Hi,
I changed the thermal paste today. And because I read your posts before, I documented it to some extend.
Executive summary: Temperatures went down, noise level for balanced and full power modes in microsft windows went down, noticeable, but not much. I will try to reduce the available CPU computing power manually, if possible. We need an effective power mode between power saving and balanced.
Ok, here come the details:
Pictures from the old thermal paste situation:
The third picture shows some cleaning effort from me. I didn’t remove all the paste from the small capacitors to avoid chipping one off. Unfortunately, I didn’t make a photo after re-applying the paste, but I distributed it all over the die, and thicker than needed, so there would be even some room for improvements. I used Thermal Grizzly Kryonaut, but I guess there are also other comparable or even better thermal pastes.
How I compared the before and after situation: It was not perfectly the same condition, but I did the following: I started CPU-Z and started the CPU stress test for multi core, 16 threads. Further, I opened sons of the forest, single player, new game. With the old paste, the starting point was in the moutains in snow, and after applying the paste, the game started at the water front. So there might be some difference in GPU and CPU load. I further was running around a bit to ensure that the load increases a bit. I measured in the modes energy saver, balanced, and full power.
The following figures show the results:
Energy saver, first old paste, then new paste
Next:
Balanced mode, first old paste, then new paste
Next:
High power mode, first old paste, then new paste:
You see an improvement, and to a low extend this was reflected by the fan noise. But as I said, to play with reduced noise, I guess we have to limit the cpu/gpu load manually.
Another factor to be considered: I applied the new paste, and directly started the notebook and started the tests. I don’t know how the results will be in a few days when the paste is in its “final” position. Maybe it is more thermal conductive at the moment. lets see.
So I hope that helped a bit!
Framework team: can you support us with some hints about manual power limitations for CPU and GPU? SO a more granular steering of the power usage?
br
Tom