MagicBeanDev/Hookshot

Archived

Watch

Fork 0

[Steam Deck / RADV] Device lost at intro cutscene → gameplay transition (12288 depth pass); SKIP_CODE not enough #1318

Closed

opened 2026-11-09 21:36:10 +00:00 by LucianoMSS · 3 comments

LucianoMSS commented

2026-11-09 21:36:10 +00:00

(Migrated from github.com)

Hookshot version

v2.0.11 (source 10f983ad)

What kind of problem

Crash, freeze or black screen (Vulkan device lost / AMDGPU reset)

What happened

On a Steam Deck (RADV VANGOGH), the intro cutscene itself plays fine. The hang is 100% reproducible at the transition from that cutscene into gameplay (when the first 3D scene / Kokiri-area load starts after the movie).

Sequence every time:

Title / cutscene run at ~60 fps
Cutscene ends → gameplay transition begins
Log allocates a 12288×12288 depth/shadow target (~1.2–1.7 GB of live textures)
Long submits / a ~1.2 s dispatch
radv/amdgpu: The CS has been cancelled because the context is lost
vk: device lost
Kernel: amdgpu: ring gfx_0.0.0 timeout → GPU reset

In Desktop Mode this GPU reset also killed Xwayland (DrKonqi popup). That looks like collateral damage from the GPU reset, not a separate Wayland bug.

Where in the game

Right after the first intro cutscene, at the cutscene → gameplay transition (before/as the first playable 3D scene comes in). Cutscene playback itself is OK.

How often

Every time

Graphics card and driver version

Device: AMD Custom GPU 0405 (RADV VANGOGH) (Steam Deck)
Vulkan 1.4.350
Mesa 26.2.0-devel (git-5c65d4fc31) / driverVersion 26.1.99
SteamOS 3.9.5 (BUILD_ID=20261023.100), kernel 7.2.7-valve2-2-neptune-72-...

CPU, RAM and operating system

Steam Deck, SteamOS 3.9.5 (Desktop Mode for these runs; same Hookshot hang pattern across runs)

Changes from the defaults

hookshot_options.ini (graphics):

quality_preset=1 (Minimum / 4GB)
frame_generation=0
resolution=4 (50%)
output_res=0
handheld_profile=1
upscaling=0
sharpness=0
fps_counter=1
game_speed=0 (per frame)

hookshot_env.txt for the latest run:

RX_VK_VRAM_MB=2500
RX_VK_TEX_MAX_MB=768
RX_VK_TEX_MAX_DIM=4096
RX_VK_LOW_SPEC=1
RX_VK_KEEP_RETIRED=1
RX_VK_SKIP_CODE=2c56dc1006942f74
RX_VK_SKIP_PROG=ffb53ed102dd9908

Other software hooked into the game

None

What we already tried (still hangs)

Attempt	Result
Minimum preset + handheld + FG off + 50% res	Still hangs at same transition
v2.0.8 → v2.0.10 → v2.0.11	Still hangs
`RX_VK_VRAM_MB` / `TEX_MAX_*` / `KEEP_RETIRED=1`	Env applies (`budget 2500 MB`), still hangs
`RX_VK_SKIP_CODE=2c56dc1006942f74` (#817 / v2.0.11 readme)	Does engage (many programs left out), still hangs
Also `RX_VK_SKIP_PROG=ffb53ed102dd9908`	Does engage; 12288 pass draw count drops (~3400 → ~337); last program on the clear becomes `f2999864d78ae1d3`; still device lost after a ~1.2 s dispatch

So this looks related to but not fixed by the Deck culling workaround in #1165 / #817. Skipping 2c56dc1006942f74 is not enough for this intro→gameplay transition.

Notable log lines (latest run, with both skips)

vk: draws of program ffb53ed102dd9908 left out (RX_VK_SKIP_PROG); its code: 0:365b92d206544e3f 1:acfb185a608654e9
...
vk: texture 0x113acf6000 12288x12288x1 x1 levels 1 NVN 0x32 -> vk 124 (258 textures, 1252 MB)
vk: a clear after 337 draws of its pass (12288x12288, 0 colour targets + depth, last program f2999864d78ae1d3): a pass of its own
vk: a clear after 337 draws of its pass (4096x4096, 0 colour targets + depth, last program f2999864d78ae1d3): a pass of its own
...
vk: slow submit: 1265.7 ms: texture lookups 4.5 uploads 4.3 dispatch 1258.3 queue submit 7.8
radv/amdgpu: The CS has been cancelled because the context is lost. This context is innocent.
vk: device lost: reading the fault info (RX_VK_CHECKPOINTS=1 for markers)
vi: 11531 frames presented (0.4 fps over the last 12.2 s, interval 1)

Kernel:

amdgpu: ring gfx_0.0.0 timeout
amdgpu: GPU reset begin!
amdgpu: GPU reset succeeded

Without SKIP_PROG, the same 12288 clear previously showed last program ffb53ed102dd9908 with ~3400+ draws — same transition, same end state.

