My Lenovo Legion 5-15ARH05H has started crashing unpredictably during games over the past few months. It may blue-screen when launching a game, 20–30 minutes into a session, after several hours, or not at all. The most frequent problems occur in Apex Legends, although other games have also triggered crashes. The failures include BSODs, complete freezes requiring a forced shutdown, instant black-screen restarts, and occasional power-state behavior. I have seen several stop codes, especially IRQL_NOT_LESS_OR_EQUAL, along with KERNEL_SECURITY_CHECK_FAILURE, DRIVER_POWER_STATE_FAILURE, KMODE_EXCEPTION_NOT_HANDLED, INVALID_PUSH_LOCK_FLAGS, SYSTEM_SERVICE_EXCEPTION, and WORKER_THREAD_RETURNED_AT_BAD_IRQL. One Apex crash also reported DXGI_ERROR_DEVICE_HUNG. WinDbg has mentioned components such as the Intel wireless driver, Easy Anti-Cheat, and Riot's kernel driver, but no single driver appears consistently responsible. The system uses a Ryzen 7 4800H, GTX 1660 Ti, 16 GB RAM, Windows 11, and an SSD, usually with an external monitor and Ethernet connected. I have already tested the memory, CPU, GPU, storage, thermals, BIOS, chipset and graphics drivers, system files, clean boot configuration, networking devices, overlays, anti-cheat software, and power settings without finding an obvious problem. All 13 available mini-dumps reportedly terminated on logical processor 8, with many involving CPU idle, timer, scheduling, or context handling. Coreinfo shows that the processor has eight physical hyperthreaded cores. What other tests could help determine whether this is a faulty CPU core, motherboard or power issue, driver problem, or something specific to how certain games use the system?
1 Answer
The repeated pattern in the dumps is more interesting than the individual stop-code names. All 13 crashes ended on logical processor 8, and many were connected to CPU idle, timer, scheduler, or context-switch activity. That does not prove core 8 itself is defective, since Windows may simply schedule certain work there, but it is a useful lead. Conventional stress tests can miss this because they keep every core busy and may not exercise idle-to-load transitions. Run Coreinfo to confirm the logical-to-physical mapping, then try a variable-load test with OCCT or a similar tool. Pay particular attention to tests that repeatedly let cores idle and wake rather than only running a constant all-core benchmark. If crashes stop when the suspected logical processor is disabled or when processor idle states are limited, that would point toward a marginal CPU core, motherboard power delivery, firmware, or power-management issue rather than ordinary overheating.
Coreinfo reports eight physical processors, all hyperthreaded, so the system has 16 logical processors. I have not yet isolated or disabled the logical processor associated with the repeated crashes, but that seems like a worthwhile next test.