VK_ERROR_DEVICE_LOST (-4) on compute dispatch after intro movie (GTX 1660 SUPER) #266

Closed
opened 2026-11-07 18:15:41 +00:00 by entonimontoril-del · 1 comment
entonimontoril-del commented 2026-11-07 18:15:41 +00:00 (Migrated from github.com)

Hey folks, running into a Vulkan device crash right as the 3D scene loads after the intro video. Tested on two driver versions, same behavior.

System Specs

  • GPU: NVIDIA GeForce GTX 1660 SUPER (6 GB)
  • Driver: 610.88 / 621.14
  • CPU: AMD Ryzen 5 5600GT
  • RAM: 16 GB DDR4
  • Storage: 256 GB NVMe SSD (Patriot) (~100 GB free space)
  • OS: Windows 10/11
  • Render backend: Vulkan (hookshot_vk.dll)

Issue Details

  • Crash: VK_ERROR_DEVICE_LOST (-4) at frame ~1879
  • VRAM usage at crash: ~1.1 GB out of 5.2 GB (doesn't look like an OOM issue)
  • Flags enabled: movie-sync, less-stutter | Frame Gen: OFF

Key log findings

  1. Massive compute dispatch stalls right after uploading 12288x12288 texture (0x113acf6000):
    • slow dispatch 501.2 ms: program 3de6d86ec7e6720e
    • slow submit: 530.4 ms: dispatch 530.2
  2. TAAU network fd029a317a9a4b94 fallback triggers a ~340ms submission delay.
  3. Unbound storage buffer warnings during compute dispatches (53caac5f38279735 and 7984161db3e053b5).

Looks like the TDR timeout is kicking in due to the heavy compute dispatch chain/TAAU fallback. Setting RX_VK_TAAU_KEYS=- seems to mitigate it.

Log attached for reference.

Hey folks, running into a Vulkan device crash right as the 3D scene loads after the intro video. Tested on two driver versions, same behavior. ### System Specs - GPU: NVIDIA GeForce GTX 1660 SUPER (6 GB) - Driver: 610.88 / 621.14 - CPU: AMD Ryzen 5 5600GT - RAM: 16 GB DDR4 - Storage: 256 GB NVMe SSD (Patriot) (~100 GB free space) - OS: Windows 10/11 - Render backend: Vulkan (hookshot_vk.dll) ### Issue Details - Crash: `VK_ERROR_DEVICE_LOST (-4)` at frame ~1879 - VRAM usage at crash: ~1.1 GB out of 5.2 GB (doesn't look like an OOM issue) - Flags enabled: `movie-sync`, `less-stutter` | Frame Gen: OFF ### Key log findings 1. Massive compute dispatch stalls right after uploading 12288x12288 texture (`0x113acf6000`): - `slow dispatch 501.2 ms: program 3de6d86ec7e6720e` - `slow submit: 530.4 ms: dispatch 530.2` 2. TAAU network `fd029a317a9a4b94` fallback triggers a ~340ms submission delay. 3. Unbound storage buffer warnings during compute dispatches (`53caac5f38279735` and `7984161db3e053b5`). Looks like the TDR timeout is kicking in due to the heavy compute dispatch chain/TAAU fallback. Setting `RX_VK_TAAU_KEYS=-` seems to mitigate it. Log attached for reference.
NotRexed commented 2026-11-07 19:38:50 +00:00 (Migrated from github.com)

Thanks for the detailed report. This is the device-lost crash tracked in #150. Your notes (the unbound storage buffers in 53caac5f38279735 and 7984161db3e053b5, and RX_VK_TAAU_KEYS=- seeming to avoid it) match other NVIDIA reports there, so they help. Closing as a duplicate; if you can, attach the log in #150, and please try https://github.com/NotRexed/Hookshot/releases/tag/foliage-linux-updater.

Thanks for the detailed report. This is the device-lost crash tracked in #150. Your notes (the unbound storage buffers in 53caac5f38279735 and 7984161db3e053b5, and RX_VK_TAAU_KEYS=- seeming to avoid it) match other NVIDIA reports there, so they help. Closing as a duplicate; if you can, attach the log in #150, and please try https://github.com/NotRexed/Hookshot/releases/tag/foliage-linux-updater.
Commenting is not possible because the repository is archived.
No milestone
No assignees
1 participant
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
MagicBeanDev/Hookshot#266
No description provided.