GPU hotspot?

[dr.stevil]

I was doing a bit work on my machine, trying to tune my fan speeds, and happened to notice the hotspot sensor in HWInfo after a run in superposition with a temperature of 105*

This is a bit baffling since all of my other temps are significantly lower. The second highest temp I’m seeing is one of the memory modules in the low 80’s… core never gets hotter than the mid 60’s.

I haven’t noticed any throttling to speak of so I’m sort of stumped why this sensor is reading so high. I went and ordered some new thermal paste and pads to see if that will help, but the temperature delta is stupid high. I’ve read some forum posts that suggest a delta of 10-15* is fairly normal from the core, but 40*+ seems really excessive.

Anyone have any input about this sensor? Should I be worried? Does anyone happen to know exactly where this sensor is located?

Almost forgot, the GPU is a RTX3090 FTW3 with a Hybrid cooler. To check my sanity, I did a second run of super position with every fan at 100% (very loud) and was still seeing hotspot temps well over 100*

 

[NattyKathy]

Hotspot isn't a single fixed sensor on newer NV and AMD cards, it's the highest reading from an array of sensors in the GPU die

What I've generally seen from a high-power GPU on AiO is 60° or below core average with 15° or less hotspot delta with the card running heavy real-world workload. Mid-60s core avg is on the high end but not out of range, 40° core hotspot delta is very not right though.

I'd check thermal pad thickness (could be affecting contact w core), verify the is AiO functioning properly, and check case airflow.

 

[pendragon1]

that^^ mostly, my normal difference is about 20c. i would try something else, like afterburner, to see if it sees the same oddly high temp first.

 

[CAD4466HK]

that^^ mostly, my normal difference is about 20c. i would try something else, like afterburner, to see if it sees the same oddly high temp first.

GPU-Z is good for logging hotspot thermals as well.

 

[arestavo]

Hotspot isn't a single fixed sensor on newer NV and AMD cards, it's the highest reading from an array of sensors in the GPU die

What I've generally seen from a high-power GPU on AiO is 60° or below core average with 15° or less hotspot delta with the card running heavy real-world workload. Mid-60s core avg is on the high end but not out of range, 40° core hotspot delta is very not right though.

I'd check thermal pad thickness (could be affecting contact w core), verify the is AiO functioning properly, and check case airflow.

Can't the hotspot also be a VRM on an AMD card? That's what I've seen others say at least.

 

[Koldur]

Hotspot is the hottest part of the GPU die itself, which in this case means the die doesn't have full contact with the heatsink and or not enough thermal paste between the heatsink and GPU die.

This is the case for 30X0 series card at least, probably with 40X0 series too, but not enough definitive info on that yet.

 

[dr.stevil]

Thanks for the clarification folks. Good information to know.

I’ll repaste and apply some new pads and will let y’all know how it goes.

 

[dr.stevil]

Wanted to update this thread in case someone else stumbles on this in the future.

I ordered some new thermal pads and grease and reinstalled the block… temps across the board are slightly better but the hotspot dropped from 105* down to around 88*. Higher than I would like, but still a significant improvement all the same.

it was difficult to identify exactly what the issue was as there wasn’t any clear indication that the block wasn’t making good contact with the die or anything, but I will say that the paste was much crustier than any other GPU I’ve changed it on, so maybe it was using old paste from the factory?

At any rate, I’ll certainly be keeping an eye on it but I think it’s safe to say that it’s fixed now.

 

[NattyKathy]

Glad to hear it's improved, OP! It's possible there's a mounting pressure issue with that AiO that was compounded by the dried-up thermal paste. There won't necessarily be visible signs of bad contact when temps are like yours were, if the mount was really bad the average temps would be super high too- but ok average with high hotspot usually seems to mean a moderate mounting pressure deficit, thermal paste problem, or slight heat saturation issue.

 

[mnewxcv]

Wanted to update this thread in case someone else stumbles on this in the future.

I ordered some new thermal pads and grease and reinstalled the block… temps across the board are slightly better but the hotspot dropped from 105* down to around 88*. Higher than I would like, but still a significant improvement all the same.

it was difficult to identify exactly what the issue was as there wasn’t any clear indication that the block wasn’t making good contact with the die or anything, but I will say that the paste was much crustier than any other GPU I’ve changed it on, so maybe it was using old paste from the factory?

At any rate, I’ll certainly be keeping an eye on it but I think it’s safe to say that it’s fixed now.

nothing wrong with an 88c hot spot. That means the core temp is probably in the 60s.

 

[dr.stevil]

nothing wrong with an 88c hot spot. That means the core temp is probably in the 60s.

Yeah, that’s exactly where it is (at least when I’m stressing it)

What actually really surprised me is, during testing, I was trying less demanding games (the long dark) @4k and noticed the temps hovering in the mid 40’s where previously it would start to creep into the high 60’s. The fan noise would bother me so I’d step it down to 1440p (and even then, the core temp was still in the 50’s)

Can’t believe this has been a problem for so long without noticing anything was wrong.