GT 650M core clock decreases steadily in-game

[Ebeb]

Hello everyone,

I have a Samsung NP550P5C-S02 which includes a GT 650M (GDDR3) graphics card. I was quite satisfied with it until it started to give low fps values in games. I noticed it with League of Legends first; fps dropped to 15-20 range during teamfights; so I swithced LoL to integrated graphics (which gives better performance -_-). But whenever I try to open Skyrim, the fps drops in the first couple of minutes drastically, becoming unplayable after 3-5 minutes. I have played lightly modded Skyrim in this computer 3-4 months ago and I had reasonable fps at High setting.

The GPU-Z logs show that GPU temperature increases steadily over 30 seconds hitting 90 degrees C (while core and memory clock speeds are normal) and then core clock starts to decrease, eventually decreasing to idle values 135 MHz. During this time GPU temperature is constant at 90 degrees C. I have tried latest driver for this also stock driver for this laptop, similar results. I remember playing this game with my hand nearly burning at palm rest, but frames were still okay. Now I don't feel like the temperature is going that high but it still decreases clock frequency 🙁 I actually think it reads wrong temperature values.

I appreciate any kind of help

 

[Robert Cook]

How old is this laptop? it may be full of dust (air can it), also, if its warranty is gone, try opening it up and giving it fresh thermal paste. I find it unlikely that the temp probe is off by that amount.

 

[Ebeb]

Well it's a little more than 2 years but it has seen service once since the plastic thing under the keyboard (i guess it was the wiring, it had metal roads on it) melted right on top of GPU fan 😀 And I explicitly told them to remove the dust and reapply thermal paste, but of course I can't be
sure. The issue is even my idle temperatures are at 60-70 C and I am not doing anything! I will try to remove the dust, but I still don't think these are the right values.

Also is there a software that overrides GPU temperatures and tells it to carry on? Maybe that will work or just burn the card 🙁

 

[Robert Cook]

Either you fan got killed or the "service" people botched their jobs, it seems they may have even forgotten thermal paste if the temps are that bad. There is no safe way to override the GPU, a custom Bios, may work, but at the cost of bricking the whole laptop. Once again, I think there is almost no way to fool a temp sensor. if your warranty is gone open it up and see what the GPU has for thermal paste, and if the heat sink is even on right, also see if the fan works.

 

[Ebeb]

Well right now I can't apply a new paste so I don't want to remove heat sinks, but I will surely check them. I don't think service did anything at all. Thanks for your quick response

 

[Ebeb]

Just wanted to add that applying new paste solved my problem. Today I removed the heat sink, it was both dusty and there were thermal paste everywhere, except on the gpu core...I cleaned it and put a little amount of Xigmatek paste. Problem solved