My two-year-old Samsung 990 Pro 2TB causes the entire system to crash within a few minutes of launching any game. The game closes, Windows becomes unresponsive briefly, the display goes black, and the computer restarts without creating a dump file. CPU, GPU, and SSD temperatures remain normal.
The system uses a Ryzen 9 7950X3D, Noctua NH-D15, ASRock X670E Steel Legend, MSI RTX 5080, and 64GB of DDR5-6000 CL30 memory. Windows, motherboard drivers, and the SSD firmware are fully updated, and I have performed a fresh Windows installation.
A WD Black SN850X 4TB works perfectly in every M.2 slot, including while gaming, and the system remains stable with it. The 990 Pro crashes in every slot, even when it is the only installed drive. CrystalDiskMark does not trigger a crash, but actual gaming does. Samsung inspected the drive and returned it saying that it passed their tests.
CrystalDiskInfo reports unusual SMART values: the 990 Pro shows about 121,406 power-on counts over 15,970 hours, while an older 970 EVO shows only about 15,250 power-on counts over 30,836 hours. The system information also shows unexpected shutdowns, but I cannot find any minidump files. What else should I test, and does this point to a defective SSD despite Samsung's repair results?
4 Answers
Make sure the 990 Pro has the current firmware and that the motherboard chipset and storage-controller drivers are current, but you have already covered the usual software checks. If the drive still crashes the machine in every M.2 slot after that, Samsung’s test result is not consistent with your real-world failure. Ask for a replacement drive rather than continuing to troubleshoot a disk that may corrupt data or the operating system.
A benchmark passing does not prove the drive is healthy under every workload. Gaming can produce a different mix of small reads, queue activity, and power-state changes. Try a long mixed read/write or random-read stress test, but back up anything important first. If that reproduces the restart, document the SMART data and failure pattern for the warranty claim.
The missing dumps may mean this is a hard reset rather than a normal BSOD. Check that Windows is configured for automatic memory dumps, that the page file is enabled on the system drive, and that automatic restart is disabled temporarily. Also check Event Viewer for WHEA, disk, storahci, or controller errors around the restart time. However, those steps are mainly for confirmation; the fact that another SSD is completely stable makes a motherboard, memory, or temperature problem much less likely.
The power-on count is extremely suspicious. It could reflect repeated resets or power-cycle events while the drive is connected, especially if the SSD has been causing crashes for a long time. Since the SN850X is stable in every slot and the 990 Pro fails regardless of slot, the 990 Pro itself is the main suspect. I would stop using it for important data and request another replacement or refund from Samsung.
That makes sense. CrystalDiskMark only runs briefly and does not seem to reproduce what happens during real game loading, so I will test it more cautiously after backing up the drive.