I have a 2021 LG Gram 17Z90P-G.AA78F with an Intel i7-1165G7, 16 GB RAM, and a 1 TB Samsung OEM NVMe SSD (MZVLB1T0HBLR-00000) running Windows 11. For about 10 weeks it has randomly crashed or restarted with several different bugchecks, including INACCESSIBLE_BOOT_DEVICE (0x7B), UNEXPECTED_STORE_EXCEPTION, CRITICAL_PROCESS_DIED, KMODE_EXCEPTION_NOT_HANDLED, and SYSTEM_SERVICE_EXCEPTION. Some crashes remain stuck at 0%, and Windows sometimes restarts without showing a blue screen.
SFC, DISM, CHKDSK, memory testing, Windows Update, a repair reinstall, and Reset this PC with a cloud download have not fixed it. The SSD reports healthy, has no recorded wear, and appears normally in Device Manager using the Microsoft Standard NVM Express Controller. A new Lightroom catalog also made no difference, and the laptop can crash shortly after reaching the desktop without Adobe software running. After repeated failures, Automatic Repair reported that it could not repair the installation.
Disabling hibernation/Fast Startup, PCIe ASPM, and disk idle power-down made the system stable for about an hour, but launching Lightroom soon triggered another INACCESSIBLE_BOOT_DEVICE crash. The SSD has reached a historical maximum temperature of about 84°C, although it is usually around 57°C. Safe Mode has been stable, while Lightroom is a reliable trigger under heavier storage or system activity.
Samsung Magician recognizes the OEM drive but cannot access its SMART data or run its diagnostic scan because the model is unsupported. Its performance benchmark completed successfully at roughly 3394 MB/s read and 1471 MB/s write. No minidumps are being created despite the pagefile and small-dump settings being configured.
The laptop has a second M.2 slot, so I am considering installing another NVMe drive and performing a completely clean Windows installation. Before replacing hardware, would you suspect the SSD, the M.2 connection or NVMe controller, BIOS, power or thermal management, the motherboard, or something else? What free diagnostics or controlled tests would you run next?
5 Answers
Samsung Magician is worth trying, but OEM Samsung drives are often only partially supported. A successful benchmark proves the drive can work under that particular load; it does not eliminate intermittent faults. I would also check the SSD firmware with an appropriate tool, inspect and reseat the module if practical, and test the second M.2 slot if the machine supports booting from it.
The cleanest software test is a true installation from a Windows USB installer: back up everything, delete the existing Windows partitions on the test drive, and install only the required drivers. If the same failures occur on a fresh installation, software and the existing catalog become much less likely. Installing a known-good NVMe drive in the second slot would help distinguish the SSD from the laptop's storage path, although a shared motherboard or power problem could still affect both slots.
The missing dump files are important. If Windows cannot write a dump and the crashes include INACCESSIBLE_BOOT_DEVICE, storage access may be failing during the crash itself. Check C:WindowsMinidump and the Event Viewer logs, but do not assume that a drive showing “Healthy” rules out an intermittent NVMe problem. SMART tools can miss controller, firmware, power-state, or connection failures.
I would avoid replacing parts blindly, but the two most useful next tests are a bootable diagnostic environment and a clean install on a known-good NVMe drive. If a Linux live USB or other independent environment also freezes or loses the drive under activity, suspect the SSD, slot, controller, or motherboard. If only the original Windows installation fails, focus on drivers and firmware instead. Keep verified backups before repeatedly testing a drive that may be dropping offline.
The fact that Safe Mode is stable and power-saving changes temporarily help points toward a driver, PCIe power-state, firmware, or hardware fault that appears under normal operating conditions. Reset BIOS settings to defaults, check whether LG has a newer BIOS directly rather than relying only on its updater, and test with PCIe power management disabled. Also monitor temperatures and reliability events, but do not treat the successful benchmark as a conclusive health test.
That makes sense. Since Windows Safe Mode is stable but normal Windows and Lightroom can trigger the failure, I will test with a live USB and a clean installation on a separate NVMe drive before deciding whether the original SSD or the motherboard needs replacement.