At 100% CPU Pc throws a black screen and a GPU debug, BUT under 100% GPU nothing happens.

olafperk159502dc

New member
Joined
Apr 2, 2026
Messages
10
Hey this issue is annoying as hell, and it keeps happening more and more recently.
My PC specs:
CPU: AMD Ryzen 7 3800x
Cooler: Alphacool Eisbaer Extreme 280mm
GPU: Palit RTX 3070 Gaming Pro
Motherboard: MSI MPG X570 GAMING PRO CARBON WIFI
RAM: G.Skill Trident Z Neo 32GB DDR4 3600mhz
OS: Windows 11 (who would have thought)
Storage: 1x 950 Samsung Pro 256GB, 2x Samsung 860 QVO 1TB
PSU: Be quiet! Pure Power 11 600W
Case: Asus ROG Helios

It started off seemingly randomly, then I managed to conclude that it only happens when my CPU hits 100% usage. Weirdly it when that happens it shows a GPU debug light. I tried testing the GPU only and it operated normally. Any ideas what might cause this weird thing?
I think it might be related to temps so I am planning on upgrading to the 'new' nzxt H5 and Liquid Freezer III Pro.
 
A 280mm AIO should have no problem dealing with the 3800X. The case is also not that bad, considering it has a glass front, but that glass front is protruding from the air intake, so there's enough room for decent airflow. It's weird that you don't have any other ASUS components for such a heavily ASUS-branded case that lends itself to an ASUS build, but hey. It seems to be a decent case. So something is not adding up.

From your description, we have to look at the sensor data, especially the temperatures. Don't worry about which EZ Debug LED is showing, that might be a false lead. The board goes into an abnormal state, and we need to look for the potential cause in a methodical manner. It could be that your AIO has developed a problem.

Please do the following: Open HWinfo Sensors and leave it running in the background, with the sensors expanded via <-> arrows on the bottom left. Let it sit for maybe five minutes like that. Then attempt to do a full run of Cinebench R23 Multi (takes ten minutes). In the end, take a screenshot of HWinfo Sensors with all the sensors visible at once, like this:

yes.png



Also note down the Cinebench result score. If the issue starts to happen before the benchmark can finish, then run the benchmark, but take a screenshot a few minutes in, or before the issue is likely about to happen.
 
Haha I decided on the case when I was younger and it just looked cool.

I appreciate your help but the moment i start cinabench or even the HWinfo inbuilt stress test the issue appears, and it always follows the same way. The reason I think it may be cooling, is that even on 5% cpu it sits around 45C and under 60% it reaches around 80C, it may be as simple as repasting but i fear the D5 inside my aio might be dying out after years of use.
 
A 280mm AIO should have no problem dealing with the 3800X. The case is also not that bad, considering it has a glass front, but that glass front is protruding from the air intake, so there's enough room for decent airflow. It's weird that you don't have any other ASUS components for such a heavily ASUS-branded case that lends itself to an ASUS build, but hey. It seems to be a decent case. So something is not adding up.

From your description, we have to look at the sensor data, especially the temperatures. Don't worry about which EZ Debug LED is showing, that might be a false lead. The board goes into an abnormal state, and we need to look for the potential cause in a methodical manner. It could be that your AIO has developed a problem.

Please do the following: Open HWinfo Sensors and leave it running in the background, with the sensors expanded via <-> arrows on the bottom left. Let it sit for maybe five minutes like that. Then attempt to do a full run of Cinebench R23 Multi (takes ten minutes). In the end, take a screenshot of HWinfo Sensors with all the sensors visible at once, like this:

yes.png



Also note down the Cinebench result score. If the issue starts to happen before the benchmark can finish, then run the benchmark, but take a screenshot a few minutes in, or before the issue is likely about to happen.
This is what happens after turning on cinabench. It also shows Event 41 Kernel-Power.

 
Ok, temperatures quickly climb to over 80°C, but there was no sign of thermal throttling in the sensors yet, so the issue appeared very suddenly. In general, the 3800X is one of the hottest-running CPUs of the 3000-series anyway. It's the model with the highest power draw to only have a single processor die, for which it can use roughly 105W (per TDP) or up to 142W, while the CPU models below that draw less power, and the CPU models above that can spread that same power draw over two processor dies. So it can run a bit hot, but I don't know if the temperatures should climb this fast.

What you can do, update the BIOS to the latest beta version from https://www.msi.com/Motherboard/MPG-X570-GAMING-PRO-CARBON-WIFI/support

Then, once the latest version is on, don't change anything CPU-related, only enable A-XMP for your RAM. Check if the issue still appears.
 
No problem, use the Flash BIOS Button to flash the BIOS properly:

1. Download the latest (beta) BIOS from the MSI support site for your board and extract it.
2. Take a USB stick and format it to FAT32 (it's best for it to have a size of 32 GB or less). You can also try Rufus for formatting, as FAT32 non-bootable.
3. Rename the extracted (!) BIOS file to MSI.ROM and save it to the root of your USB flash drive / USB stick (not in a subfolder).
4. Plug the USB flash drive / USB stick that contains the MSI.ROM file into the marked Flash BIOS Port (!) on the rear I/O panel.
5. With the PC off, press the Flash BIOS Button, and the LED should start flashing for about 5-8 minutes and then the board reboots.

Under a minute of flashing, or more than ten minutes, and it didn't work. After ten minutes it will never finish, you can just turn it off then. I would then advise to try with different USB sticks, the boards can sometimes be annoyingly picky in which one they will accept for flashing.

I checked before that your board model has this button, otherwise I wouldn't have suggested a BIOS update, as we can't fully ensure that things are well enough for that. But with the button method, other issues don't matter, it's just a minimal routine on the board that doesn't care about anything else that's wrong with the system. As long as the PSU is ok, but I saw nothing weird going on with the voltages in your video, and the PSU model is pretty decent.
 
No fin way its currently half way done with the cinabench, its crazy that all those issues were seemingly caused by the BIOS. I cancelled the case order buuuut i will upgrade the AIO just for peace of mind. Then again my bios version was from like 2021 XD.
 
Nice one, glad to hear. Yeah, sometimes a BIOS update can do wonders.

That Arctic AIO should be one of the best ones on the market. Not sure if it's entirely necessary here anymore, but it sure should cool well.
 
Aditional observation, IT SEEMS that issues only occur with game boost thing ON in bios. I will just keep it off for now and check with the new cooler? Maybe its VRMs kicking the bucket.
 
Back
Top