• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

GPU Temp: 68°C, Hotspot Temp: 100°C (Sapphire Pulse RX 6700XT)

I sent the graphics card to the technical service under warranty. I hope I get good results.
 
Good luck with the repair/swap. I recently sent my 4080 in for the same reason, lovely 41C delta and instant throttling + fan speeds pegged. Meanwhile my ancient 1070 is still running along happily at an 11C delta on its stock paste... Kinda feel like quality's gone out the window recently.
 
I was using Gainward GTX 1070 Phoenix GPU until 2 months ago. Even though the GPU is 6-7 years old and has never been opened, it did not exceed 68°C temperature and 75°C hotspot temperature even when I played for 10 hours under load. Manufacturers have neglected the quality of the hardware while giving importance to technology. Then we, the consumers, became testers. Not only in hardware, even games have optimization problems, but improvements are made after consumers' feedback.

They have to give us money for this. :laugh:
Good luck with the repair/swap. I recently sent my 4080 in for the same reason, lovely 41C delta and instant throttling + fan speeds pegged. Meanwhile my ancient 1070 is still running along happily at an 11C delta on its stock paste... Kinda feel like quality's gone out the window recently.
 
When undervolted in pure-CUDA, my 4070 ventus 2x has 12 delta between hotspot and gpu and gigabyte 4060 ti gaming has 13 delta. When no undervolt but 150MHz oc, both go ~15-20 delta. Imo the hotspot is power module / voltage regulator that does not have enough contact pressure on cooler block.

But in non-CUDA, i.e. gaming, both cards run below 80C and only 8-9 delta between hotspot and gpu.

Graphics cards are designed to work well under graphics tasks. Or its just my luck to have worse CUDA thermals on my non-golden chips.

When you limit the temperature from afterburner, it depends on your cooler & fan performance. Gigabyte 4060ti gaming is faster than 4070 ventus 2x when both are limited to 60C in a hot day. But in winter, 4070 always wins.
 
When undervolted in pure-CUDA, my 4070 ventus 2x has 12 delta between hotspot and gpu and gigabyte 4060 ti gaming has 13 delta. When no undervolt but 150MHz oc, both go ~15-20 delta. Imo the hotspot is power module / voltage regulator that does not have enough contact pressure on cooler block.

But in non-CUDA, i.e. gaming, both cards run below 80C and only 8-9 delta between hotspot and gpu.

Graphics cards are designed to work well under graphics tasks. Or its just my luck to have worse CUDA thermals on my non-golden chips.

When you limit the temperature from afterburner, it depends on your cooler & fan performance. Gigabyte 4060ti gaming is faster than 4070 ventus 2x when both are limited to 60C in a hot day. But in winter, 4070 always wins.
Hotspot temperatures on Nvidia's side are lower than AMD's. Generally, temperatures do not exceed 85-90°C. On AMD, this temperature is generally between 85-95 degrees, but the difference between normal temperature and hotspot temperature is at most 20 degrees, according to other sites and test videos. I usually see a difference of around 15°C. On my graphics card, there is a difference of 30-35°C.

On some sites it was interpreted as a heatsink assembly error, and on some sites it was interpreted as a silicone problem. Let's see what the technical service will answer. I hope the repair takes place.

Additionally, I also mentioned that the fan noise is quite loud at 2200 rpm. I also added the video to the description.

 
Nothing to worry there. These are my temps after Superposition 4K optimized test.

1698039849842.png



1698039907212.png


Powercolor 6700 XT Fighter with a slightly modified Morpheus II cooler
 
Check that upper pic. 62C on core, 89.5C on memory, 100C hotspot at maximum.

Though they are usually lower under typical gaming, I'll check and report later when I play something.
 
Check that upper pic. 62C on core, 89.5C on memory, 100C hotspot at maximum.

Though they are usually lower under typical gaming, I'll check and report later when I play something.
This temperature is different and it worries me in terms of GPU health. I want my GPU to last long. I requested maintenance under warranty. If not, I will do the maintenance myself after the warranty expires.
 
This temperature is different and it worries me in terms of GPU health. I want my GPU to last long. I requested maintenance under warranty. If not, I will do the maintenance myself after the warranty expires.
Personally I took a risk and put an aftermarket cooler on my card. That stock cooler sucked but that's what I got when bought the cheapest possible 6700 XT. :laugh:

The noise was untolerable (could be heard even with headphones on), now it's barely audible (2x bequiet Pure Wings 120mm on the cooler)

EDIT: Similar temps when gaming, running Life is Strange: True Colors with 4K very high RT on.
 
