There seems to be huge performance deficit in KDE compared to just about any other DE(below are comparisons between KDE and GNOME both on wayland)
KDE Plasma:
GNOME:
Nvidia stop generating frames when Vsync is on while playing with wine-wayland (native wayland) or gamescope.
in vk3d title you have to disable the extension which work, but for dxvk you have to disable vsync or the game hangs at start.
As wine wayland is almost ready, this should get fixed. Basically according to my understanding if this vk extension is disabled, we can’t have actual vsync working? Either way it 's able to get vk3d and gamescope to work with disabling but dxvk you can’t disable the extension, either you have to disable vsync in game, which is hard if the game keep getting to stuck at start.
VK_KHR_present_wait
Gamescope issue mention here
Attach Game log from AC Brotherhood running on native wine-wayland 10.3
game-log.log (15.6 KB)
nvidia-bug-report.log.gz (598.5 KB)
Would be great if you can let us know if it’s being worked or acknowledged
Can you make your game is using direct scannout?
You can test it out Settings->Desktop Effects->Show Compositing
When the game is full screen, the Compositing Red Text on top right should disappear, if it doesn’t disappear,
Try to set Color Profile to None in display setting and HDR off and try again. Make sure game is in fullscreen and run the test.
I have tried all that and the compositing does not disappear.
color profile: none
HDR: off
color accuracy: prefer efficiency
VRR: on
tearing: on
Same results, additionally I tried VRR: on, tearing off, vrr: off, tearing: off, no dice.
This is on a clean cachyos result, additionally I tried on a laptop with amd+nvidia and the same thing, compositing stays top right.
Yup that can be an issue for you losing the performance.
I think you already reported a bug there?
Also can you check if gnome is using direct scannout? You can check it using Alt+F2, type lg and there should be an option to show opaque or something, been a while but there should be an option in last tab.
It is working on gnome, also if I check journalctl, drm_info and everything else it appears that kde “should” also be doing direct scanout, but it isn’t.
For reference I also tried adding a window rule to the game with the rule “block compositing” which also didn’t help.
Not sure why it’s happening to you, for me it’s working fine.
Gnome other hand sometime the direct scannout stop working for me there, if I alt tab to many time
That’s very odd, I do not have anything HDR capable so I can’t test that, but I tried Cyberpunk and Compositing does disappear here.
Hi all
Just reporting an additional problem that could be related to some of the issues others are experiencing with this new version. I’m on Arch Linux 6.13.4 and I have an RTX 4070 Ti.
After upgrading to 570.124.04, my screen (a single monitor) would just go dark after a few hours as if no signal was being detected. I couldn’t discern any pattern, sometimes it took a couple of hours, sometimes more than a day.
I’ve downgraded to 570.86.16 since then and the issue went away. Unfortunately I couldn’t grab any relevant session details because it always forced me to restart, but you can find some additional details here:
Getting a lot of crashes. Very “unstable” stable drivers.
Just by browsing on Firefox (KDE).
Mar 15 12:00:39 zzzzzz kernel: NVRM: GPU at PCI:0000:08:00: GPU-738d510d-abfd-7290-16db-025b3eb23693
Mar 15 12:00:39 zzzzzz kernel: NVRM: Xid (PCI:0000:08:00): 79, GPU has fallen off the bus.
Mar 15 12:00:39 zzzzzz kernel: NVRM: GPU 0000:08:00.0: GPU has fallen off the bus.
