RTX 3050 idle power too high

Hello,

First some details about my setup. I have a RTX 3050 without anything connected via HDMI or other ports. I use this card only for computing. I run it on opensuse 15.6. and my computing jobs are running in a container. The kernel module is loaded with the parameter “NVreg_DynamicPowerManagement=0x02”

Now, I observed a strange behavior. After a reboot or a kernel module reload, my RTX 3050 graphic card stays at 3w idle power.

marvin:~ # nvidia-smi -q -d PERFORMANCE

==============NVSMI LOG==============

Timestamp                                 : Thu Oct 17 10:45:29 2024
Driver Version                            : 550.120
CUDA Version                              : 12.4

Attached GPUs                             : 1
GPU 00000000:01:00.0
    Performance State                     : P8
    Clocks Event Reasons
        Idle                              : Active
        Applications Clocks Setting       : Not Active
        SW Power Cap                      : Not Active
        HW Slowdown                       : Not Active
            HW Thermal Slowdown           : Not Active
            HW Power Brake Slowdown       : Not Active
        Sync Boost                        : Not Active
        SW Thermal Slowdown               : Not Active
        Display Clock Setting             : Not Active
    Sparse Operation Mode                 : N/A
marvin:~ # nvidia-smi -i 0 -q
...
    GPU Power Readings
        Power Draw                        : 3.27 W
        Current Power Limit               : 70.00 W
        Requested Power Limit             : 70.00 W
        Default Power Limit               : 70.00 W
        Min Power Limit                   : 20.00 W
        Max Power Limit                   : 70.00 W
...

after some computing jobs which consumes around 60W, it falls back to 7w, but not 3w.

marvin:~ # nvidia-smi -i 0 -q
...
    GPU Power Readings
        Power Draw                        : 7.79 W
        Current Power Limit               : 70.00 W
        Requested Power Limit             : 70.00 W
        Default Power Limit               : 70.00 W
        Min Power Limit                   : 20.00 W
        Max Power Limit                   : 70.00 W
...

Performance state falls back to P8 too.

Does anyone knows why this happens. If I unload all nvidia related kernel modules like “nvidia”, “nvidia_drm”, “nvidia_modeset”, “nvidia_uvm” and load it via modprobe again. It stays at 3w again.

My use case is that I need the computing power to analyze and group my photos in nextcloud once or twice per week. Only when new photos arrives my private cloud. The rest of the time I is idle.

I updatet to driver version 560.35.03 and still have the same problem

after a modprobe “nvidia” and running “nvidia_smi” I get “3W” Power Usage

then I run some computing jobs with a power usage between ~20W and 60W

after that the idle power usage is “7-8W”

then I unload “nvidia_drm”, “nvidia_modeset”, “nvidia_uvm” and “nvidia” followed by a modprobe “nvidia”

after that the idle power usage is “3W” again

btw. nvidia-persistanced is disabled and not running. That means there are no references to character device resources in background which prevents the gpu ressources from tearing down.

If I call

echo ‘suspend’ > /proc/driver/nvidia/suspend

followed by

echo ‘resume’ > /proc/driver/nvidia/suspend

it falls back to 3W too.

before someone asks, I will provide some more details about my setup.

Server: Headless running ASUS PRIME B760M-K with Intel 12400T
Graphiccard: Headless PNY GeForce RTX™ 3050 6GB Verto Dual Fan Graphics Card
OS: opensuse leap 15.6
Module Options: Only the ones provided by installation process in /etc/modprobe.d/50-nvidia-default.conf

I tried nvidia driver 550.xx.xx provided by opensuse and I tried also 560.35.03 provided by nvidia

Everytime I get the same result.

  • after a reboot 3W idle power consumption

  • running a computing app for some seconds (nextcloud recognize classify)

  • falls back to 7-8W idle power consumption
    grafik

  • unload all nvidia related kernel modules (rmmod nvidia_uvm, nvidia_drm, nvidia_modeset & nvidia)

  • modprobe nvidia

  • stays at 3W again
    grafik

also a ‘suspend’ followed by a ‘resume’ to /proc/driver/nvidia/suspend will ‘fix’ it

I played also with “options nvidia NVreg_DynamicPowerManagement=0x02” and removed “options nvidia-drm modeset=1 fbdev=1” and “NVreg_PreserveVideoMemoryAllocations=1” without any noticeable effect