T-Troll / alienfx-tools

Alienware systems lights, fans, and power control tools and apps
MIT License
458 stars 39 forks source link

Inaccurate GPU temperature #265

Closed Charles-IX closed 1 year ago

Charles-IX commented 1 year ago

It seems the GPU temp detected in alienFX is always lower than that displayed in taskmgr and GPU-Z. I can only guess that alienFX can not read the data sent from Nvidia driver but use a different BIOS sensor. image Is there any possibility that alienFX can read the temp more accurately ?

T-Troll commented 1 year ago

This is a known bug for some systems. Dell take incorrect sensor data into BIOS, and they not fix it for months. As a fix, use one of ESIF sensors (seems like ESIF 3 in your case) instead - they are correct.

BTW, what you BIOS version to report this to Dell?

PS: Seems like this time they take data from Ambient sensor instead of GPU ^_^

Charles-IX commented 1 year ago

After using for a while I found none of the ESIF sensors actually stands for my GPU temp. In one situation temp from ESIF sensor 1 might be more close, in another situation ESIF sensor 2 gave me more "accurate" temp. Fun in fact, none of them were even close to the temp shown in taskmgr in some cases. image I went to Dell.com and found a newer version, which is 1.21.0. I will update it and see if they had fixed it.

Charles-IX commented 1 year ago

Now my BIOS version is 1.21.0. They had not fixed it at all. None of the ESIF sensor value match the taskmgr GPU temperature. I added a curve in each of the 3 ESIF sensors. I can't think of a better idea to solve this... Or maybe you could offer some suggestions? I'm all ears.

T-Troll commented 1 year ago

Well... My tools support Libre Hardware Monitor readings (but start it before tools). So you can use sensors from it as well (they use different method to retrieve it). This also help to configure out which one across ESIF is really GPU.

Don't rely on task manager, it's quite inaccurate (approximate).

T-Troll commented 1 year ago

I got an answer from Dell. This is... Optimus. They always take dGPU temp, and then chip is not active... you will see random value. Can you please check (start game and look at temperatures).