Mar 15 12:00:39 zzzzzz kernel: NVRM: Xid (PCI:0000:08:00): 154, GPU recovery action changed from 0x0 (None) to 0x1 (GPU Reset Required)
Mar 15 12:00:39 zzzzzz kernel: [drm:__nv_drm_semsurf_wait_fence_work_cb [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00000800] Failed to register auto-value-update on pre-wait value for sync FD semaphore surface
Mar 15 12:00:39 zzzzzz kwin_wayland[1071]: kwin_wayland_drm: Pageflip timed out! This is a bug in the nvidia-drm kernel driver
Mar 15 12:00:39 zzzzzz kwin_wayland[1071]: kwin_wayland_drm: Please report this at https://forums.developer.nvidia.com/c/gpu-graphics/linux
Mar 15 12:00:39 zzzzzz kwin_wayland[1071]: kwin_wayland_drm: With the output of 'sudo dmesg' and 'journalctl --user-unit plasma-kwin_wayland --boot 0'
Mar 15 12:00:40 zzzzzz kwin_wayland[1071]: kwin_wayland_drm: Pageflip timed out! This is a bug in the nvidia-drm kernel driver
Mar 15 12:00:40 zzzzzz kwin_wayland[1071]: kwin_wayland_drm: Please report this at https://forums.developer.nvidia.com/c/gpu-graphics/linux
Mar 15 12:00:40 zzzzzz kwin_wayland[1071]: kwin_wayland_drm: With the output of 'sudo dmesg' and 'journalctl --user-unit plasma-kwin_wayland --boot 0'
Mar 15 12:00:43 zzzzzz kwin_wayland[1071]: kwin_wayland_drm: Pageflip timed out! This is a bug in the nvidia-drm kernel driver
Mar 15 12:00:43 zzzzzz kwin_wayland[1071]: kwin_wayland_drm: Please report this at https://forums.developer.nvidia.com/c/gpu-graphics/linux
Mar 15 12:00:43 zzzzzz kwin_wayland[1071]: kwin_wayland_drm: With the output of 'sudo dmesg' and 'journalctl --user-unit plasma-kwin_wayland --boot 0'
Mar 15 12:00:44 zzzzzz kernel: [drm:__nv_drm_semsurf_wait_fence_work_cb [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00000800] Failed to register auto-value-update on pre-wait value for sync FD semaphore surface
nvidia-bug-report.log.gz (794.3 KB)
Clean install: Arch Rolling
Kernel 6.13.6-arch1-1
CPU: AMD Ryzen 9 9900X3D
GPU: nvidia RTX 5070 (non-TI)
Driver: NVIDIA 570.124.04 – nvidia-open-dkms
Desktop Env: KDE Plasma
Ran a clean install of Arch, installed KDE and the nvidia drivers through CLI, everything appears to be working except the GPU.
Tests conducted:
Phoronix:
Unigine Heaven: ~30 FPS
CPU Stress Test: 100% on all cores, cancelled early, only testing CPU capability
via Steam:
Kingdom Come: Deliverance (Proton): ~13 FPS
V Rising (Proton, GE Proton): 15 - 17 FPS
– Testing variations: Wayland, X11, launch option: DXVK_HUD=1 __GLX_VENDOR_LIBRARY_NAME=nvidia __GL_THREADED_OPTIMIZATIONS=1 __GL_SHADER_DISK_CACHE=1 __GL_SHADER_DISK_CACHE_PATH=/tmp __GL_GSYNC_ALLOWED=1 __GL_VRR_ALLOWED=1 __VK_LAYER_NV_optimus=NVIDIA_only %command%
The Talos Principle (Native): ~15 FPS
via Arch:
Xonotic (Native): 60 FPS, occassionally getting up to 120 FPS
SuperTuxKart: 60 FPS
0 A.D. : Menu ran at 60 FPS, unable to enter match due to crashing
Blender Render - BMW Benchmark
GPU (Cuda): 2:56.77
CPU: 1:03.12
Additional information:
The GPU never went over 75 watts
CPU seemed to never work hard until stress testing
I’ve run out of ideas. Overall, it just seems that nothing will make the GPU actually kick in and run like it’s supposed to. Any assistance or suggestions would be appreciated.
sudo nvidia-bug-report.sh
nvidia-bug-report.log.gz (618.9 KB)
I have exactly the same problem. See above - I get 95% performance loss with a 5070. I believe the 5070 driver is completely broken.
Here’s a new bug report with the latest 570.124 drivers. No performance gains.
nvidia-bug-report.log.gz (452.7 KB)
Can you run the test with compositing showing? cause for me the result are really minimal so i can’t tell if it effect performance for me, you can just select the color profile to ICC built in, it should make direct scannout to not turn on.
Fedora 41 Xfce updated everything and steam games freeze still and again (560-565-570) try to add bug report when games freeze at starting. Good luck.
nvidia-bug-report.log.gz (120.5 KB)
@BlueGoliath @shelter From what I can tell, there are applications that don’t get listed in nvidia-smi.
I’ve once tried running Plasma and Chrome with Vulkan-renderer instead of OpenGL - and most of the VRAM allocations weren’t showing up in nvidia-smi. Sometimes, it wouldn’t even list an application despite it being open as a window. If I closed those applications, total usage would go down again - so it was not leaked.
I think the real problem is nvidia-smi not listing all VRAM consumers properly or reliably.
Since, according to those experiments I did, the VRAM work-around doesn’t properly work for Vulkan applications, I’ve gone back to the default renderer to get a “working work-around” back.