Last edited:
Personally I took a risk and put an aftermarket cooler on my card. That stock cooler sucked but that's what I got when bought the cheapest possible 6700 XT. :laugh:

The noise was untolerable (could be heard even with headphones on), now it's barely audible (2x bequiet Pure Wings 120mm on the cooler)

EDIT: Similar temps when gaming, running Life is Strange: True Colors with 4K very high RT on.
Hello, I sent the graphics card to warranty on October 20 due to the temperature problem and the unstable operation of the GPU fan. They gave me a new GPU yesterday, same model and brand.

I can say that the temperatures have returned to normal. There is a difference of around 15-20 degrees between the two temperatures right now. However, the instability in GPU fans did not improve. I don't think the problem is caused by the GPU anymore. Methods I have tried before to solve this:

1- I tried many drivers up to 23.8.2 using AMD Cleanup Utility or Display Driver Uninstaller, but it did not work.
2- When I wanted to control it via MSI Afterburner, the same problem continued.
3- I reset the motherboard's BIOS, removed and replaced the battery, cleared CMOS, and did CMOS Reset, but it still didn't work.
4- I installed Sapphire Trixx, but I don't see the fan menu there.
5- I turned off the Windows Fast Boot option, but the problem still persisted.

There are 2 options left: The first is to test the GPU on a different computer, and the second is to reinstall Windows.

I don't know what else I can do.
 
Hello, I sent the graphics card to warranty on October 20 due to the temperature problem and the unstable operation of the GPU fan. They gave me a new GPU yesterday, same model and brand.

I can say that the temperatures have returned to normal. There is a difference of around 15-20 degrees between the two temperatures right now. However, the instability in GPU fans did not improve. I don't think the problem is caused by the GPU anymore. Methods I have tried before to solve this:

1- I tried many drivers up to 23.8.2 using AMD Cleanup Utility or Display Driver Uninstaller, but it did not work.
2- When I wanted to control it via MSI Afterburner, the same problem continued.
3- I reset the motherboard's BIOS, removed and replaced the battery, cleared CMOS, and did CMOS Reset, but it still didn't work.
4- I installed Sapphire Trixx, but I don't see the fan menu there.
5- I turned off the Windows Fast Boot option, but the problem still persisted.

There are 2 options left: The first is to test the GPU on a different computer, and the second is to reinstall Windows.

I don't know what else I can do.
So a custom fan profile doesn't work? Even with the Radeon software?
 
So a custom fan profile doesn't work? Even with the Radeon software?
Exactly. So I did just about everything I could. I didn't do two things:

- Testing the GPU in a different case.
- Reinstalling Windows.

Could it be a problem with Windows? For example, when I install Sapphire Trixx, only 3 menus appear. I don't see the fan menu.

cf30a06dde5e10146139e1d9b11c69c3.png
 
I've tried Trixx last time back in HD 3850 days so haven't used it myself, but could multiple GPU tuning tools cause a conflict? I personally made a fan profile (and overclocked) using the AMD software, I use Afterburner only for OSD/monitoring.
 
I've tried Trixx last time back in HD 3850 days so haven't used it myself, but could multiple GPU tuning tools cause a conflict? I personally made a fan profile (and overclocked) using the AMD software, I use Afterburner only for OSD/monitoring.
I normally don't use any fan software. The computer had nothing but AMD Adrenalin, games and gaming platforms, personal hardware and software applications.

However, when I noticed that the fan was working erratically and made special settings in the Adrenalin program, I noticed that the fans were still spinning at a constant speed of 75%. Then I installed other software for trial purposes, but none of them worked.
 
I normally don't use any fan software. The computer had nothing but AMD Adrenalin, games and gaming platforms, personal hardware and software applications.

However, when I noticed that the fan was working erratically and made special settings in the Adrenalin program, I noticed that the fans were still spinning at a constant speed of 75%. Then I installed other software for trial purposes, but none of them worked.
 
Unfortunately, this was not a solution.

I updated the motherboard's BIOS yesterday. Additionally, I reinstalled Windows on the computer, but the fan speed is still unstable.

Let me show you an example with default settings:

1- Fan speed 1000 rpm when temp is 68°C
1698900221225.png


2- Fan speed 950 rpm when temp is 71°C
1698900239152.png


3- Fan speed 850 rpm when temp is 72°C
1698900261241.png


4- If I make a special fan curve using any program, this time the fans always spin at around 70-75% speed as soon as the temperature exceeds 50.

1698900469863.png



In this way, it works inversely proportional. When the temperature reaches around 80 degrees, the fans spin very fast and sometimes I experience a momentary drop in frequency.

What could be the source of this problem other than the GPU?
 
The fan speed is controlled by the GPU BIOS. If you know your exact card model and BIOS (I think there was a tool to get them in TPU), you could re-flash it and see if it's fixed.
 
I still don't think it was worth the trouble, is there any difference in clocks ?
 
Also does the "fan instability" prevail if you leave the load running for a while? Hysteresis (the difference in temperature in an amount of time required before the fan speed goes up) is a thing. If it eventually climbs up to an acceptable level that keeps temperatures in check, that's ok.
 
