Troubleshoot in layers: file integrity, Windows security, application startup, device detection, network connection, accepted work, and long-term stability.
Check the exact Windows Security event, file source, signature, required runtime, permissions, and log location. Re-download only from a verified source.
Practical note: Do not solve an unknown detection by disabling all protection.
Restart, verify Device Manager, return clocks to default, update to a supported driver, and confirm the selected engine supports the device.
Practical note: Test one device at a time when multiple GPUs are installed.
Confirm date and time, DNS, firewall rules, address format, algorithm, endpoint, and network stability. Read the server message in the log.
Practical note: Persistent rejection can indicate wrong parameters, excessive tuning, latency, or an incompatible job.
The worker may not be submitting accepted work, or the dashboard may not have updated yet. Check the log first.
Mining behavior can trigger detections, but every alert still requires source and file verification.
Power delivery, heat, drivers, memory instability, or aggressive tuning are common possibilities. Return to defaults and test systematically.