VK_KHR_present_{id,wait} causes device loss on Nvidia 525.60.11 on PRIME setups
similar issue https://github.com/HansKristian-Work/vkd3d-proton/issues/1315#issuecomment-1333836687
have u tried VKD3D_CONFIG=swapchain_legacy ?
does all ur dx12 not working or this only ?
@TURING-V2 No, it's not. That issue is about totally different games.
@TURING-V2 Still doesn't work with VKD3D_CONFIG=swapchain_legacy. However, I rolled back vkd3d-proton to whatever version Experimental had before 20221129 update and that one (3fdbec990d3dbbff63e66ab8e9fb1a226fabf4a1) still works so I'll try to bisect.
Would be funny if it turned out to be 3c2302276fe9112070a74d40cdb96b496f546662 here as well.
45e67a34ec1fec3d5ef152bb71d1d25d61b17e29 is the first bad commit
commit 45e67a34ec1fec3d5ef152bb71d1d25d61b17e29
Author: Hans-Kristian Arntzen <[email protected]>
Date: Tue Oct 18 12:08:43 2022 +0200
vkd3d: Enable VK_KHR_present_wait.
Signed-off-by: Hans-Kristian Arntzen <[email protected]>
libs/vkd3d/device.c | 14 ++++++++++++++
libs/vkd3d/vkd3d_private.h | 4 ++++
libs/vkd3d/vulkan_procs.h | 3 +++
3 files changed, 21 insertions(+)
Surely enough, latest master starts working just fine if I add VKD3D_DISABLE_EXTENSIONS=VK_KHR_present_id,VK_KHR_present_wait to launch options.
Hmm? The legacy swapchain with VKD3D_CONFIG=swapchain_legacy also shouldn't actually make use of VK_KHR_present_wait at all, so is it just enabling the extensions at all that breaks it? Sounds like a driver bug if so.
Indeed, the first bad commit also doesn't seem to make use of the extension so it would make sense that VKD3D_CONFIG=swapchain_legacy doesn't resolve it if the extension is enabled even if the new swapchain is not used.
I suppose I should send email to NV about this :/
@TURING-V2 No, it's not. That issue is about totally different games.
Allow me to shed some light , tried on another game Death Stranding same result opens black screen with half mangohud and freezes. dmesg spam NVRM Xid 32.
With VKD3D_DISABLE_EXTENSIONS=VK_KHR_present_id,VK_KHR_present_wait Death Stranding and CP2077(deleting app local cache) both starts working . with 525.60.11 driver latest vkd3d
On 515 driver without VKD3D_DISABLE_EXTENSIONS=VK_KHR_present_id,VK_KHR_present_wait Both games working fine .
Tried same results with Dying Light 2 . Any other dx12 u want me to test ?
I can say both are same issue Indeed A driver bug in Nvidia 525.60.11 .New Issue detailed report on there open-kernel-modules will maybe . wait for new driver )
EDIT: Can't confirm for desktop user Im mobile prime render offload. 525.60.11 release notes did some performance changes for prime offload .
Non prime setup, same issue. I'm the one mentioned by the Warframe issue
Hm, I am also able to reproduce this under specific circumstances. It reproduces in my Granite engine as well, so it's not vkd3d-proton doing something weird.
- Desktop: Does not repro
- My laptop not connected to HDMI with nvidia-xconfig --prime (NV shows up in glxinfo). Does not reproduce, but everything tears.
- Laptop not connected to HDMI without xorg.conf, Intel shows up in glxinfo. Now it hangs immediately by just enabling the extension.
I'm also able to reproduce this issue on my setup. In my case, the game giving me trouble is Control. It runs just fine on DX11 mode, but freezes my whole system when starting it on DX12 mode. Using the env variables VKD3D_DISABLE_EXTENSIONS=VK_KHR_present_id,VK_KHR_present_wait seems to solve the issue.
The game was running just fine on nvidia 520 driver.
Everything else seems to be running just fine - proton games, emulators, etc. The only other issue I found was Retroarch's Mupen64plus-Next libretro core crashing at startup when using the ParaLLEl-RDP plugin, which may or may not be related to this issue.
My system specs:
OS: Arch Linux
CPU: Intel i5-12500H
GPU: NVidia RTX 3050 Mobile
Kernel: 6.0.11-arch1-1
DE: KDE 5.26.4
Driver: nvidia 525.60.11-2
parallel-rdp uses Granite, which uses present_wait, so that's expected I guess.
Just updated to 525.60.13 which apparently still has this problem and additionally now the game started to crash for me with
either more or less right after entering the game or in the start/middle of a mission.
NVIDIA GeForce RTX 2060 SUPER with 8GB of VRAM - which was happily working before and still is after downgrading back to 525.60.11.
Fwiw latest NV Vulkan beta driver 525.47.04 has the following note in its changelog:
Fixed a bug where usage of VK_KHR_present_id could cause applications to crash with Xid 32 errors [Linux]
And indeed, this version works even with ac7e7de701ad336887c1cb38f8adc4e588005962 reverted on top of current master.
Launch options
VKD3D_DISABLE_EXTENSIONS=VK_KHR_present_id,VK_KHR_present_wait`x3 2022-12VKD3D_DISABLE_EXTENSIONS=VK_KHR_present_id,VK_KHR_present_waitx1 2022-12VKD3D_CONFIG=swapchain_legacy`x1 2022-12VKD3D_CONFIG=swapchain_legacyx2 2022-12VKD3D_CONFIG=swapchain_legacy`.x1 2022-12PROTON_LOG=1x1 2022-12VKD3D_CONFIG=breadcrumbs`:x1 2022-12
Attempting to launch Deep Rock Galactic in DX12 mode always results in a device loss.
DE: KDE Plasma / kwin X11
GPU: GTX 1660 Ti Mobile (PRIME render offload)
Driver: 525.60.11
Proton: experimental-7.0-20221129
Commit: 8c0f277b7c3880c6242c1aded189b1ae43ec161a (this is the current master, version bundled with Proton dies because of assert failure which has been fixed since then)
Log created with
PROTON_LOG=1 VKD3D_CONFIG=breadcrumbs: steam-548430.logWhen the error happens, an error is printed to dmesg: