GPU Hot Spot Temperature 105 Celsius, fans at 3000 RPM, while GPU Temperature is max 70 Celsius

[Mitohondri]

A 7900XT packed in a "Glass box" really nice... that's a non-breathing case for such hw set...
And definitely to much delta values as previous droopyRO mentioned...

my native language is not english, but i am sure i was clear when i wrote, PLEASE READ WHAT I AM WRITING

what "glass box" and 7900 XT are you talking about?

My screen shot of GPU-Z clearly says RTX 3070!!!

the test is done on the desk, outside of the case, since i am trying to fix the machine and its easier to keep it out of the case

 

[droopyRO]

i am trying to fix the machine

Start by undervolting it an posting the results here. It should drop about 5-10ºC without performance loss if done right:

 

[meatwar]

my native language is not english, but i am sure i was clear when i wrote, PLEASE READ WHAT I AM WRITING

what "glass box" and 7900 XT are you talking about?

My screen shot of GPU-Z clearly says RTX 3070!!!

the test is done on the desk, outside of the case, since i am trying to fix the machine and its easier to keep it out of the case

Easy buddy... i understood correctly what you wrote.
And i not also a English native, same as a very high percentage of this forum users.

You had a 7900XT and now its a 3070, you faced the same issue...

The 'Glass box" is my comment to the "tiny" Lian Li O11D Mini... choice of pc case for your build.

And if the test are done in bench out of the PC case...even worst when assembled inside that case.
All the best, over_n_out.

 

[evernessince]

ok, good idea

here is my screen shot.

do you see anything that is off, or anything that can indicate what the problem is?

only thing that i notice is different from yours, is you have full green line, while mine is grey/green, is that something that shows that my PSU is faulty?

The grey just indicates idle. It means he was running the test long enough to where you can long longer see any grey on the chart.

What is a bit odd is that your GPU is running at base clock. Have you tried running a benchmark with both your systems and comparing to similar systems? That might give everyone a better picture as to what's going on.

 

[Vya Domus]

You are really stuck on this PSU thing, it's not possible for the PSU to cause high GPU temps, actually it's not possible for anything except a poorly mounted cooler to cause that, this is especially true in case of that 3070 which almost certainly doesn't have proper contact with the die.

Which model of a 7900XT exactly you have ? Even a reference model should not hit 105C at stock even in a poorly ventilated case, I would just RMA it.

 

[taka]

No, a 36ºC delta between the GPU temp and GPU hotspot is not good.

Salut !

I was talking about the input voltage and the VRM output not the temperatures. Yes that Delta is to big.

 

[Chomiq]

Show us a photo of 3070 mounted in your case. High hotspot temp is either a result of bad contact on the die due to GPU sag or bad paste job/assembly issues on the GPU itself.

 

[Mitohondri]

the 3070 is not in the computer case, and the problem is not in any of the GPUs

today i did the following, both 3070 and 7900 XT have been tested in 10 other computers, by now.

One at my home, two at my friends home, and 7 at my office, they are working fine in all of them.

Also i have tested 5 other GPU in my system, and all of them have same issue - 1050Ti, 1060, 1660, 470 and 580 that i have found around the office.

Its only in mine computer that the problem comes up.

If I am able to put both GPUs correctly in 10 other computers I am sure that I can put it in the same way in my personal.

its not te problem with MBO PCIe slot, since i tried in two motherboards, with 2 cpus, 2 memory kits, 2 ssd and 2 OS instalations

today i will take my system to service centre, and let them find what the hell is wrong

 

[droopyRO]

Also i have tested 5 other GPU in my system, and all of them have same issue - 1050Ti, 1060, 1660, 470 and 580 that i have found around the office.

No way 5+ GPUs have a 100ºC hotspot on your rig. Stop using GPUZ and use this instead, post a screenshot with HWinfo64 https://www.hwinfo.com/download/

 

[Mitohondri]

No way 5+ GPUs have a 100ºC hotspot on your rig. Stop using GPUZ and use this instead, post a screenshot with HWinfo64 https://www.hwinfo.com/download/

I am using HWinfo. I used GPU-Z only for the screenshot to compare with taka screen shot that he posted

System is at service center. We will se what they say

 

[Schnipey]

Hey all, just registered to let you know of a issue a reddit user found:

Reddit
Pin-20 problem

Maybe swap your DP-Cable?
This would also explain the problem occurring with different GPUs

 

[heraq]

Don't buy POWERCOLOR graphic cards! they don't have any sort of quality control
Here's my brand new 1 month old RX 7600 XT Hellhound edition GPU

the Hotspot sometimes reaches even 106C and is this with 2 intake fans and the side panel off
I made the mistake of buying trashcolor and now I'm stuck with this

 

[P4-630]

I made the mistake of buying trashcolor and now I'm stuck with this

Why didn't you return it earlier?

