PC Graphics

[Fire_raptor]

Hi, my junction temp in Division 2 can go to 110C (GPU RPM is at 1100). After that I had taken care of my case airflow, I put some better fan and installed upper fan. Things were better but huge difference was when I put my GPU BIOS to silent option (lower clock) and set custom fan curve to max of 1700RPM. In Firestrike 4K test GPU hotspot is now 87C which is pretty good, right?

I would like to lower temperatures it a little more, can you tell me what settings do I need to change for undervolt? What values to aim?

I read before that some GPU can have problems with bad heatsink contact and because of that it will have hotter hotspot, that only solution is to change thermal paste but my GPU is still under warranty and since I have never change thermal paste of GPU I don't dare to mess with this. 

[johnnyenglish]

If find strange a massive card like that with 3 fans get those temps. Anyway, the way you can test if bad contact exists, do a gentle press on the card heatsink, against the pcb.

Have you tried the auto undervolt option? 

[Fire_raptor]

I already tried tighten up the screws. I understand that this is problem but this is not nearly as bad as some GPU which hotspot is always near 110C.

Is it better to try manual undervolt? What value (mV) should I pursue?

[Fire_raptor]

In what program and how long to test stability? For GPU I have 3dMark and Unigine Superposition.

[Fire_raptor]

This is my current default settings with custom fan curve (up to 1700RPM). 

Tested in Fire Strike extreme (third column is max value during test)

This is with 1120mV, max frequency is set to 2500Mhz and custom fan curve up to 1700RPM

Little loss in perfomance, but hotspot is 9C lower.

How will I recognise if I went to low with voltage? Artefact on screen, black screen or computer freeze?