Hi
I’m fighting against a random Vulkan crash.
I get a VK_DEVICE_LOST after random time (seconds to minutes), when i run my program, without doing nothing else but waiting (not moving mouse, nor keyboard, etc…)
I use core 1.2 features only, nothing about RTX specifics extensions.
It seams to only appear on RTX cards.
-
my RTX2080 Win10, lastest drivers : crashs mostly < 2min
-
friend RTX2070 Win10, lastest drivers : crashs mostly < 2min
-
client RTX2060 Win10, lastest drivers : crashs mostly < 2min
-
my GTX 980 Win10, lastest drivers : run hours
-
my GTX 970m Win10, lastest drivers : run hours
-
friend GTX 1080 Win10, lastest drivers : run hours
-
my core i7 IGP Win10, lastest drivers : run hours (but slowly :D)
My program run like that :
- update 2 buffers (Time, transform matrix)
- draw cascaded shadows (multi view)
- draw deferred
- lighting and fx passes (many compute shaders)
- render on screen
- HUD (fps counter)
i have try reduce the gpu work, with only few lowpoly objects
- update 2 buffers (Time, transform matrix)
- draw cascaded shadows (multi view)
- no draw deferred
- no compute shaders
- no render on screen (yes, we can’t see anything, not even a clean screen)
- HUD (fps counter) (crash with or without)
it crash randomly, after saying device is lost : in debug mode, my callback debugbreak from everywhere in all my code. mostly in vkWaitForFences, and a few times in vkQueueSubmit. The error location is not always at the same place, it happen from start to the end instruction of a frame, it seams to be randomly too.
The first frames are doing initialisations, after that, nothing change : it’s always the same rendering, always always identical.
I try :
- different recent drivers version, no differences
- 2 most recent vulkan SDK, no differences
I spend so many days trying finding the problem, i can’t work without resolving it. I really ask for help !!