3080 hotspot constantly 10C above GPU temp

[liberteus]

Whether idling or under load, the measured GPU hotspot is always 10C above. As I'm writing for instance, 26C for my GPU now, 36 for the hotspot, playing GPU was at 66, hotspot at 76.

I repasted myself the card before installing it in my build, I saved it from another build that I disassembled (that's what I had to do to get the card at the time...), and the OEM I bought it from basically didnt put any thermal paste!

 

Should I open the card again and repaste it? Or is a constant 10C difference somewhat normal?

 

Note: gpu mem junction sits just one or two degrees higher than the hotspot, and is not a concern to me 

[Origami Cactus]

Really normal, that is what hotspot means. It is the hottest part of the gpu die afterall, and no matter how much you repaste, some part will be hotter than the average, it is just how average temperature works.

[Kilrah]

10°C is just about right.

[YoussefShaala13]

10 Degrees Seems Ok

[xg32]

the cards ive had recently are all about +15C

[liberteus]

1 minute ago, Origami Cactus said:

Really normal, that is what hotspot means. It is the hottest part of the gpu die afterall, and no matter how much you repaste, some part will be hotter than the average, it is just how average temperature works.

Good, I have to say I really dont want to open the card a second time, I mean, these things aint cheap. Using a more aggressive fan curve I clamped down on temp also, and before repasting memory went as high as 98C, GPU itself 88, after repaste it went down a good 10C, and fan curve helped another 5/10C depending on how demanding the game is (and res. I use, switching 1080p and 3440*1440).

[liberteus]

Thanks all for your replies, sharing this kind of knowledge really is gold. (and I mean it)

[Falkentyne]

11 hours ago, liberteus said:

Thanks all for your replies, sharing this kind of knowledge really is gold. (and I mean it)

The absolute minimum hotspot delta is fused into hardware as a certain amount above core temp.  This depends on SKU, precisely.

On 3080 FE and 3090 FE it's exactly 10C.  On 3080 Ti FE, it's 7C.

On AIB cards, it can be as low as 6C, depending on exact card.

 

Usually you'll see the minimum delta when the card is at full idle, *with* the Nvidia power saving mode set at default (NOT at prefer max performance) and NOT with clocks/voltages locked with MSI Afterburner.

 

At full load, the delta will vary.  It's known that both the GPU Core AND the VRM's (NOT VRAM), report to the hotspot from various sensors, and the hotspot picks up the hottest reported temp among any of them.  On most cards, if the hotspot stays within 5C of the minimum delta for that SKU, then everything is fine.

 

A high hotspot delta (>15C) most often results from insufficient core to heatsink contact pressure or badly pasted parts of the core, especially on cards with thermal pad jobs done on them, where too thick thermal pads were used.  It's rare for a high hotspot delta to be caused by bad or too *thin* VRM pads (pads not making proper contact with the heatsink), but some people with stock cards have seen terrible hotspots, then took apart the card and found a completely ripped or missing thermal pad over the VRM's (mosfets), in which case it was the VRM causing the problem.

 

VRAM (memory) hotspots report to "memory junction temperature", not to "hotspot" temperature.