I overclocked my GPU but the voltage decreases during load which makes it crash.

SyMix_KD

Reputable
Mar 14, 2014
66
0
4,640
So I overclocked my GTX 650 Ti using msi afterburner
I added +100 to the core voltage (1.125v)
+200 to the core clock (1124 Mhz)
and +350 to the memory clock (3055 Mhz)

I then ran Unigine Heaven Stress test while paying attention to the on-screen display.

And I noticed this, the overclock was stable for the first minutes while the voltage was at 1.125v, then for some reason it decreased automatically to 1.112v, still stable, a few mins later it decreased again to 1.10v which made the gpu driver crash 3 seconds later.

I have "Force constant voltage" checked in the MSI afterburner settings so I don't know why this is hapenning, it's just stupid, why would the GPU undervolt itself with no reason?

The GPU temps are around 73ºC max.

I'm using the latest drivers and windows version.

I would really like someone to help me.

Thanks! Let me know if you need extra info.
 
Solution
Your GPU chip temps may only be hitting 73 degrees but it's possible that the VRMs (voltage regulator module) may be overheating and causes the forced decrease and subsequent crash.

SyMix_KD

Reputable
Mar 14, 2014
66
0
4,640
The thing is, if I reduce the overclock to +25/+50/+100 (voltage/core/memory), it still goes from1.050 to 1.037 for no reason a few seconds later after being under load. If VRM temperatures were the problem here, shouldn't it take at least longer to undervolt since this OC is much lower than the first one?

EDIT: Just tried now: even underclocking the card to -25/-50/-100, it still undervolts! (from 1.000v to 0.987v few seconds later)

EDIT2: It was indeed the temps, I did some testing, basically let it run on stock speeds but at +100 voltage on furmark, it decreased after some time to 1.100v, after that I crancked up the fans to max and the voltage increased to 1.112v shortly after. I noticed my graphics card has no heatsinks on the VRM's, I had some tiny lying around aswell as some thermal pads, so I applied them on the VRM's and I managed to stay at a stable 1.112v which is enough to not make my games crash. Thanks for reminding me of the VRM's!

 


No problem, glad that I could help. Do you happen to have a link to the small heatsinks and thermal pads that you used for the VRMs?
 

SyMix_KD

Reputable
Mar 14, 2014
66
0
4,640
I didn't buy these, I just had some lying around that I took from old broken motherboards and power supplies, I also had a big thermal pad that you can cut off into little ones that fit your needs, but I bought this one a long time ago, I'm pretty sure you can find them pretty easily. I also ended up switching the cooler. Here I'll show you some pics:

https://imgur.com/a/OeVQd
 


That is one wicked looking GPU, I've been debating for a long time getting a universal GPU water block and using small heatsinks on the VRMs instead of buying a full waterblock that is specific to the card I own. That way if I upgrade I don't have to buy a new block.
 

SyMix_KD

Reputable
Mar 14, 2014
66
0
4,640
The problem is securing it in, I actually had to turn my case upside down because the heatsink is so heavy it wouldn't lean on the dye properly which would make the temps sky rocket when I ran anything. How will you get a GPU water block that fits on every single GPU? I'm kind of curious