I’m having the same issue on my Dell G3 15 3500 with the 545 driver (both the open kernel module and the non-open one). Here is the bug report for the open one:
nvidia-bug-report.log.gz (1.7 MB)
Having the same issue, but it was working fine before, I think after moving to Linux 6.6 I started facing this issue
nvidia-bug-report.log.gz (469.4 KB)
Same issue on Arch Linux with 545.29.06 drivers on an Acer Nitro 5 laptop.
Can’t make the GPU go into D3cold mode, neither on X11 nor Wayland. It’s pretty much halving my battery life on the laptop. Not entirely sure on which version it has started, since I haven’t used my laptop that often off battery for a couple of updates.
nvidia-bug-report.log.gz (1.2 MB)
Same issue here, Razer Blade 15 2021 with rtx 3070
Looks like this is still an issue, I’m on Arch running the latest versions of everything as available Dec 2023
nvidia-bug-report.log.gz (490.8 KB)
DId some more digging, it appears that my device will enter D3cold about every 20 seconds but immediately return to D0. I tested this without my display service (sddm) running in a TTY environment only.
Additional troubleshooting steps I’ve taken can be found here:
https://bbs.archlinux.org/viewtopic.php?pid=2136464#p2136464
I dont know if it is relevant here but on my razer blade 14, 3070ti i have a similiar problem. I am running windows 11 though. GPU randomly powers on even though no processes use it and shortens battery life. I have used NVCleanstall to upgrade driver
Same exact issue as OP to the T. RZ09-0427 razer blade 2022 14 6900hx + 3070 ti. dgpu will only power down with 525xxx.xx versions no matter what desktop environment / distro / X11/wayland combination is used.
550.54.14 still not fixed, 525.147.05 still working.
edit: so figured out if you have the nvidia-powerd service enabled this won’t work so that also has to be disabled, also need SDDM running in wayland mode.
found a fix in opensuse tw if you guys wanna try, it also goes cold almost instantly now. ver 550.54.14
options nvidia-drm modeset=0 fbdev=0
options nvidia “NVreg_DynamicPowerManagement=0x03”
blacklist nouveau
blacklist nvidiafb
blacklist rivafb
1 Like
Lenovo Legion 5 Slim 16aph8 with 4070. Kubuntu 23.10 kernel 6.5, X11, 550.54.14,
The issue with d3cold also present. In my case when I sleep the laptop for a while and wake up, my dGPU goes to d3cold as expected and stays that way till some request from an app… The laptop supports only modern sleep [s2idle], no s3 support.
Managed to get this working on fedora 39 as well, keep in mind this is wayland only, i haven’t tested xorg but i assume it will not work and it seems like the latest drivers do break suspend too…
edit: fixed suspend too by setting NVreg_PreserveVideoMemoryAllocations to 0. Either way this should be fixed, it took me many months to figure this all out, maybe instead of working on wayland fix the basics first…
sudo grubby --args=“NVreg_PreserveVideoMemoryAllocations=0” --update-kernel=ALL
sudo grubby --update-kernel=ALL --args=“nvidia-drm.modeset=0” --update-kernel=ALL
sudo systemctl disable nvidia-powerd.service
rm -f /usr/share/glvnd/egl_vendor.d/10_nvidia.json
reboot
The above file 10_nvidia.json keeps it awake.
If on gnome (keep in mind this will get replaced with updates too) you need to edit /usr/lib/udev/rules.d/61-gdm.rules and comment out this line ATTR{parameters/modeset}!=“Y”, GOTO=“gdm_disable_wayland”
Other users, please keep in mind this is for the same laptop as OP, Razer Blade 14 2022 - RZ09-0427.
Nvidia, can we please get a fix for this? It’s quite annoying to choose between having 1/4th or having to power off each time we close the lid, since suspend isn’t working. I have attached a bug report of the laptop working on battery dgpu sleeping, driver ver 550.54.14.
To verify this is working, cat /proc/driver/nvidia/gpus/0000:01:00.0/power will show ? for everything until the dgpu is activated. You can also watch -n 1 cat /sys/class/drm/card*/device/power_state and unplug the power, it should show D3Cold.
As far as i’m aware nvidia-drm.modeset=0 goes against everything the official docs say to do.
nvidia-bug-report.log.gz (451.0 KB)
Hi @JayNope
Can you check that the vendor is indeed nvidia for the card0 (or card1 / 2) which is staying perpetually in D0 using below command.
cat /sys/class/drm/cardX/device/vendor
Hi amritis,
cat /sys/class/drm/card1/device/vendor shows 0x10de which is the card that stays in D0 with anything past 525.
apologies for the edit, got a tad confused.
I’m currently running into the exact same issue (Lenovo Legion 5 Pro, 4070) as described above, but I am using optimus-manager from the Arch Linux AUR. All drivers above 525 are unable to power down my GPU.
Hey, I’m just tried 555 drivers Lenovo Legion 5 2060 and still the same issue… at least 12W of power are being consumed instead of 8W when I use nouveau.
I thnk I’ll stick with open source nouveau for now. It doesn’t flicker on Wayland, which is also a huge bonus.
I have it! Legion5slim 16aph8 with 4070. The crucial option is to set NVreg_DynamicPowerManagementVideoMemoryThreshold=0. The driver 555.52.04, X11, my full working parameters are:
options nvidia NVreg_DynamicPowerManagementVideoMemoryThreshold=0
options nvidia “NVreg_DynamicPowerManagement=0x02”
options nvidia-drm modeset=1
blacklist nouveau
blacklist nvidiafb
blacklist rivafb
Can confirm this exact issue on Razer Blade 14 2021 (3080, latest bios). dGpu enters D3Cold on AC power but on battery power only does so for a fraction of a second before going back to D0 (about every 20 seconds or so). I’m on Fedora 40 and I’ve had this issue with all recent drivers. Driver version 470xx (from rpmfusion; maintained for older cards) worked as expected in this regard but that version is not ideal since it does not support Wayland. Switching to X11 does not help either on new drivers nor do any “solutions” I’ve been able to find, including the ones in this thread. I should mention Nouveau drivers work as expected.
It seems like this problem is not specific to Linux. I have the same issue on Windows 11. The workaround I’ve been using there is to restart the dGpu (disable and then enable after 5 seconds using pnp tools) with the task scheduler every time the charger is connected/disconnected. Again, older drivers work as expected on Windows (before about 525, I haven’t pinned down the exact version where this breaks).
One more. MSI GP66 11UG with RTX3070. This laptop has a LED light at the power button. When it uses only iGPU there is white color, when it uses dGPU it colored red. The LED is always red. But, if I set nvidia_drm.modeset=0 in kernel command line and I wait a little bit at the LUKS password prompt the LED turns white. But when load process reaches graphic target it becomes red again. The dGPU always in D0 state. Even when no one process uses it (according to nvidia-smi and fuser -v /dev/nvidia*).
kernel: 6.10.6-arch1-1
nvidia driver: 560.35.03
nvidia-bug-report.log.gz (1.4 MB)