The fan speed is controlled by the GPU BIOS. If you know your exact card model and BIOS (I think there was a tool to get them in TPU), you could re-flash it and see if it's fixed.
My graphics card model is Sapphire Pulse RX 6700XT. But I have doubts about GPU Bios flash. They say it is a very risky transaction.

I still don't think it was worth the trouble, is there any difference in clocks ?
Also does the "fan instability" prevail if you leave the load running for a while? Hysteresis (the difference in temperature in an amount of time required before the fan speed goes up) is a thing. If it eventually climbs up to an acceptable level that keeps temperatures in check, that's ok.
Actually, I want to test this too. Is there a program that can play without any measurement program on the game screen and create statistics of the temperature and frequency values in this process after about 3-4 hours?

Additionally, we will test my GPU on my friend's computer today. I want to be absolutely sure whether the problem is with the GPU or not.

It would be great if I could find someone else using the Sapphire Pulse RX 6700XT to compare it to, but I don't. :(

I found a solution, guys.

In the custom fan setting, the fan percentage was not as I wanted. When I tried to reach the rpm values I wanted by manually adjusting them one by one, I came to the following conclusion:


WhatsApp Image 2023-11-03 at 15.58.23.jpeg



After adjusting this curve, I came to this conclusion in a 15-minute furmark test (1920x1080, 8xMSAA):

1699016564988.png


When you set it this way, it runs around 2000 RPM. Since the max fan speed of Sapphire Pulse 6700XT is 3500 RPM, I probably cannot get the exact values. I don't understand it exactly, but now I can adjust it as I want.

I cannot calculate it numerically, but I can say that I can adjust it as I wish, since I can determine how many rpm is spinning at which % value by changing the setting during the test.

Update:
The fan speed was entered incorrectly in the BIOS of this GPU. It should be written "-15" units less than normal fan speeds. So I learned how to calculate.

Thanks for everything guys. :)

@kapone32 @las @Beginner Micro Device @GerKNG @kurosagi01 @Vya Domus @Zyll Goliat @Bobaganoosh @oxrufiioxo @eidairaman1 @RJARRRPCGP @physicsjunky @Nike_486DX @Bernec @ARF @watzupken @Vayra86 @sLowEnd @tugrul_SIMD @XpanD @Kissamies @faye @Kodehawa
 
Last edited:
SOLVED!!!

I was having the same issue with my GPU temps and a 20-30C OR MORE hotspot delta on a Sapphire Pulse 6700 XT. Repasted and new thermal pads, used TG Kryonaut + TG minus pads with minor difference... saw everyone saying "by design/normal behavior" so I reluctantly excepted it. Over the next couple months kept watching the temps that screams bad IHS contact to me and I took one more attempt.

The answer for me was a 25x25mm Kryosheet + Arctic TP-3 thermal pads! 1mm for the mosfets and 1.5mm for the memory.

I also shaved about 0.1-0.2MM off the height of the 2 rubber spacers opposite of the memory modules between the IHS and the board.

Temps after are what I am used to 60-70C while gaming, quieter, 10 degree delta or less, no more 95-110C with 20-30+ delta!!!

Just wanted to share since no fix online anywhere. ENJOY!

links to products used
Kryosheet https://www.amazon.com/dp/B0C61NWNZ...3d6c4048dbb11&language=en_US&ref_=as_li_ss_tl
TP-3 1mm https://www.amazon.com/ARCTIC-TP-3-...b593993f31926&language=en_US&ref_=as_li_ss_tl
TP-3 1.5mm https://www.amazon.com/ARCTIC-TP-3-...3d44a68ea4fa3&language=en_US&ref_=as_li_ss_tl

before: 99% utilization = 90C+ hotspot, constant spikes up to 95-110C, 20-30C delta
after: 99% utilization = 70C hotspot, occasional spikes up to 85C, 10 degree delta or less, outstanding thermal recovery time!

in COD MW3 = 75C max hotspot, only 2-5 degree delta, 60C memory on a 1440p ultrawide monitor using a moderate overclock

Manual overclock used:
FREQ = 2500-2600MHz
Volts = 1.20v
MEM = 2112MHz, fast timings on memory = results in 2100MHz
Power = +15%
fan = 75%, zero rpm off



This is what worked for me, do at your own risk.
 
Last edited:
Undervolt it. I run my Powercolor Fighter with these values.

View attachment 328956
I had it at 1.05v, and I can't remember where but it was freezing coming out off a specific application or benchtest, I realize 1.2v is only a mild undervolt. Runs at 1.45 or something like that out of the box. I was not trying to Max it out, was only looking to up the performance a bit to hit a Target FPS for a specific game and get my temps under control. This is why I called it a moderate overclock. If I did not, I meant to.
 
Back
Top