Originally Posted by cplifj
does anyone happen to be running corsair's link software and also happen to get clocks stuck at max after gaming or compute loads ?
it doesn't seem to happen when corsair's link software isn't running, this also monitors temps and fanspeeds of everything in the system.
more monitoring just seems to conflict with each other and then some.
I would highly recommend to limit the number of monitoring softwares running, especially ones polling the gpu as it will have an effect on the devices overall performance and stability... now it can't be helped when your trouble shooting to tuning, but in general; my humble suggestion is to avoid it. As you have suggested the different softwares pull from the same resources to read back the / poll the state.
Originally Posted by orlfman
anyone know what's the max, safe hotspot temperature? i tried googling but find conflicting information. from 90c to 115c. i noticed on my air cooled vega at stock settings, bios, and profile hitting up to 86c. though average is around 80c.
I have read / heard many numbers tossed around... and on this topic... to my knowledge it has never been identified by anyone credible from AMD / Radeon in the technical department. I have called it a nuthin' burger... and "believe" it to be a current - table referenced - diode near the bottom of the die mass on top of the base of the interposer (just between the mem and proc/soc). It scales beat for beat with memory loading with all my testing that I have done. I have seen no appreciable collaboration between it and performance / stability. The caveat to that statement is that I run about a 10-15 C delta from it to my HBM temps on my setup under load so I "maybe" see 45 C after an hour or two of straight' gaming or benching. This is on a non-molded chip first gen ref card.
Originally Posted by Grummpy
Just ignore hotspot.
ive done tests and it looses its temp instantly when you close stress test
its just nonsense .
you can apply any result from resistance on thermal diodes.
thats i i think anyway, anything that has instant temp drops tellls me it was just a estimate
It can be precision rather than an estimate, but most of these setups are designed to be "more" accurate within a certain range. Depends on the granularity of the look up table. I do agree that its a load dependent... in that sense, an estimate perhaps of the thermal delta or gradient of the entire package.
All this reminds me of my ole' Phenom II, oc' it to kingdom come, but if the temp read ever >50C a crash was imminent. Mind you stock... 70C + was a non-issue. AMDs like it cold for sure, not sure if this is at all relate-able to the hot spot talk though.