I recently built a new Windows 11 PC that frequently crashes with a blue screen while playing VALORANT, shortly after closing the game, and occasionally during shutdown. It usually boots normally again afterward.
The crashes originally seemed to occur after two to four matches, but they also continued after uninstalling VALORANT, the Riot client, and Vanguard. I have already used DDU in Safe Mode, clean-installed the NVIDIA drivers, changed the power plan, disabled and re-enabled XMP, updated the motherboard BIOS from version 4.03 to 4.10, installed the AMD chipset drivers, and tested the memory with MemTest86 without errors. The RAM has also been replaced previously after the original module caused boot-loop CPU and DRAM diagnostic lights.
The system has a Ryzen 7 7800X3D, one 16 GB APACER PANTHER DDR5 module, an ASUS STRIX GTX 1080 Ti, an ASRock B650M X3D WiFi motherboard, and a 1 TB KIOXIA EXCERIA PRO SSD. The memory module was installed in slot B2 according to the motherboard instructions.
One dump reported stop code 0xBE, ATTEMPTED_WRITE_TO_READONLY_MEMORY. WinDbg showed nt!MiRaisedIrqlFault+2c3, with ntkrnlmp.exe listed as the image and no obvious third-party driver identified. What should I test next to determine whether this is caused by hardware, Windows, or a driver such as Vanguard?
4 Answers
Since the crashes also happen while shutting down and during unrelated games, I would broaden the investigation beyond VALORANT. Check Windows Event Viewer and Reliability Monitor for hardware or driver errors around each crash, and make sure the SSD firmware and all motherboard device drivers are current. It is also worth temporarily testing with only essential hardware connected and checking temperatures, power connections, and the graphics card. A clean Windows installation should be considered only after the hardware and firmware checks, not as the first step.
The WinDbg output showing nt!MiRaisedIrqlFault and ntkrnlmp.exe is not enough to identify a specific driver. Use the full !analyze -v output and inspect the call stack, MODULE_NAME, IMAGE_NAME, and failure bucket for every new dump. If the failures have different locations but similar memory-related symptoms, unstable hardware is more likely than one faulty application driver. If all dumps consistently implicate the same driver, update or remove that driver and retest.
The 0xBE stop code does not automatically mean the NVIDIA driver is responsible. It means kernel-level code attempted to write to read-only memory, and ntkrnlmp.exe is the Windows kernel where the crash was detected—not necessarily the original culprit. With MemTest86 passing, the next priority is still stability testing at completely default BIOS settings: leave XMP or EXPO disabled, load BIOS defaults, and verify the single DIMM is in the slot the exact board manual recommends. Also collect two or three additional dumps; if they all point to the same third-party module, that is much more useful than a single crash that only names the kernel.
Passing one MemTest86 run does not completely rule out platform instability, especially after the previous RAM failure. Test the system for several passes or overnight with BIOS defaults, and if possible try a known-good DDR5 module or test this module in another compatible system. If crashes continue at stock settings with different memory, the motherboard, CPU memory controller, GPU, power supply, or storage becomes more suspicious. Because the machine is new, repeated failures at default settings are a good reason to use the warranty or return process rather than spending weeks forcing it to work.
The BIOS is already on 4.10 and the chipset drivers are installed, so I would avoid changing random voltages or using driver-tuning tools. The fact that crashes continued after removing VALORANT and Vanguard makes a game-specific cause less likely.