Most of these faults trace back to software or heat, not a dead component. Below is an ordered, South African gaming checklist for the symptom in the heading, written so each step is something you can verify yourself before booking a repair or buying a replacement.
Quick Answer
Most of these faults trace back to software or heat, not a dead component. This usually comes down to heat, a hot NVMe drive or background load rather than a faulty Ryzen 7 9800X3D. Run a fixed 20-minute test after each change and keep the one that holds at a steady frame rate or a stable temperature under 85C.
Temperatures under load
Confirm the case has a clear front-to-back path and that no cable is resting against a 120mm or 140mm fan. Reseat the cooler and apply a fresh pea of paste if temps spiked after a recent strip-down.
Drive health and thermals
Check SMART data for a failing 1TB or 2TB SSD that is silently retrying reads. Check SMART data for a failing 1TB or 2TB SSD that is silently retrying reads.
Software and firmware checks
Check that a recent Windows feature update did not swap your tuned driver for a generic Microsoft one overnight. Roll the GPU driver back to the last stable release, then do a clean reinstall using DDU so leftover files do not clash.
PSU and connector checks
Re-seat the 12V-2x6 and 8-pin EPS connectors fully; a half-clicked plug shows up as random shutdowns under load. Re-seat the 12V-2x6 and 8-pin EPS connectors fully; a half-clicked plug shows up as random shutdowns under load.
FAQ
How long should each test run?
Use a fixed 20-minute load that reliably triggers the fault. A shorter burst can hide a thermal or memory problem that only appears once the system is fully warmed up at 80C or higher.
Should I reset my overclock or EXPO profile?
Yes, as a test. Turn off EXPO and any GPU overclock, run the same load, and see if the fault disappears, since an unstable 6000MT/s kit is a common hidden cause.
Does heat explain random crashes?
Often. A CPU sitting at 95C or a GPU hotspot past 100C will throttle or shut down to protect itself, so log temperatures across a full session rather than reading one quick number.
Pro Tip
one variable per test: adjust a setting, re-run the same 20-minute load, then decide.