Hi @con1,
Please provide details of your system configuration, a NVIDIA bug report and the list of games that show low FPS. We will file a bug and check if we can reproduce the issue on our systems. Thank you
Hi @con1,
Please provide details of your system configuration, a NVIDIA bug report and the list of games that show low FPS. We will file a bug and check if we can reproduce the issue on our systems. Thank you
Thank you.
Kubuntu 24.04
Intel(R) Core™ i9-14900K
MSI PRO Z790-P DDR4 motherboard Bios version MS-7E06
128GB DDR4 Ram
ASUS GeForce RTX 5070 Prime OC GDDR7 12GB
4k@120Hz or 60Hz monitor resolution via either HDMI or DP
Kernel: 6.13.5 (but tested multiple different kernels, including distribution kernel with same result)
Game: Path of Exile 2 on steam running directx 12, Vulkan, or directx 11. Resolution does not affect performance - same performance at 1280x720 windowed and 4k windowed fullscreen ~6fps. Changing setting to all the lowest settings shows the same performance. Nvidia system monitor shows 100% GPU utilisation but power consumption low at 50W. Path of exile shows a baseline 120ms delay from GPU at all settings. A 4070 GPU with the same settings and driver version 550 was averaging 120fps. This is a 95% performance loss.
Diablo 4 running via battle.net on proton gets about 14fps where the 4070 was getting 60.
Xorg or Wayland any desktop environment (KDE, XFCE, Ubuntu)
bug report attached as requested.
nvidia-bug-report.log.gz (575.5 KB)
VRAM issues???
Can you please stop making oneliners with no reference, it helps no one.
I can only guess here but are you referring to the “Excessive VRAM usage” or the “VRAM allocation/crashes/shared VRAM” issue?
The “excessive VRAM usage” in apps/Xwayland have a workaround for now and the devs know about it.
The “VRAM allocation/crashes/shared VRAM” issue does not have a tracking ID as far as I know and we have no real response from Nvidia regarding this.
Thank you for taking Action i hope it will get solved by the next driver release and keep us updated this is a huge issue for all of us
Hi,
The issue mostly happens at any dx12 / vkd3d games. This results into around 20-30% less performance and is known issue since years.
This happend across all my GPUs:
Any recent released game the issue can be reproduced.
Edit:
Here is a full thread with reports and benchmarks:
Really glad to hear the performance drop is being tracked. I’ll make recordings this weekend to track the PRESENT_WAIT issues causing crashes with Wine-Wayland & gamescope.
Thank you for taking the time to work on this! I know you have other priorities, and fixing issues like this goes a long way in making our computers more usable.
@RyzenDew And thank you Matt for raising awareness.
My data, FWIW.
AMD7600X GPU
Nvidia 4070 Ti Super GPU
64 GB RAM
Starfield using DLSS Balanced, 4k, frame generation on
Windows 11 24H2, 572 drivers, Starfield in Akila City averages 100 FPS
Linux Mint 22.1, 570 drivers, Starfield in Akila City averages 63 FPS
This is using the same mod loadout and same character save.
I appreciate any attention this issue receives. Thank you.
My data.
7950X3d
RTX 4090 XLR8
32GBs RAM
Black Myth Full Max RT and everything with dlss and framegen on windows i get about 160 fps, and linux i get 100… Probably the worst performance ive seen in a dx12 title.
I’ve noticed worse performance in Black Myth Wukong and Alan Wake 2, especially when ray tracing is enabled, compared to Windows.
Distro: Arch Linux
Kernel: 6.13.6.zen1-1
GPU Driver: NVIDIA-open 570.124.04-1
GPU: NVIDIA GeForce GTX 3080
CPU: AMD Ryzen 9 5950X
Edited to add bug report.
nvidia-bug-report.log.gz (420.1 KB)
Any update on the CTM freeze bug? I’ve mentioned this multiple times now with no response from the NVidia team, n ot even a bug has been filed: 565 release feedback & discussion - #370 by faz
That post contains all info including the nvidia-bug-report and repro steps.
