Question 6700 XT and 6750 XT Hotspot Temps ? (Solved?)

[PC_Daddio]

Solution: Reverted power plan from High Performance to Balanced (default). I had been using High Performance exclusively since I bought the PC because I had done the same with previous PCs and laptops.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Hello, thanks for reading. I currently have an ASRock Challenger Pro 6750 XT GPU in my system. I also have an ASRock Challenger Pro 6700XT awaiting funds to be used in a PC for my son. Both have identical problems and have received identical treatment.

Spoiler: PC Components

CPU- Intel i7-11700F
GPU- AMD RX 6750XT
RAM- 16GB (8x2) DDR4 @ 3000 MHz
PSU- Corsair CX750M
MOBO- Gigabyte Z590 UD AC
Display- 2560x1080 @60Hz

I forget the specifics, and I don't think it matters, but I have a 7.1 channel Creative sound card and a Wi-Fi / Bluetooth PCIE card.

All drivers, firmware, and BIOS are up to date.

The following applies to both GPUs.
When I first got the GPU everything was fine. Having upgraded from a 1050 TI I was very happy with the performance. Temps were only concerning in demanding AAA titles but adjusting the settings took care of that. After around 6 months is when I began having problems with my GPU temps. I adjusted the fan curve to be more aggressive, which had a marginal impact. When that didn't work I applied a 5% under volt. That helped for a short period and now it's necessary to keep hot spot temps down, even in low demand games. I repasted with CryoFuze Violet , and that worked for a short time. I replaced the thermal pads on the 6750XT with 1.5mm as well

Regardless of what helped keep my hot spot temps down, the overall GPU temp has always been low. I don't keep a log of the exact numbers, but GPU temp is typically ~45C when the hot spot hits 110C and does a thermal shut down. This even happens with my GPU utilization below 20%! Having been trying to find an answer for this for a while this seems to be an issue for many others and all AIBs.

I came across a discussion on a forum that mentioned a possible cause that made sense. I haven't been able to explore this yet, but the user brought up that the mounting pressure of modern GPUs and CPUs combined with the difference in thermal expansion between the dye and heatsink causes thermal pastes to be squeezed out over time. They proposed using a PCM type of thermal pad like this. The reasoning was that the pad will liquify during use, but after the GPU cools the pad will re-solidify and contract as it does preventing the squeeze out seen by thermal pastes. I'm hoping to have $20 to spare next week to test this. I'll provide an update if I see immediate improvement or not, and hopefully remember to come back in a few months to follow up.

Currently I can only game in low demanding titles with settings set to low and capping FPS to 30. Even then I have to be mindful of areas that hit the GPU a little harder.
I'm open to suggestions and any questions. However, the following is a list of responses that won't be helpful because they would not be helpful or are things I have already tried.

Spoiler: Non-responses

Hotspot delta is normally 20C.
Yes, but not 60+. That's why I'm here.

Improper thermal paste application.
Had this not been an issue prior to applying new paste and pads there would be merit to this. That being said, I'm no stranger to repasting. The extreme temperature delta here would indicate potential improper pasting, normally. In this case it appears to be a symptom of another problem though.

Pads not correct thickness.
Just use Google. 1.5mm is the thickness needed. I also compared the new pad to the old where it was not indented before replacing to verify.

Case fans and GPU fan curve.
Case fans are set to 100% with a positive pressure bias. 3 240mm on the bottom and 1 240mm on the front as intake with a 240mm on top as exhaust. I remove the screen on the top of the case when I'm using the PC too, but that doesn't have a noticeable effect. The air being exhausted is a bit warmer than ambient, but not hot enough to suggest my case is being heat soaked. GPU fan curve is pretty aggressive already. I've tried setting the fans to 100% to see if that made a difference but it did not. Fan Curve

Clean my PC
I do, and it is.

RMA it.
Far to late for that. I don't think it's a deflective unit though since I have had the same issue with 2 new GPUs. Could be extremely unlucky I guess but that doesn't seem likely.

