[SOLVED] Nvidia system wide crash GPU Reset Required

Al.Piotrowicz · 2026-06-21 21:20:17

Here is what happens after few days of uptime on the latest driver (610.43.02):

cze 21 22:54:05 testowy kernel: NVRM: GPU0 kgmmuInvalidateTlb_GM107: TLB invalidation failed waiting for completion (status=0x00000065) for PDB 0x10e63000, vaspaceFlags 0x84004020, scope 0x2, GFID 0
cze 21 22:54:05 testowy kernel: NVRM: GPU at PCI:0000:07:00: GPU-fb51809b-350a-99db-67b5-e16fc6d498ab
cze 21 22:54:05 testowy kernel: NVRM: Xid (PCI:0000:07:00): 62, 000121ea 00012246 00011d77 00015c6c 00016077 00014092 00000011 00000000
cze 21 22:54:05 testowy kernel: NVRM: GPU0 _kgspRpcGspEventPmuHalted: Received signal from GSP that PMU has halted.
cze 21 22:54:05 testowy kernel: NVRM: Xid (PCI:0000:07:00): 154, GPU recovery action changed from 0x0 (None) to 0x1 (PF FLR)
cze 21 22:54:09 testowy at-spi2-registryd[5043]: Disabling unresponsive app with pid 5112
cze 21 22:54:09 testowy at-spi2-registryd[5043]: Disabling unresponsive app with pid 1205682
cze 21 22:54:12 testowy accounts-daemon[5373]: Unable to open /etc/tcb: Nie ma takiego pliku ani katalogu
cze 21 22:54:12 testowy dbus-daemon[4133]: [system] Activating via systemd: service name='org.freedesktop.home1' unit='dbus-org.freedesktop.home1.service' requested by ':1.24' (uid=0 pid=5373 comm="/usr/lib/accounts-daemon")
cze 21 22:54:12 testowy dbus-daemon[4133]: [system] Activation via systemd failed for unit 'dbus-org.freedesktop.home1.service': Unit dbus-org.freedesktop.home1.service not found.
cze 21 22:54:12 testowy accounts-daemon[5373]: couldn't list homed users: GDBus.Error:org.freedesktop.systemd1.NoSuchUnit: Unit dbus-org.freedesktop.home1.service not found.
cze 21 22:54:35 testowy kernel: NVRM: kgmmuInvalidateTlb_GM107: TLB invalidation failed waiting for prior invalidate (status=0x00000065), vaspaceFlags 0x101, scope 0x2, GFID 0
cze 21 22:54:35 testowy kernel: NVRM: nvAssertFailedNoLog: Assertion failed: pEntries != NULL @ gmmu_walk.c:842
cze 21 22:54:35 testowy kernel: NVRM: nvAssertFailedNoLog: Assertion failed: progress == entryIndexHi - entryIndexLo + 1 @ mmu_walk_fill.c:130
cze 21 22:54:35 testowy kernel: NVRM: nvAssertFailedNoLog: Assertion failed: NV_OK == status @ mmu_walk.c:541
cze 21 22:54:35 testowy kernel: NVRM: mmuWalkUnmap: Failed to unmap VA Range 0x128800000 to 0x1289fffff. Status = 0x00000040
cze 21 22:54:35 testowy kernel: NVRM: nvAssertFailedNoLog: Assertion failed: 0 @ mmu_walk_unmap.c:62
cze 21 22:54:35 testowy kernel: NVRM: nvAssertOkFailedNoLog: Assertion failed: Generic Error: Invalid state [NV_ERR_INVALID_STATE] (0x00000040) returned from mmuWalkUnmap(userCtx.pGpuState->pWalk, vaLo, vaHi) @ gpu_vaspace.c:2080
cze 21 22:55:05 testowy kernel: NVRM: kgmmuInvalidateTlb_GM107: TLB invalidation failed waiting for prior invalidate (status=0x00000065), vaspaceFlags 0x101, scope 0x2, GFID 0
cze 21 22:55:05 testowy kernel: NVRM: nvAssertFailedNoLog: Assertion failed: pEntries != NULL @ gmmu_walk.c:842
cze 21 22:55:05 testowy kernel: NVRM: nvAssertFailedNoLog: Assertion failed: progress == entryIndexHi - entryIndexLo + 1 @ mmu_walk_fill.c:130
cze 21 22:55:05 testowy kernel: NVRM: nvAssertFailedNoLog: Assertion failed: NV_OK == status @ mmu_walk.c:541
cze 21 22:55:05 testowy kernel: NVRM: mmuWalkUnmap: Failed to unmap VA Range 0x128800000 to 0x1289fffff. Status = 0x00000040
cze 21 22:55:05 testowy kernel: NVRM: nvAssertFailedNoLog: Assertion failed: 0 @ mmu_walk_unmap.c:62
cze 21 22:55:05 testowy kernel: NVRM: nvAssertFailedNoLog: Assertion failed: NV_OK == status @ gpu_vaspace.c:4913

