| Field | Details |
|---|---|
| Container Version | NVDisplay.Container.exe 1.39.3323.1171 (compiled 2023-08-25) |
| Exception Code | 0xC0000409 (STATUS_STACK_BUFFER_OVERRUN) |
| Fault Offset | 0x00000000000932E5 (consistent across all crashes) |
| Crash Timing | ~181 seconds after service start (runtime crash, not initialization) |
| Hardware | 6× NVIDIA RTX 5880 Ada Generation (all WDDM mode) |
| Motherboard | Supermicro X12DPG-OA6, Dual Intel Xeon Gold 6354, 1TB RAM |
| OS | Windows 11 Enterprise 24H2 (Build 26100.1) |
| GPU Bus IDs | PCI 4F:00.0, 52:00.0, 57:00.0, CE:00.0, D1:00.0, D6:00.0 |
| Impact | 4 of 6 GPUs show “Error” status in Device Manager / WMI after crash; nvidia-smi and CUDA remain fully functional; status self-recovers in ~30s when service restarts |
| Reproduction | 100% reproducible on every boot; crash always at same fault offset |
| Related Bug | Forum Topic #360403 — “Handle leak in NVDisplay.Container.exe version 1.39.3323.1171” (same Container version, reports 47,905 leaked handles) |
After every system boot, the NVIDIA Display Container LS service (NVDisplay.Container.exe, version 1.39.3323.1171) crashes with exception code 0xC0000409 (STATUS_STACK_BUFFER_OVERRUN) at a fixed fault offset of 0x932E5, approximately 181 seconds after the service starts.
The crash causes 4 of 6 RTX 5880 Ada GPUs to temporarily report “Error” status via WMI/PnP (Win32_VideoController), while nvidia-smi continues to report all 6 GPUs as healthy throughout the event. The PnP error status is transient (~30 seconds) and self-recovers when the service restarts via its FAILURE_ACTIONS configuration.
This suggests the crash only affects the WDDM user-mode driver stack connection managed by the Container service, not the kernel-mode driver (nvlddmkm.sys) or the NVML/CUDA path.
The consistent fault offset (0x932E5) across all crashes indicates a deterministic code path failure, potentially related to the handle leak documented in Forum Topic #360403 for the same Container version (1.39.3323.1171), where accumulated unreleased handles eventually corrupt the stack.
Expected Behavior: NVDisplay.Container.exe should run stably without crashing, and all 6 GPUs should maintain consistent “OK” status in both nvidia-smi and WMI/Device Manager.
Workaround: Disabling the NVIDIA Display Container LS service entirely has no impact on CUDA/NVML compute workloads, as these use the kernel driver path directly.