As far as I can tell, gnome-shell now uses Vulkan by default. So this may be an issue. You may try booting your session with GSK_RENDER=ngl or GSK_RENDER=gl set in your environment to see if this fixes the behavior. I do not use gnome-shell, so I cannot try or verify.
Please be aware that Gnome considers the old render paths untested, so you may see bugs when not using the Vulkan renderer.
Xid 79 again:
NVRM: GPU at PCI:0000:01:00: GPU-1dfdd8a3-609f-ab69-f862-db0ee121b8ce
[ 39.078494] NVRM: GPU Board Serial Number: 0
[ 39.078495] NVRM: Xid (PCI:0000:01:00): 79, pid=3407, name=nvidia-settings, GPU has fallen off the bus.
[ 39.078498] NVRM: GPU 0000:01:00.0: GPU has fallen off the bus.
[ 39.078499] NVRM: GPU 0000:01:00.0: GPU serial number is 0.
[ 39.078503] NVRM: kgspRcAndNotifyAllChannels_IMPL: RC all channels for critical error 79.
[ 39.078508] NVRM: _threadNodeCheckTimeout: API_GPU_ATTACHED_SANITY_CHECK failed!
[ 39.078516] NVRM: _threadNodeCheckTimeout: API_GPU_ATTACHED_SANITY_CHECK failed!
[ 39.078520] NVRM: _threadNodeCheckTimeout: API_GPU_ATTACHED_SANITY_CHECK failed!
[ 39.078578] NVRM: _issueRpcAndWait: rpcSendMessage failed with status 0x0000000f for fn 78!
[ 39.078580] NVRM: nvCheckOkFailedNoLog: Check failed: GPU lost from the bus [NV_ERR_GPU_IS_LOST] (0x0000000F) returned from nvdEngineDumpCallbackHelper(pGpu, pPrbEnc, pNvDumpState, pEngineCallback) @ nv_debug_dump.c:274
[ 39.078602] NVRM: _issueRpcAndWait: rpcSendMessage failed with status 0x0000000f for fn 78!
[ 39.078603] NVRM: nvCheckOkFailedNoLog: Check failed: GPU lost from the bus [NV_ERR_GPU_IS_LOST] (0x0000000F) returned from nvdEngineDumpCallbackHelper(pGpu, pPrbEnc, pNvDumpState, pEngineCallback) @ nv_debug_dump.c:274
[ 39.078606] NVRM: _issueRpcAndWait: rpcSendMessage failed with status 0x0000000f for fn 78!
[ 39.078607] NVRM: nvCheckOkFailedNoLog: Check failed: GPU lost from the bus [NV_ERR_GPU_IS_LOST] (0x0000000F) returned from nvdEngineDumpCallbackHelper(pGpu, pPrbEnc, pNvDumpState, pEngineCallback) @ nv_debug_dump.c:274
[ 39.078623] NVRM: RmLogGpuCrash: RmLogGpuCrash: failed to save GPU crash data
[ 39.078626] NVRM: _kgspLogRpcSanityCheckFailure: GPU0 sanity check failed 0xf waiting for RPC response from GSP. Expected function 103 (GSP_RM_ALLOC) (0x0 0x78).
[ 39.078626] NVRM: GPU0 GSP RPC buffer contains function 78 (DUMP_PROTOBUF_COMPONENT) and data 0x0000000000000000 0x0000000000000000.