Just buy and Nvidia GPU
Not here for fan boy BS, nor do I have the ability to buy a new GPU.

 

[helpstar]

how old is your power supply?
would check with a different one.
how do you know it's thermal shutdown?

did you reset the bios by jumper yet?

 

[PC_Daddio]

how old is your power supply?
would check with a different one.
how do you know it's thermal shutdown?

did you reset the bios by jumper yet?

PSU is less than a year old. I got it at the same time I switched the 6700 for the 6750. Clearing the BIOS wouldn't do anything, but yes the BIOS has been reset a couple times in the last year. I know it's a thermal shutdown because if the hotspot hits 110C for more than a second the PC powers off.

 

[helpstar]

please take some pictures of your builds
overview and detailed, uoload to imgur.com and post the link

did you install any tweaking software like afterburner, asrocck tweak or similar, uninstall and reboot

does it have a dual bios switch onboard?

try a holder to be sure the gpu is not bent
be sure the cooler is not tightened too much to the gpu

how hot is your cpu getting?`which cooler is installed?

how do you test the gpu´s hot spot? furmark or just while gaming?

how much rpm are the fans spinning? are all spinning?

how hot is it in your room?

case fans installed?

check again the paste and thermal pads. all in the correct position? nothing missing?

 

[PC_Daddio]

please take some pictures of your builds
overview and detailed, uoload to imgur.com and post the link

did you install any tweaking software like afterburner, asrocck tweak or similar, uninstall and reboot

does it have a dual bios switch onboard?

try a holder to be sure the gpu is not bent
be sure the cooler is not tightened too much to the gpu

how hot is your cpu getting?`which cooler is installed?

how do you test the gpu´s hot spot? furmark or just while gaming?

how much rpm are the fans spinning? are all spinning?

how hot is it in your room?

case fans installed?

check again the paste and thermal pads. all in the correct position? nothing missing?

Turn standard orientation 90 degrees clockwise.

Only tweaking software I've ever used is Adrenaline.

No dual UEFI on this mobo

My current case has what would normally be rear IO, on top. So instead of the usual horizontal orientation it is vertical and supported by an extra stiff bracket. My old case was the standard orientation, but I did use a support bracket in that one.

My CPU hits 90C with an extended 100% load. Otherwise normal range is between 40 and 60 depending on use case. Cooler is stock.

I've done testing with Heaven, but now I just monitor temps with the overlay.

Fans are currently reporting 3000 RPMs. All 3 are functioning. My case's unusual orientation let's me see the GPU fans easily. I take a look at them whenever I see a spike in temps just in case they do fail. No way to know if there's any difference between their RPMs because they daisy chain to 1 header.

Currently winter with a basement office so ambient is 60F, give or take. Probably closer to 50F first thing in the morning.

Non-response

Non-response

 

[helpstar]

does the GPU have a dual bios switch onboard? like silent mode/performance/boost or whatever

try turning the case so the GPU is horizontal (fans downwards), open up the case to see if the airflow gets better or not

 

[PC_Daddio]

does the GPU have a dual bios switch onboard? like silent mode/performance/boost or whatever

try turning the case so the GPU is horizontal (fans downwards), open up the case to see if the airflow gets better or not

I realized I misread this the first time. Mostly the same answer though as the GPU does not have a mode switch either.

Mobo does not have dual UEFI, still. No switch or button to change modes. I double checked the specs page for your last comment.

This case is vertical, the last one was horizontal, issue persists regardless. The vertical orientation is more ideal for air flow as a horizontal GPU causes impedance when intake fans are mounted to the bottom of the case, as mine are. Removing the side panel results in a ~3C dip when the slightly cooler ambient air rushes in. It then goes right back up. Airflow is not the problem.

You asking about a mode switch made me think of something I haven't thought of yet. I just reinstalled Heaven to run some tests. I'll report back when that is finished.

