The GPU isn't receiving sufficient energy.
The GPU isn't receiving sufficient energy.
I possess an outdated Rx 570 XFX GPU that seems to be reaching its limit. The issue began over a year ago when the machine would shut down whenever the GPU consumed excessive power, around 80 Watts. I disassembled it and cleared the dust from the VRMs, which resolved the problem until recently. Now, under load, the system restarts again. After thorough cleaning and inspection, no visible damage was found. Testing the GPU on another machine reproduced the same issue. This suggests the power supply or other components aren’t the cause. Temperatures remain stable, only reaching about 70°C. It’s likely the GPU is struggling to supply enough power. Any advice would be appreciated—what steps could help fix this? My specs are: CPU AMD Ryzen 5 2600, GPU Rx 570 XFX, motherboard Micro-Star B450M PRO-M2, PSU Antec 500W, RAM 2x8GB DDR4, SSD 240GB ADATA SU630.
It might be a simple issue with the older graphics card, though it's hard to say for sure.
Temperatures remained consistent across the GPU, only reaching close to 60°C before the system rebooted. At the crash moment, the voltage was 1V and the VRM at 11.4V. The GPU was drawing 90 Watts. It’s worth noting the failure happened almost immediately. Gradually increasing power and clock speed on the MSI Afterburner caused a brief blackout, preventing temperatures from rising further. This resembles overclocking beyond safe limits and receiving a 'no' response.
The two machines you tested the GPU on aren't a full picture of all computers worldwide. It's just a test to avoid jumping to conclusions. The GPU isn't working in those devices, which means it won't work for you either. Whether it breaks or not doesn't matter—what matters is that it does break. Time to upgrade then.
Assuming the other two machines are reliable, running the 570 in both and experiencing a crash similarly is strong evidence pointing to the GPU as the issue. The graphics card may be outdated, with possible poor SMD contact or internal wear. Right now you can verify if thermal pads are properly connecting to VRAM—GPUs sometimes hide VRAM temperatures. If not, consider baking or using a heat gun for a short period, though lasting use is unlikely. Older GPUs generally degrade over time; some models from that era may still work but with reduced performance. I’ve worked with XFX cards previously, and the ones I purchased a few years back were probably used for mining and failed quickly. This particular lineup (RX 470, 480, 570, 580) seems to have had reliability problems.