In my country I can send it back without reason within 30 days.

 

[Dragam1337]

Don't buy POWERCOLOR graphic cards! they don't have any sort of quality control
Here's my brand new 1 month old RX 7600 XT Hellhound edition GPU
View attachment 346462
the Hotspot sometimes reaches even 106C and is this with 2 intake fans and the side panel off
I made the mistake of buying trashcolor and now I'm stuck with this

Yeah, you deffo wanna repaste that card.

 

[heraq]

Why didn't you return it earlier?

In my country I can send it back without reason within 30 days.

I would honestly but I live in Europe and I ordered the GPU off Amazon US
and I don't want to deal with the return hassle and not being able to use my pc for who knows how long until they send it back

Yeah, you deffo wanna repaste that card.

Yeah I will, I already ordered some PTM7950 off aliexpress, hopefully it's legit and not some fake stuff

 

[Zach_01]

Have you consider maybe if there is any mount pressure issue?
For example trying to tighten screws a bit. But without too much effort…

 

[#22]

Big difference (~30C) between chip temperature and hot spot rather always is not evenly distributed paste. Cards have built-in protections with starting to run fans like crazy when hot spot hits some temp, usually 105C. There're two things to do in such situation: repaste yourself or RMA. If RMA, it's worth to check if card doesn't achieve declared boost clock and mention that.

 

[Dragam1337]

I would honestly but I live in Europe and I ordered the GPU off Amazon US
and I don't want to deal with the return hassle and not being able to use my pc for who knows how long until they send it back

Yeah I will, I already ordered some PTM7950 off aliexpress, hopefully it's legit and not some fake stuff

Yeah, thermal sheet would be a good longterm fix

 

[Zach_01]

To need repaste after one month? Could be, but I find the potential mounting issue more likely to be the case.
We've seen this before not long a go with a 7900XTX liquid cooled.

 

[Nubi]

So I had an issue dealing with this case as well and it started doing random power loops really fast and destroying components replaced every piece of hardware in this case more money than I had couldn’t fix it going through 3 motherboards just to find out the riser cable was the issue.

 

[eidairaman1]

So I had an issue dealing with this case as well and it started doing random power loops really fast and destroying components replaced every piece of hardware in this case more money than I had couldn’t fix it going through 3 motherboards just to find out the riser cable was the issue.

Cables are a hit and miss

 

[Noci]

Just some info to look at the hotspot temps from another perspective; eventhough my GPU is watercooled, when put to work it still has a delta-T of about 30 ~ 35 degrees Celsius between GPU & hotspot.
This is only under load only, when idle or low usage the delta-T is max 5 degrees Celsius.

Because of the fact that my GPU is watercooled, the temps are still well within an acceptable range, with aircooling it could be a little higher of course depending on the quality of the cooling device and the airflow.

Picture below, I tickled the settings in Adrenalin to the max to get some serious power usage on my RX 6700 XT (> 240W). These numbers are recorded with ambient temp 29 degrees Celsius and cooling liquid temp fluctuating around 33 degrees Celsius.

As soon as I stop the load, temps return to a few degrees over ambient within about 10 seconds and the delta-T between GPU and hotspot to about 5 degrees Celsius. So there is excellent heat transfer even for the hotspot.

So my point is, I know the contact of the sillicon with the cold plate is good in my case, but still the high delta-T occurs. This seems to have another reason.
For example I don't know where these sensors are physically located and how much thermal resistance there is between them induvidually and the cooling solution, but that might have some influence. For all I know, this may vary chip by chip as this occurs on a macro scale.

I'm not saying it can't be the TIM, but repasting is not always or the only solution as far as I'm concerned, you have to keep an open mind.
And yes, everything exceeding the 90 degrees Celsius threshold is a no go in my book, I prefer it much cooler.

To the OP, airflow at the front of the GPU alone is nice, but don't forget a proper airflow on the backside/backplate also. A lot of heat is emitting from there as well. In a small case that could quickly lead to undesired results due to the overcrowded cramp space in shuch a small volumetric area creating a plethora of restrictions. There may be less airflow on the desired places as you expect, despite the overall airflow should be sufficient.