[ 39.078628] NVRM: GPU0 RPC history (CPU -> GSP):
[ 39.078629] NVRM: entry function data0 data1 ts_start ts_end duration actively_polling
[ 39.078630] NVRM: 0 103 GSP_RM_ALLOC 0x0000000000000000 0x0000000000000078 0x00063070c382cf67 0x0000000000000000 y
[ 39.078632] NVRM: -1 76 GSP_RM_CONTROL 0x00000000402c0105 0x0000000000000038 0x00063070c37d458b 0x00063070c37d960a 20607us
[ 39.078635] NVRM: -2 76 GSP_RM_CONTROL 0x00000000402c0105 0x0000000000000038 0x00063070c37cf53e 0x00063070c37d4575 20535us
[ 39.078636] NVRM: -3 76 GSP_RM_CONTROL 0x00000000402c0105 0x0000000000000038 0x00063070c37ca4e0 0x00063070c37cf52b 20555us
[ 39.078638] NVRM: -4 76 GSP_RM_CONTROL 0x00000000402c0105 0x0000000000000038 0x00063070c37c4d54 0x00063070c37ca4cc 22392us
[ 39.078639] NVRM: -5 76 GSP_RM_CONTROL 0x000000002080e634 0x0000000000000188 0x00063070c37c3bdd 0x00063070c37c4d47 4458us
[ 39.078641] NVRM: -6 76 GSP_RM_CONTROL 0x00000000402c0105 0x0000000000000038 0x00063070c37beae9 0x00063070c37c3b88 20639us
[ 39.078642] NVRM: -7 76 GSP_RM_CONTROL 0x00000000402c0105 0x0000000000000038 0x00063070c37b99a8 0x00063070c37bead7 20783us
[ 39.078643] NVRM: GPU0 RPC event history (CPU <- GSP):
[ 39.078644] NVRM: entry function data0 data1 ts_start ts_end duration during_incomplete_rpc
[ 39.078645] NVRM: 0 4099 POST_EVENT 0x0000000000000021 0x0000000000000020 0x00063070c37beb4a 0x00063070c37beb4e 4us
[ 39.078647] NVRM: -1 4099 POST_EVENT 0x0000000000000021 0x0000000000000100 0x00063070c32148ea 0x00063070c32148f0 6us
[ 39.078648] NVRM: -2 4108 UCODE_LIBOS_PRINT 0x0000000000000000 0x0000000000000000 0x00063070c279b6f7 0x00063070c279b6f7
[ 39.078650] NVRM: -3 4108 UCODE_LIBOS_PRINT 0x0000000000000000 0x0000000000000000 0x00063070c279b5be 0x00063070c279b5be
[ 39.078651] NVRM: -4 4128 GSP_POST_NOCAT_RECORD 0x0000000000000002 0x0000000000000027 0x00063070c27996c8 0x00063070c27996ca 2us
[ 39.078653] NVRM: -5 4098 GSP_RUN_CPU_SEQUENCER 0x000000000000061c 0x0000000000003fe2 0x00063070c278e019 0x00063070c278f369 4944us
[ 39.078656] CPU: 14 UID: 1000 PID: 3407 Comm: nvidia-settings Tainted: G O 6.13.7-200.fc41.x86_64 #1
[ 39.078658] Tainted: [O]=OOT_MODULE
[ 39.078658] Hardware name: Dell Inc. Dell G15 5510/0H0F1D, BIOS 1.30.0 12/11/2024
[ 39.078660] Call Trace:
[ 39.078662] <TASK>
[ 39.078664] dump_stack_lvl+0x5d/0x80
[ 39.078670] _kgspRpcRecvPoll+0x37f/0x7a0 [nvidia]
[ 39.079014] _issueRpcAndWait+0x70/0x360 [nvidia]
[ 39.079179] rpcRmApiAlloc_GSP+0x1ed/0x500 [nvidia]
[ 39.079310] ? iterate_fd+0x82/0xa0
[ 39.079314] deviceConstruct_IMPL+0x52e/0x830 [nvidia]
[ 39.079494] ? deviceConstruct_IMPL+0x782/0x830 [nvidia]
[ 39.079672] ? nvDbg_Printf+0x30/0x40 [nvidia]
[ 39.079849] ? gpuGetByRef+0xc9/0x110 [nvidia]
[ 39.080024] ? __nvoc_ctor_RsResource+0x4e/0x70 [nvidia]
[ 39.080174] __nvoc_objCreate_Device+0x19e/0x2b0 [nvidia]
[ 39.080280] ? os_acquire_spinlock+0x12/0x30 [nvidia]
[ 39.080367] __nvoc_objCreateDynamic+0x46/0x80 [nvidia]
[ 39.080486] ? os_alloc_mem+0xc0/0xd0 [nvidia]
[ 39.080578] resservResourceFactory+0xa1/0x1d0 [nvidia]
[ 39.080721] _clientAllocResourceHelper+0x290/0x730 [nvidia]
[ 39.080852] ? os_acquire_spinlock+0x12/0x30 [nvidia]
[ 39.080948] ? tlsEntryGet+0x30/0x80 [nvidia]
[ 39.081069] serverAllocResourceUnderLock+0x3ec/0x990 [nvidia]
[ 39.081215] ? __serverFindClientEntryByHandle+0xcc/0x150 [nvidia]
