Running this file with latest master causes intermittent crashes on my machine under the debugger.
It throws VK_ERROR_DEVICE_LOST
demod.scopesession.txt
This is with Windows 11, MSYS2, gpu is a GTX 1080
It is possible the GPU is dying, but I haven't noticed this elsewhere.
It does not crash when running FurMark for a while - if you know a more appropriate stress test I am happy to run it and see.
The GPU info printed on startup is below:
Initializing Vulkan
VK_KHR_get_physical_device_properties2: supported
VK_EXT_debug_utils: supported
Loader/API support available for Vulkan 1.4
Vulkan 1.2 support available, requesting it
Initializing glfw 3.4.0 Win32 WGL Null EGL OSMesa MinGW-w64 DLL
GLFW required extensions:
VK_KHR_surface
VK_KHR_win32_surface
Physical devices:
Device 0: NVIDIA GeForce GTX 1080
API version: 0x00403118 (0.1.3.280)
Driver version: 0x8c178000 (560.94.0.0)
Vendor ID: 10de
Device ID: 1b80
Device type: Discrete GPU
int64: yes
int16: yes (allowed in SSBOs)
int8: yes (allowed in SSBOs)
atomic int64 in SSBO: yes
atomic int64 in shared: yes
Max image dim 2D: 32768
Max storage buf range: 4095 MB
Max mem objects: 4294967295
Max compute shared mem: 48 KB
Max compute grp count: 2147483647 x 65535 x 65535
Max compute invocs: 1536
Max compute grp size: 1536 x 1024 x 64
Memory types:
Type 0
Heap index: 1
Type 1
Heap index: 1
Type 2
Heap index: 1
Type 3
Heap index: 1
Type 4
Heap index: 1
Type 5
Heap index: 1
Type 6
Heap index: 1
Type 7
Heap index: 0
Device local
Type 8
Heap index: 1
Host visible
Host coherent
Type 9
Heap index: 1
Host visible
Host coherent
Host cached
Type 10
Heap index: 2
Device local
Host visible
Host coherent
Memory heaps:
Heap 0
Size: 7 GB
Device local
Heap 1
Size: 15 GB
Heap 2
Size: 214 MB
Device local
Selected device 0
Queue families (5 total)
Queue type 0
Queue count: 16
Timestamp valid bits: 64
Graphics
Compute
Transfer
Sparse binding
Queue type 1
Queue count: 2
Timestamp valid bits: 64
Transfer
Sparse binding
Queue type 2
Queue count: 8
Timestamp valid bits: 64
Compute
Transfer
Sparse binding
Queue type 3
Queue count: 1
Timestamp valid bits: 32
Transfer
Sparse binding
Video decode
Queue type 4
Queue count: 2
Timestamp valid bits: 32
Transfer
Sparse binding
Video encode
Driver: vk::DriverId::eNvidiaProprietary
Enabling 64-bit float support
Enabling 64-bit integer support
Enabling 16-bit integer support
Enabling 32-bit atomic float support for SSBOs and shared memory
Enabling 16-bit integer support for SSBOs
Enabling 8-bit integer support
Enabling 64-bit atomic int support for SSBOs
Enabling 64-bit atomic int support for shared memory
Enabling 8-bit integer support for SSBOs
Device has VK_KHR_push_descriptor, requesting it
Device has VK_KHR_shader_atomic_int64, requesting it
Device has VK_KHR_shader_non_semantic_info, requesting it
Device has VK_EXT_memory_budget, requesting it
Device has VK_EXT_shader_atomic_float, requesting it
Using heap 1, type 9 for pinned host memory
Using heap 0, type 7 for card-local memory
Sorted queues:
Family=1 Index=1 Flags=0000000c
Family=1 Index=0 Flags=0000000c
Family=4 Index=1 Flags=0000004c
Family=4 Index=0 Flags=0000004c
Family=3 Index=0 Flags=0000002c
Family=2 Index=7 Flags=0000000e
Family=2 Index=6 Flags=0000000e
Family=2 Index=5 Flags=0000000e
Family=2 Index=4 Flags=0000000e
Family=2 Index=3 Flags=0000000e
Family=2 Index=2 Flags=0000000e
Family=2 Index=1 Flags=0000000e
Family=2 Index=0 Flags=0000000e
Family=0 Index=14 Flags=0000000f
Family=0 Index=15 Flags=0000000f
Family=0 Index=0 Flags=0000000f
Family=0 Index=13 Flags=0000000f
Family=0 Index=12 Flags=0000000f
Family=0 Index=11 Flags=0000000f
Family=0 Index=10 Flags=0000000f
Family=0 Index=9 Flags=0000000f
Family=0 Index=8 Flags=0000000f
Family=0 Index=7 Flags=0000000f
Family=0 Index=6 Flags=0000000f
Family=0 Index=5 Flags=0000000f
Family=0 Index=4 Flags=0000000f
Family=0 Index=3 Flags=0000000f
Family=0 Index=2 Flags=0000000f
Family=0 Index=1 Flags=0000000f
Running this file with latest master causes intermittent crashes on my machine under the debugger.
It throws VK_ERROR_DEVICE_LOST
demod.scopesession.txt
This is with Windows 11, MSYS2, gpu is a GTX 1080
It is possible the GPU is dying, but I haven't noticed this elsewhere.
It does not crash when running FurMark for a while - if you know a more appropriate stress test I am happy to run it and see.
The GPU info printed on startup is below: