something with the thermal pads. Do you happen to know how much thicker they are compared to the stock ones?

 

[Phaaze88]

The overclock probably was never really stable. The Gpu Boost algorithm gave you the illusion that it was.
The lower core temperature allowed for longer and higher boost bins, so combined with the +125 core clock, it spent more time at that and crashed because it wasn't stable.
At the previous thermals, Gpu Boost likely dialed down clocks earlier.

This software monitors the gpu's parameters, and will dynamically make its own adjustments to the gpu - ignoring your own settings - based on said parameters.
It loses boost bins depending on how frequently the board power limit is hit.
It loses boost bins the hotter it runs.

With Gpu Boost present, manual overclocks can't be verified stable if the frequency can't be kept static. For example, my 1080Ti likes to boost to 1949mhz core clock:
-It'll hold 1949 all throughout Game A.
-In Game B, it'll run 1949 for the first 5 mins, after which it'll sit at 1936 for the remainder of the session.
-In Game C, it bounces around between 1835-1949.
-In Game D, it bounces around between 1911-1936.


GDDR5 and R5X Vram can be cooled just fine on air alone - pads and sinks are ok, but not necessary.
It's kind of like the gimmick with Arctic's Liquid Freezer II VRM fan: if that little fan is actually saving motherboard VRM operating thermals in any meaningful way, you've got bigger problems.
The Vram doesn't have boost bins based on temperature - only the gpu core does, so whether it runs a few degrees higher or lower doesn't change anything for it.
Now, GDDR6 and 6X, that won't work; those need pads and sinks.


Nvidia killed off Gpu OC for more casual users, as well as taking the fun out of it, with Gpu Boost. It's built into the card, so no, you can't remove it.