580 release feedback & discussion

580.65.06

  • Fixed a bug that could cause Vulkan applications to hang when destroying swapchains after a lost device event.
  • Fixed a bug that could allow atomic commit and other DRM operations to return success status despite having failed due to handling an interrupt: nv_drm_atomic_commit can silently fail due to down_interruptible in nvkms_ioctl_common being interrupted, causing the Jay wayland compositor to hang · Issue #832 · NVIDIA/open-gpu-kernel-modules · GitHub
  • Fixed a bug that could cause GTK 4 applications to crash when using the Vulkan backend on Wayland.
  • Fixed a bug that could intermittently cause llama.cpp to crash on exit when using the Vulkan backend: Misc. bug: Inconsistent Vulkan segfault · Issue #10528 · ggml-org/llama.cpp · GitHub
  • Added support for the fifo-v1 Wayland protocol on Vulkan.
  • Updated GPU clock value reporting in nvidia-settings, NVML, and nvidia-smi to show clocks before thermal and idle slowdowns for better consistency with the equivalent functionality on Windows.
  • Fixed compatibility with Bigscreen Beyond Head Mounted Displays.
  • Fixed a bug that could result in a black screen when setting specific modes on HDMI displays.
  • Fixed a bug that caused blank or frozen screens under the following conditions: nvidia-drm is loaded with the modeset=1 and fbdev=1 parameters, using a Maxwell or Pascal series GPU, and more than one display device of differing resolutions are connected.
  • Fixed a bug that caused nvidia-suspend.service to fail when available system memory is low.
  • Enabled RMIntrLockingMode by default. This feature can help reduce stutter especially when using virtual reality. This feature was originally introduced in the r570 series. It can be disabled by loading nvidia.ko with the NVreg_RegistryDwords=RMIntrLockingMode=0 kernel module parameter.
  • Implemented another feature that can reduce time spent in the interrupt top half for low latency display interrupts by deferring the work until later. This feature is experimental and disabled by default. This feature can be enabled by loading nvidia.ko with the NVreg_RegistryDwords=RmEnableAggressiveVblank=1 kernel module parameter.
  • Fixed a bug that could cause blank rendering on some single-buffered GLX applications when running on Xwayland.
  • Fixed a bug that could cause a kernel use-after-free on pre-Turing GPUs.
  • Fixed a bug that could cause OpenGL applications and compositors to stall when using NVIDIA as a PRIME Display Offload sink (“Reverse PRIME”), potentially resulting in a black screen.
  • Fixed a bug that led to increasing memory usage in X11 OpenGL and Vulkan applications after suspend/resume cycles.
  • Fixed a bug that could cause 32-bit x86 applications running on recent builds of glibc to crash on dlopen().
15 Likes

Thank you

Edit: DX12 is not the needful.

When will the 580 drivers come out?

580.65.06 (beta) is available here: Driver Details | NVIDIA

Still nothing about a fix for dx12 games, but still thank you for your hard work.

1 Like

Let’s benchmark it first. It may not be that explicit in the changelog. 😙

1 Like

That’s what I am waiting for before I start crying about it.

Hello. I am using Gnome 48 on my GTX 1050ti. After installing this driver whenever I open and then close any Gtk4 app, The app freezes and then after force quitting the app the DE crashes and takes me back to GDM Login Screen. This only happens on wayland not on x11.

1 Like

NVDEC still forces GPU into P2 CUDA state.

1 Like

This driver seems to have fixed stutters with GSP, Allocating GPU memory seems to be a system-wide blocking operation · Issue #164 · NVIDIA/egl-wayland · GitHub and Micro stutter with specific config · Issue #1647 · flightlessmango/MangoHud · GitHub. But unfortunately my Alienware AW2725DF monitor still randomly looses video signal when using VRR with low FPS: No signal with VRR when framerate is low. This issue can be easily reproduced by setting Adaptive Sync to “Always” in KDE Plasma.

Edit: I also have the same issue as @debestinator11 with GTK4 apps that hang at exit.

Edit2: Setting NVreg_RegistryDwords=RMIntrLockingMode=1 with 575 and probably 570 drivers also fixes stutters issue.

1 Like

I can confirm that all GTK4 apps now hang when attempting to exit with the the 580.65.06 Beta drivers.

  • kernel 6.15.9
  • hyprland 0.50.1 (-DNO_XWAYLAND:STRING=true)
  • gtk4 4.18.6
  • nVidia 580.65.06 Beta

I did not test X11 as I don’t have it installed/available.

Nothing too telling at first glance with strace. Futexes spinning per usual:

[pid  3142] futex(0x7ff08c000fc0, FUTEX_WAIT_BITSET_PRIVATE|FUTEX_CLOCK_REALTIME, 0, {tv_sec=1754342465, tv_nsec=694558000}, FUTEX_BITSET_MATCH_ANY <unfinished ...>
[pid  3122] <... poll resumed>)         = 0 (Timeout)
[pid  3122] poll([{fd=52, events=POLLIN|POLLPRI}], 1, 10) = 1 ([{fd=52, revents=POLLIN|POLLPRI}])
[pid  3122] poll([{fd=52, events=POLLIN|POLLPRI}], 1, 10) = 1 ([{fd=52, revents=POLLIN|POLLPRI}])
[pid  3122] poll([{fd=52, events=POLLIN|POLLPRI}], 1, 10 <unfinished ...>
[pid  3142] <... futex resumed>)        = -1 ETIMEDOUT (Connection timed out)
[pid  3142] futex(0x56041cbc0230, FUTEX_WAKE_PRIVATE, 1) = 0
[pid  3142] futex(0x7ff08c000fc0, FUTEX_WAIT_BITSET_PRIVATE|FUTEX_CLOCK_REALTIME, 0, {tv_sec=1754342465, tv_nsec=704685000}, FUTEX_BITSET_MATCH_ANY <unfinished ...>
[pid  3139] <... futex resumed>)        = -1 ETIMEDOUT (Connection timed out)
[pid  3139] futex(0x56041ce14a60, FUTEX_WAKE_PRIVATE, 1) = 0
[pid  3139] futex(0x7ff090000fc0, FUTEX_WAIT_BITSET_PRIVATE|FUTEX_CLOCK_REALTIME, 0, {tv_sec=1754342465, tv_nsec=795452000}, FUTEX_BITSET_MATCH_ANY <unfinished ...>
[pid  3122] <... poll resumed>)         = 0 (Timeout)
[pid  3122] poll([{fd=52, events=POLLIN|POLLPRI}], 1, 10 <unfinished ...>
[pid  3142] <... futex resumed>)        = -1 ETIMEDOUT (Connection timed out)
[pid  3142] futex(0x56041cbc0230, FUTEX_WAKE_PRIVATE, 1) = 0
[pid  3142] futex(0x7ff08c000fc0, FUTEX_WAIT_BITSET_PRIVATE|FUTEX_CLOCK_REALTIME, 0, {tv_sec=1754342465, tv_nsec=714778000}, FUTEX_BITSET_MATCH_ANY <unfinished ...>
[pid  3122] <... poll resumed>)         = 1 ([{fd=52, revents=POLLIN|POLLPRI}])
[pid  3122] poll([{fd=52, events=POLLIN|POLLPRI}], 1, 10) = 1 ([{fd=52, revents=POLLIN|POLLPRI}])
[pid  3122] poll([{fd=52, events=POLLIN|POLLPRI}], 1, 10 <unfinished ...>) = ?
[pid  3170] <... futex resumed>)        = ?
[pid  3161] <... futex resumed>)        = ?
[pid  3160] <... futex resumed>)        = ?
[pid  3159] <... futex resumed>)        = ?
[pid  3158] <... futex resumed>)        = ?
[pid  3167] <... futex resumed>)        = ?
[pid  3157] <... futex resumed>)        = ?
[pid  3182] <... futex resumed>)        = ?
[pid  3156] <... futex resumed>)        = ?
[pid  3155] <... futex resumed>)        = ?
[pid  3154] <... futex resumed>)        = ?
[pid  3153] <... futex resumed>)        = ?
[pid  3152] <... futex resumed>)        = ?
[pid  3151] <... futex resumed>)        = ?
[pid  3150] <... futex resumed>)        = ?
[pid  3149] <... futex resumed>)        = ?
[pid  3148] <... futex resumed>)        = ?
[pid  3147] <... futex resumed>)        = ?
[pid  3146] <... futex resumed>)        = ?
[pid  3145] <... futex resumed>)        = ?
[pid  3144] <... futex resumed>)        = ?
[pid  3143] <... futex resumed>)        = ?
[pid  3142] <... futex resumed>)        = ?
[pid  3140] <... futex resumed>)        = ?
[pid  3139] <... futex resumed>)        = ?
[pid  3137] <... futex resumed>)        = ?
[pid  3136] <... futex resumed>)        = ?
[pid  3131] <... futex resumed>)        = ?
[pid  3127] <... ppoll resumed> <unfinished ...>) = ?
[pid  3126] <... futex resumed>)        = ?
[pid  3125] <... ppoll resumed> <unfinished ...>) = ?
[pid  3124] <... ppoll resumed> <unfinished ...>) = ?
[pid  3123] <... futex resumed>)        = ?
[pid  3182] +++ killed by SIGKILL +++
2 Likes

I can reproduce this. I filed internal bug #5436037.

9 Likes

Most DXVK games do not run, vkd3d games do not run, mangohud overlay missing, steam overlay not working, obs-vkcapture does not work…

Some games still start but I think this affects games that require specific features.

See Elite Dangerous (359320) · Issue #150 · ValveSoftware/Proton · GitHub for an example.

1 Like

Is NVML PowerMizer supported on Windows too? What’s the relationship between it and the nvidia control panel setting? Is it different than before?

Tried a few games: HZD Remastered, TLOU, and L4D2. The games seem to run ok without issue in Wayland and NTsync. I enabled

NVreg_RegistryDwords=RmEnableAggressiveVblank=1

The games all ran fine and smooth. HZD Remaster though it looked like I got a 10-20 FPS bump. Before I was getting around 105 and now it is 120 ish. This is maxed out settings at 4K with DLAA.

I will keep testing my library and will report if I come across any issues.

The only GTK4 app I have on my system is Handbrake. As far as I can tell, it runs ok. The only thing borked is the window decoration for close because it will hang the program. If I quit the application through the File menu, it works fine.

The only oddity since enabling the agressive vblank is that it looks like when opening a child window sometimes (open/save file browser), the screen will go black for a few seconds.

GTK applications freeze on close.

2 Likes

I think this is related to the GTK4 close bug, but when using video player mpv, no controls work, and it does not close either (without forcing)..

1 Like

Fixed a bug that could cause GTK 4 applications to crash when using the Vulkan backend on Wayland.

I don’ know what’s fixed, but now GTK4 application are unusable on Vulkan.
Here’s quick fix if someone have that problem:

echo "export GSK_RENDERER=ngl" >> ~/.profile
5 Likes

with the 580 beta driver, everytime I try to close gtk4 apps they hang, and I’m not using the vulkan backend, I’m using the automatic renderer in KDE Plasma 6.4 which is OpenGL.

nvidia-bug-report.log.gz (888.5 KB)