[ 39.081335] ? portThreadGetCurrentThreadId+0x1d/0x30 [nvidia]
[ 39.081454] ? rmclientValidateLocks_IMPL+0x21/0xa0 [nvidia]
[ 39.081598] ? _serverLockClientWithLockInfo.constprop.0+0x105/0x280 [nvidia]
[ 39.081745] serverAllocResource+0x567/0x6e0 [nvidia]
[ 39.081897] rmapiAllocWithSecInfo+0x204/0x440 [nvidia]
[ 39.082052] rmapiAllocWithSecInfoTls+0x70/0xa0 [nvidia]
[ 39.082205] Nv04AllocWithAccessSecInfo+0x6a/0x80 [nvidia]
[ 39.082357] ? security_capable+0x50/0x150
[ 39.082362] RmIoctl+0xa3a/0xdc0 [nvidia]
[ 39.082548] ? os_get_current_tick+0x3b/0xa0 [nvidia]
[ 39.082635] ? os_acquire_spinlock+0x12/0x30 [nvidia]
[ 39.082726] ? portSyncSpinlockAcquire+0x1d/0x50 [nvidia]
[ 39.082841] rm_ioctl+0x67/0x4d0 [nvidia]
[ 39.083016] ? syscall_exit_to_user_mode+0x10/0x210
[ 39.083019] nvidia_ioctl.isra.0+0x458/0x7f0 [nvidia]
[ 39.083098] nvidia_unlocked_ioctl+0x1e/0x30 [nvidia]
[ 39.083177] __x64_sys_ioctl+0x94/0xc0
[ 39.083179] do_syscall_64+0x82/0x160
[ 39.083182] ? do_sys_openat2+0x9c/0xe0
[ 39.083184] ? get_close_on_exec+0x34/0x40
[ 39.083185] ? do_fcntl+0x2f0/0x640
[ 39.083188] ? __x64_sys_fcntl+0x6c/0xe0
[ 39.083190] ? syscall_exit_to_user_mode+0x10/0x210
[ 39.083191] ? do_syscall_64+0x8e/0x160
[ 39.083193] ? syscall_exit_to_user_mode+0x10/0x210
[ 39.083194] ? clear_bhb_loop+0x25/0x80
[ 39.083195] ? clear_bhb_loop+0x25/0x80
[ 39.083196] ? clear_bhb_loop+0x25/0x80
[ 39.083197] entry_SYSCALL_64_after_hwframe+0x76/0x7e
[ 39.083200] RIP: 0033:0x7fcf3ad9b4ad
[ 39.083221] Code: 04 25 28 00 00 00 48 89 45 c8 31 c0 48 8d 45 10 c7 45 b0 10 00 00 00 48 89 45 b8 48 8d 45 d0 48 89 45 c0 b8 10 00 00 00 0f 05 <89> c2 3d 00 f0 ff ff 77 1a 48 8b 45 c8 64 48 2b 04 25 28 00 00 00
[ 39.083222] RSP: 002b:00007ffcdcde2540 EFLAGS: 00000246 ORIG_RAX: 0000000000000010
[ 39.083225] RAX: ffffffffffffffda RBX: 0000000000000030 RCX: 00007fcf3ad9b4ad
[ 39.083226] RDX: 00007ffcdcde26c0 RSI: 00000000c030462b RDI: 0000000000000012
[ 39.083226] RBP: 00007ffcdcde2590 R08: 00007ffcdcde26c0 R09: 00007ffcdcde26e8
[ 39.083227] R10: 0000000000000000 R11: 0000000000000246 R12: 00007ffcdcde26c0
[ 39.083228] R13: 0000000000000012 R14: 00000000c030462b R15: 000000000000002b
[ 39.083229] </TASK>
[ 39.083333] NVRM: Xid (PCI:0000:01:00): 154, GPU recovery action changed from 0x0 (None) to 0x1 (GPU Reset Required)
[ 40.192396] nvidia-modeset: ERROR: GPU:0: Failed detecting connected display devices
[ 40.192457] nvidia-modeset: ERROR: GPU:0: Failed detecting connected display devices
[ 40.192480] nvidia-modeset: ERROR: GPU:0: Failed detecting connected display devices
[ 42.036485] NVRM: Error in service of callback
After that running nvidia-bug-report.sh prints this:
Summary of Skipped Sections:
Skipped Component | Details
================================================================================
mst output | mst not found
--------------------------------------------------------------------------------
nvlsm-bug-report.sh output | nvlsm-bug-report.sh not found
--------------------------------------------------------------------------------
nvidia-bug-report.log.gz (1.0 MB)
Does anyone know how long it was from the 5070Ti till a working driver came out? I’ve had this regular 5070 in my PC for over a week and it’s unusable for the primary purpose I bought it for with the broken driver. Should I keep waiting or just give up and return it for something else?