 

[PC_Daddio]

@helpstar

The testing wasn't the most precise, but I think the results wouldn't change drastically given a more stringent series with more accurate logging.

What I did was change Windows power plan setting from High Performance, to Balanced. Your suggestion about a mode switch for my BIOS is what made me think about trying that because that switch would effectively do the same, as I understand it.

Since I hadn't set hotkeys to begin logging and benchmarking at the same time, my performance log began right before launching Heaven. Some settings in Heaven can't be made until you are in the application. Because of that I noted where I was when I started the High Performance benchmark, then waited to start the Balanced benchmark until I was at that same spot. This was to have equal levels of heat soak for each test.

I ran the first benchmark in High Performance mode after posting my reply to you. I kept my GPU settings the same (-5% under volt, aggressive fan cure). After completing that bench mark I noted the results from Heaven, changed to Balanced mode, then noted the results from the AMD performance log. I saved my notes then gave the PC a restart. After startup was complete I browsed these forums for a few minutes because that's what I was doing before the first test. I repeated the benchmark, then noted the results.

Below are the details

Spoiler: Heaven Settings

API- DirectX11
Quality- High
Tessellation- Normal
Anti-Aliasing- x4
Full Screen- Enabled
Resolution- 2560x1080 (System)
Ambient Occlusion- Enabled
Refractions- Enabled
Volumetric Shadows- Enabled
VSync- Enabled

Spoiler: High Performance

FPS:
59.2
Score:
1491
Min FPS:
8.7
Max FPS:
65.0

GPU Max 46
GPU Average 44
Hot Spot Max 111 (single tick in log)
Hot Spot Average 87

Spoiler: Balanced

FPS:
59.0
Score:
1487
Min FPS:
27.2
Max FPS:
60.7

GPU Max 41
GPU Average 37
Hot Spot Max 79 (Anomalous 101 at start of log / Heaven)
Hot Spot Average 68

Aside from the anomalous 101 peak at the start of the Balanced benchmark, both tests seemed to hit their peak temperature in scenes 18/19. Those scenes have intense dynamic lighting and shadows and the timestamp on the performance logs seemed to align with that point in the benchmark. My theory for the 101 peak is Heaven was performing pre-rendering / shader caching that was lost after restarting my PC.

I feel that these results show that Windows High Performance power plan is a major contributing factor to the problems I've been dealing with. The delta between GPU and Hot Spot is still rather wide though. I do intend to apply a PCM pad, but I most likely won't be able to for a while now due to unforeseen expenses.

While Balanced mode is scored slightly lower than High Performance mode, I think the real winner is Balanced mode. Scoring 4 points less is negligible, especially when High Performance was clearly throttling during the benchmark. That throttling was very visible watching the benchmark as well, and I value stability over higher average FPS.

A few points to make note of:
Heaven, and DirectX 11, do not demand as much from my GPU as being in game or a DirectX 12 title. While I suspect I will still see improvement across the board, it will almost certainly be different between games.

It is unknown how using the Balanced power plan is going to effect the performance of the rest of my PC. I rarely hit 100% utilization of my SATA SSD though. Outside of synthetic testing I have yet to see significant usage of my M.2 NVmE. If a noticeable difference is found I may run benchmarks for the rest of my PC's components.

I did try lowering the max power available in Adrenaline recently. I had to revert that back because even a 1% change caused my GPU to crash in games after a few moments.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

At this time I am inclined to consider the matter as solved. Should I experience any complications I'll report back with those. I truly do have every intention of applying a PCM pad and sharing the results. If someone happens to see this in a month (03/03/25) and I haven't provided the results of using a PCM pad feel free to ping me about it.

PS, I'm wondering if using the High Performance power plan interfered with the GPUs power management causing the drivers to allow too much amperage to be used. I'm honestly speaking from my rear-end here, but if the power management was created with the default Balanced power plan and a user switched to High Performance then the drivers could potentially call for X% of power without knowing X% is now significantly higher than expected. This is far outside my realm of knowledge so if I am way off base I would not be suprised.