Question GPU overheating issues ?

[mstojanovic]

Trying my luck here as well as I posted on other platforms and got no responses.

I’ve been experiencing a constant thermal issue with my ASUS ROG STRIX 4090 GAMING WHITE OC which I’ve had for around 2 years. The problem began around September/October last year. I noticed that during demanding games the fans would spin up loudly, much more than normal. I later discovered that one of the three fans wasn’t spinning at all because the card holder that prevents the card from sagging (the one that comes in the box) had shifted slightly, pressing against the outside of the fan. I fixed that by moving it back to proper position, and thankfully there was no fan damage occurred. But even after fixing that, the GPU temperatures remained high. Here's some data:

• Under heavy load (>60%), temperatures are consistently around 85–90°C, with hotspot hitting around 105°C.
• ⁠When idle or low load (<60%), idle temps are around 28–30°C and below 50°C during light usage.
• ⁠In games like CP2077 and Diablo IV (max settings), the temps stay at about 89–90°C for most of the gameplay (except in menus).

I tried several troubleshooting steps:

• ⁠Reverted drivers using DDU to versions that were stable when the issue started.
• ⁠Made sure that the airflow in my case is good and verified no rogue processes were stressing the GPU.
• ⁠Couple of days ago, I took the card to a service center. They repasted the GPU (the original paste was almost gone) and replaced thermal pads and they reported normal temperatures (around 70–75°C) during high-load tests on their bench (which uses a different setup).

When I returned the card and reinstalled it in my PC (with the latest drivers), the high temps (around 89–90°C) were still there. For fun, I even ran Heaven Benchmark, and it pushes it to 89°C almost immediately and throughout the test.

After further research, I ran into some posts on reddit about limiting the FPS in NVCP. I tried globally limiting the FPS in NVCP (setting it to –3 or more of my monitor’s refresh rate) and enabling VSync globally (but disabled in-game, as I use a 144Hz G-Sync monitor). This adjustment 'magically' brought down the temperatures in games I was testing to an average of 68–72°C at an 80–85% GPU load. I still ran into some additional issues where, in CP2077, I had to set the transformer model to “Balanced” instead of Auto/Quality to keep temps in check. But if I set the max FPS to 120, then the game works very well and no extreme temperatures.

So, in general, I'm now 'satisfied' with these workarounds - I get very good FPS at 1440p, while the GPU remains 'cool'. But still, I think the core issue is still unsolved. I've seen examples of people running higher framerates and GPU loads, with temps not going so wild. Has anyone experienced similar issues or have additional suggestions? Could there be something with my setup, or did maybe the repasting/thermal pad replacement not have fully addressed the issue?
For reference, here's the setup that was used:

My system configuration:

• OS: Windows 11 Pro 24H2 (Insider Preview build)
• ⁠CPU: Intel Core i9-13900K
• ⁠Motherboard: ASUS ROG MAXIMUS Z790 APEX
• ⁠PSU: Hydro PTM PRO ATX3.0 (PCIe 5.0) 1200W (using 12vhpwr connector for GPU)
• ⁠RAM: 2x G.Skill F5-6000J3040G32G (64GB total)
• ⁠GPU: ASUS ROG STRIX RTX 4090 GAMING WHITE OC (latest drivers)
• ⁠Storage: Kingston SKC3000S 1TB, Samsung SSD 990 PRO 2TB
• ⁠Case: Phanteks Evolv X
• ⁠Some pics of how card is installed: https://postimg.cc/gallery/dZRrwBF

Test bench in the shop (less precise since I got this from them via email):

• OS: Windows 10 Pro 22H2
• ⁠CPU: AMD Ryzen 7 5700X
• ⁠Motherboard: Gigabyte A520 K V2
• ⁠PSU: Antec 850W HCG850 Gold
• ⁠RAM: Kingston Fury Renegade DDR4 16GB/3200MHz
• ⁠SSD: Kingston 1TB NVMe
• ⁠(They also used the latest Nvidia drivers)

Thanks in advance!

 

[drivinfast247]

When you say "constant" do you mean since day one?

FYI: Your RAM is installed incorrectly. It should be a2-b2 or 2nd and last slot from CPU.

 

[mstojanovic]

When you say "constant" do you mean since day one?

FYI: Your RAM is installed incorrectly. It should be a2-b2 or 2nd and last slot from CPU.

When I say "constant", I mean since September/October 2024 when issue started happening. Before that I had no similar issues (since I bought it H1 2023).

Regarding the RAM modules - Z790 MAXIMUS APEX has 2 regular slots which I use (2x 32GB modules), and there's one DIMM.2 slot on the right, which I use for DIMM.2 card for the second M.2 SSD

 

[drivinfast247]

When I say "constant", I mean since September/October 2024 when issue started happening. Before that I had no similar issues (since I bought it H1 2023).

Regarding the RAM modules - Z790 MAXIMUS APEX has 2 regular slots which I use (2x 32GB modules), and there's one DIMM.2 slot on the right, which I use for DIMM.2 card for the second M.2 SSD

Okay. I understand that. That's cool!

What are you using to monitor temps?

For the sake of testing, try laying the PC on its side and check temps.

 

[mstojanovic]

I'm using HWiNFO to measure the temps.

I have tried what you suggested, and unfortunately no change. Here's a screenshot of HWiNFO window after around 5 minutes in Expedition 33 game

Additionally, here's a CSV from the test run: HWiNFO_test-run.CSV

 

[thestryker]

I'm guessing the case doesn't have sufficient airflow to cool the 450W from the GPU (when running maxed out) on top of the probably ~100W from the CPU.

Have you tried opening the side panel just to see what the temps look like?

This is a simple way to check and see if the GPU is getting enough cool air in.

 

[mstojanovic]

I'm guessing the case doesn't have sufficient airflow to cool the 450W from the GPU (when running maxed out) on top of the probably ~100W from the CPU.

Have you tried opening the side panel just to see what the temps look like?

This is a simple way to check and see if the GPU is getting enough cool air in.

Yes, I have tried that as well in several tests I performed, ever since first noticing the issue (apologies for not including that in the original post).

It makes no difference in tests sadly 🙁

Here's the screenshot of HWiNFO from the latest test with side panel open:

 

[drivinfast247]

I'd do another repaste of GPU.

 

[thestryker]

Yes, I have tried that as well in several tests I performed, ever since first noticing the issue (apologies for not including that in the original post).

It makes no difference in tests sadly 🙁

Here's the screenshot of HWiNFO from the latest test with side panel open:

Something is extremely wrong then because while everything looks normal voltage/memory clock/memory temp wise the GPU isn't clocking anywhere near as high as it should. Typically this would be the time I'd say repaste, but the shop already did that and this has been an ongoing issue. The only fixable thing I can really think of is if the thermal pad thickness is too high on the card this could occur.

Honestly this is something I'd have probably RMA'd a card this expensive for, but I'm guessing that likely isn't an option now. The best TIM setup you can use on a video card from my experience is using a PTM7950 pad on the GPU then a quality putty anywhere that used thermal. pads.