Power issues with dual NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition

[shahizat]

NVIDIA Open GPU Kernel Modules Version

NVRM version: NVIDIA UNIX Open Kernel Module for x86_64 580.95.05 Release Build (dvs-builder@U22-I3-B17-02-5) Tue Sep 23 09:55:41 UTC 2025

Please confirm this issue does not happen with the proprietary driver (of the same version). This issue tracker is only for bugs specific to the open kernel driver.

• I confirm that this does not happen with the proprietary driver package.

Operating System and Version

Ubuntu 24.04.3 LTS

Kernel Release

6.14.0-37-generic

Please confirm you are running a stable release kernel (e.g. not a -rc). We do not accept bug reports for unreleased kernels.

• I am running on a stable kernel release.

Hardware: GPU

GPU 0: NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition (UUID: GPU-ef5135f3-1177-e4c8-dd47-3818ddbe9182) GPU 1: NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition (UUID: GPU-9d278fa4-f945-cee7-9cb4-2626f3fc5

Describe the bug

Hello,

We've encountered an issue when running LLMs using inference frameworks like vLLM or Sglang in a multi GPU configuration. When I attempt to shut down the machine, either via sudo shutdown now or the desktop UI, it occasionally reboots instead of powering off. After it reboots once, I am usually able to shut it down normally. The issue is non-deterministic. It sometimes shuts down correctly, but other times it triggers a restart. We tested on the four machines with below configuration. The same issue on all machines. Please help to fix it.

• Motherboard: Gibabyte TRX50 AI TOP
• CPU: AMD Ryzen Threadripper 9960X 24-Cores
• GPU: 2xNVIDIA RTX PRO 6000 Blackwell Max-Q
• PSU: FSP2500-57APB

+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 580.95.05              Driver Version: 580.95.05      CUDA Version: 13.0     |
+-----------------------------------------+------------------------+----------------------+
| GPU  Name                 Persistence-M | Bus-Id          Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf          Pwr:Usage/Cap |           Memory-Usage | GPU-Util  Compute M. |
|                                         |                        |               MIG M. |
|=========================================+========================+======================|
|   0  NVIDIA RTX PRO 6000 Blac...    Off |   00000000:21:00.0 Off |                  Off |
| 30%   33C    P8              5W /  300W |     276MiB /  97887MiB |      0%      Default |
|                                         |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+
|   1  NVIDIA RTX PRO 6000 Blac...    Off |   00000000:C1:00.0 Off |                  Off |
| 30%   34C    P8             15W /  300W |      15MiB /  97887MiB |      0%      Default |
|                                         |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+

+-----------------------------------------------------------------------------------------+
| Processes:                                                                              |
|  GPU   GI   CI              PID   Type   Process name                        GPU Memory |
|        ID   ID                                                               Usage      |
|=========================================================================================|
|    0   N/A  N/A            2126      G   /usr/lib/xorg/Xorg                      118MiB |
|    0   N/A  N/A            2276      G   /usr/bin/gnome-shell                     24MiB |
|    1   N/A  N/A            2126      G   /usr/lib/xorg/Xorg                        4MiB |


cat /proc/driver/nvidia/params | grep DynamicPowerManagement
DynamicPowerManagement: 3
DynamicPowerManagementVideoMemoryThreshold: 200



cat /proc/driver/nvidia/gpus/0000\:21\:00.0/power
Runtime D3 status:          Disabled by default
Video Memory:               Active

GPU Hardware Support:
 Video Memory Self Refresh: Not Supported
 Video Memory Off:          Supported

S0ix Power Management:
 Platform Support:          Not Supported
 Status:                    Disabled

Notebook Dynamic Boost:     Not Supported



cat /proc/driver/nvidia/gpus/0000\:c1\:00.0/power
Runtime D3 status:          Disabled by default
Video Memory:               Active

GPU Hardware Support:
 Video Memory Self Refresh: Not Supported
 Video Memory Off:          Supported

S0ix Power Management:
 Platform Support:          Not Supported
 Status:                    Disabled

Notebook Dynamic Boost:     Not Supported

To Reproduce

1. vllm serve --model Qwen/Qwen3-VL-30B-A3B-Instruct --tensor-parallel-size 2 --gpu-memory-utilization 0.9
2. sudo shutdown now
3. It restarts instead of shutting down

Bug Incidence

Always

nvidia-bug-report.log.gz

Running nvidia-bug-report.sh... complete.


Summary of Skipped Sections:

Skipped Component                   | Details
================================================================================
ldd output                          | glxinfo not found
--------------------------------------------------------------------------------
vulkaninfo output                   | vulkaninfo not found
--------------------------------------------------------------------------------
ibstat output                       | ibstat not found
--------------------------------------------------------------------------------
acpidump output                     | acpidump not found
--------------------------------------------------------------------------------
mst output                          | mst not found
--------------------------------------------------------------------------------
nvlsm-bug-report.sh output          | nvlsm-bug-report.sh not found
--------------------------------------------------------------------------------

Summary of Errors:

Error Component                     | Details                                                      | Resolution
=========================================================================================================================

More Info

No response

[vjureta]

Filed #1111 — appears to be a silent-hang variant of the same GB202 GSP firmware death class you're seeing, but under sustained LLM inference (llama.cpp + Qwen3-32B) rather than Vulkan/Proton. No Xid raised at all in my case, which is the distinguishing signature.