Hi all,
Please add any new details for the DirectX12 performance issues on Linux to the original thread for tracking - DirectX12 performance is terrible on Linux - #104 by vesnine
@faz, Sorry I will check if we have one filed or file an internal bug to track the CTM freeze bug. Thanks
It’s extremely easy to figure out what’s being referenced.
The “excessive VRAM usage” in apps/Xwayland have a workaround for now and the devs know about it.
No, that workaround does not work. There is VRAM being consumed by nothing.
I have made a custom app profile for the apps I use, all I can say it helps keeping VRAM usage down for those apps.
Really?
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 570.124.04 Driver Version: 570.124.04 CUDA Version: 12.8 |
|-----------------------------------------+------------------------+----------------------+
| 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 GeForce RTX 4060 Off | 00000000:0C:00.0 On | N/A |
| 0% 45C P3 N/A / 115W | 2455MiB / 8188MiB | 1% Default |
| | | N/A |
+-----------------------------------------+------------------------+----------------------+
+-----------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=========================================================================================|
| 0 N/A N/A 1961 G /usr/bin/gnome-shell 606MiB |
| 0 N/A N/A 2474 G /usr/bin/Xwayland 3MiB |
| 0 N/A N/A 2543 G ...per --variations-seed-version 616MiB |
| 0 N/A N/A 2853 C+G /usr/bin/gnome-system-monitor 134MiB |
| 0 N/A N/A 3063 C+G /usr/bin/nautilus 99MiB |
| 0 N/A N/A 17191 C+G /usr/bin/gnome-control-center 134MiB |
| 0 N/A N/A 124887 G /usr/bin/cpu-x 2MiB |
| 0 N/A N/A 182959 G ...share/Steam/ubuntu12_32/steam 2MiB |
| 0 N/A N/A 183170 G ./steamwebhelper 101MiB |
| 0 N/A N/A 183196 C+G .../share/Steam/logs/cef_log.txt 5MiB |
| 0 N/A N/A 267441 C+G /usr/bin/gnome-calculator 54MiB |
| 0 N/A N/A 388277 G ./java 102MiB |
+-----------------------------------------------------------------------------------------+
Please inform me which XWayland process is eating all my VRAM because I’m not seeing it. Adding all those values is "only"1858MB. Over half a GB is nowhere to be found.
AFAIK all chromium based browsers are now fully Wayland.
Tried again with kernel-6.14.0-0.rc6.49.fc42.x86_64
I was able to get nvidia driver modules to load. Black screen. Locked up solid. No keyboard lights.
Setting nvidia-modeset=0 in kernel boot at least didn’t lock up, permitted login via ssh.
nvidia-bug-report.log.gz (334.5 KB)
And again, another log file, after installing nvidia-smi, clearing the logs, and rebooting to text mode with nvidia_drm loaded: I think I narrowed it down to xorg-x11-drv-nvidia-3:570.124.04-1.fc42.x86_64, which contains the firmware, nvidia-fallback driver service, udev rules, profiles, and quite a lot of documentation. So I went to reinstall [xorg-x11-drv-nvidia-570.124.04-1.fc42.x86_64 from rpmfusion]. But, it is missing from the rpmfusion Fedora 42 repos! That confirms my suspicions that something is wrong with it, and I can assume it’s being worked-on. I noticed xorg-x11-drv-nvidia-xorg-libs-570.124.04-1.fc42.x86_64 is not simlinking the libs. libglxserver_nvidia.so.0 should be a link to libglxserver_nvidia.so.570.124.04 for example, but it isn’t. I will try again when we get a new set of xorg-x11-drv-nvidia packages to test.
nvidia-bug-report.log.gz (660.1 KB)
It’s not really limited to Xwayland.
It may still increase VRAM for Wayland applications too, that’s why I added chromiium, firefox, heroic etc, to my custom application profile.
Try nvtop instead of nvidia-smi, and see if it reports the memory usage differently.