I bought a used gaming PC with an RTX 3060 Ti, 1TB of storage, and 16GB of RAM. It crashes during demanding games such as heavily modded Minecraft with shaders and Overwatch. I upgraded the power supply to 750W and increased the memory to 32GB, but the crashes continued.
The seller operated an internet café and said the graphics card was about two years old. I also noticed that one of its fans had been replaced and that the GPU made noticeable noise during benchmarking. Reducing the GPU power limit to 85% in MSI Afterburner and lowering game settings reduced the frequency of crashes.
The GPU reaches about 92°C at stock settings and around 82°C with the reduced power limit. Its drivers are up to date, and Windows came preinstalled. The motherboard is an A520MHP, and the memory is Kingston DDR4-2666. Should I have the system properly diagnosed, replace only the GPU, sell the PC, or simply continue using the lower power limit? I am concerned that a repair shop might recommend unnecessarily expensive parts.
4 Answers
Around 92°C on the GPU is quite hot, especially if the card is also making unusual fan noise. The reduced power limit lowering the temperature and improving stability strongly suggests a cooling or GPU-related problem, though it is not absolute proof. Check whether the fans ramp up correctly, clean the heatsink, improve case airflow, and consider having the card’s thermal paste and pads inspected. Avoid treating repeated crashes as normal.
The motherboard and memory configuration are worth checking too. Make sure the RAM is running at stable, supported settings rather than an automatically chosen timing that causes errors. Test one memory stick at a time with a memory diagnostic, reset BIOS settings to defaults, and verify that the CPU and GPU power cables are properly connected. A 750W power supply should generally be sufficient for an RTX 3060 Ti if it is a reliable model and is functioning correctly.
There is not enough information to confidently blame the GPU yet. Crashes could come from overheating, unstable memory settings, damaged hardware, drivers, or the preinstalled Windows installation. Start by recording the exact temperatures and crash behavior, then test the system methodically rather than replacing parts immediately. A clean Windows installation and a clean graphics-driver reinstall would also rule out a lot of software problems.
I would pay for a diagnostic before buying a replacement GPU. Ask the shop to test the graphics card, memory, temperatures, storage, and power supply separately and to give you the diagnosis and estimated cost before doing any work. If the GPU is confirmed defective and the rest of the system is healthy, replacing just the GPU makes more sense than selling the entire PC. Until then, using the lower power limit is acceptable as a temporary workaround, but it does not repair the underlying fault.
If the card still crashes after cleaning and confirming good airflow, test it in another known-good computer or run a dedicated GPU stress test while monitoring temperatures. That can help separate a faulty card from a problem elsewhere in the system.