I tried to test the GPU by running cuda_memtest (flawless for few hours), so could it be dying GPU? Probably sounds related despite newer GPUs

Al.Piotrowicz · 2026-07-11 11:07:54

Problem sounds to be related into DOCP RAM overclock, seems the motherboard isn't fully compatible within these, which could manifest various errors. The other pair of boots is the 5950x itself which is also hard to tune on b-550 chip. Avoid these combination on linux.

EDIT: It tends to be MMU error which manifestes: "NVRM: GPU0 kgmmuInvalidateTlb_GM107: TLB invalidation failed waiting for completion (status=0x00000065) for PDB 0xXXXXXXXX, vaspaceFlags 0xXXXXXXXX, scope 0x2, GFID 0". These "X" are always the random hex values. Seems to be some glitch in the motherboard bios itself, because it hangs on manual bios settings (nothing overlcocked/overvolted/downvloted)

Last edited by Al.Piotrowicz (2026-07-25 06:47:22)

hwertz · 2026-10-28 23:33:49

Unfortunately, I can assure you this is not the case, it's not from overclocking. (I'm running Ubuntu, not Arch, but for this case I think that doesn't matter since we're talking rather the bug is drivers/software or hardware.)

I get the same errors on a GTX1650, i7-8700 system, and nothing is overclocked. And I've found complaints online with everything up through at least a 4090 getting just the same errors; often times suspecting hardware. I think I've had firefox playing a video popped out every time? It's terrible, at one point I had it crash like several times in a few days, then nothing for many weeks so I'm like "Good, it's fixed". Then it acts up again. I found Noveau (and NVK for Vulkan) would soft-lock (so I could restart my graphical session, rather than having GPU fully locked up), but although it's MUCH better than in years past it's still only got about 66% the performance of the Nvidia driver (and no CUDA support of course). If I get too many of these frikin' freezes I'll go for it, at least it's a viable option.

Last edited by hwertz (2026-10-28 23:37:03)

seth · 2026-10-29 07:21:16

I had it crash like several times in a few days, then nothing for many weeks so I'm like "Good, it's fixed". Then it acts up again.

Varying room temperature or 3rd channel power draw?

XID 62 is a hardware or thermal issue or every single nvidia driver version on every OS you've ever used had the same weird bug - what's unlikely and also apparently nouveau "soft-lock"s as well.
Overclocking is *one* way to run into this but certainly not the only one.

Arch Linux

#51 2026-06-21 21:20:17

Re: [SOLVED] Nvidia system wide crash GPU Reset Required

#52 2026-07-11 11:07:54

Re: [SOLVED] Nvidia system wide crash GPU Reset Required

#53 2026-10-28 23:33:49

Re: [SOLVED] Nvidia system wide crash GPU Reset Required

#54 2026-10-29 07:21:16

Re: [SOLVED] Nvidia system wide crash GPU Reset Required

Board footer