#817 / #1000 — Steam Deck gfx_0.0.0 timeout (culling task shader 2c56dc1006942f74); workaround shipped in v2.0.11 as RX_VK_SKIP_CODE, but does not cover this intro transition hang
#996 / #708 — AMD device lost family (mostly Windows; same device lost / GPU hang pattern)

Happy to retest with RX_VK_CHECKPOINTS=1 (or any experimental build) if that helps name the dispatch that hangs after the 12288 clear.

Before you send it

I'm on the newest release, or the problem is still there on it.
I searched the open issues for this problem.
I'm not attaching game files, keys, hookshot.exe, the .bin caches or saves.

### Hookshot version v2.0.11 (source `10f983ad`) ### What kind of problem Crash, freeze or black screen (Vulkan device lost / AMDGPU reset) ### What happened On a Steam Deck (RADV VANGOGH), the **intro cutscene itself plays fine**. The hang is **100% reproducible at the transition from that cutscene into gameplay** (when the first 3D scene / Kokiri-area load starts after the movie). Sequence every time: 1. Title / cutscene run at ~60 fps 2. Cutscene ends → gameplay transition begins 3. Log allocates a **12288×12288** depth/shadow target (~1.2–1.7 GB of live textures) 4. Long submits / a ~1.2 s dispatch 5. `radv/amdgpu: The CS has been cancelled because the context is lost` 6. `vk: device lost` 7. Kernel: `amdgpu: ring gfx_0.0.0 timeout` → `GPU reset` In Desktop Mode this GPU reset also killed **Xwayland** (DrKonqi popup). That looks like collateral damage from the GPU reset, not a separate Wayland bug. ### Where in the game Right after the first intro cutscene, at the cutscene → gameplay transition (before/as the first playable 3D scene comes in). Cutscene playback itself is OK. ### How often Every time ### Graphics card and driver version - Device: `AMD Custom GPU 0405 (RADV VANGOGH)` (Steam Deck) - Vulkan 1.4.350 - Mesa `26.2.0-devel (git-5c65d4fc31)` / driverVersion `26.1.99` - SteamOS 3.9.5 (`BUILD_ID=20261023.100`), kernel `7.2.7-valve2-2-neptune-72-...` ### CPU, RAM and operating system Steam Deck, SteamOS 3.9.5 (Desktop Mode for these runs; same Hookshot hang pattern across runs) ### Changes from the defaults `hookshot_options.ini` (graphics): - `quality_preset=1` (Minimum / 4GB) - `frame_generation=0` - `resolution=4` (50%) - `output_res=0` - `handheld_profile=1` - `upscaling=0` - `sharpness=0` - `fps_counter=1` - `game_speed=0` (per frame) `hookshot_env.txt` for the latest run: ``` RX_VK_VRAM_MB=2500 RX_VK_TEX_MAX_MB=768 RX_VK_TEX_MAX_DIM=4096 RX_VK_LOW_SPEC=1 RX_VK_KEEP_RETIRED=1 RX_VK_SKIP_CODE=2c56dc1006942f74 RX_VK_SKIP_PROG=ffb53ed102dd9908 ``` ### Other software hooked into the game None ### What we already tried (still hangs) | Attempt | Result | |---|---| | Minimum preset + handheld + FG off + 50% res | Still hangs at same transition | | v2.0.8 → v2.0.10 → **v2.0.11** | Still hangs | | `RX_VK_VRAM_MB` / `TEX_MAX_*` / `KEEP_RETIRED=1` | Env applies (`budget 2500 MB`), still hangs | | `RX_VK_SKIP_CODE=2c56dc1006942f74` (#817 / v2.0.11 readme) | **Does engage** (many programs left out), **still hangs** | | Also `RX_VK_SKIP_PROG=ffb53ed102dd9908` | **Does engage**; 12288 pass draw count drops (~3400 → ~337); last program on the clear becomes `f2999864d78ae1d3`; **still device lost** after a ~1.2 s dispatch | So this looks **related to but not fixed by** the Deck culling workaround in #1165 / #817. Skipping `2c56dc1006942f74` is not enough for this intro→gameplay transition. ### Notable log lines (latest run, with both skips) ``` vk: draws of program ffb53ed102dd9908 left out (RX_VK_SKIP_PROG); its code: 0:365b92d206544e3f 1:acfb185a608654e9 ... vk: texture 0x113acf6000 12288x12288x1 x1 levels 1 NVN 0x32 -> vk 124 (258 textures, 1252 MB) vk: a clear after 337 draws of its pass (12288x12288, 0 colour targets + depth, last program f2999864d78ae1d3): a pass of its own vk: a clear after 337 draws of its pass (4096x4096, 0 colour targets + depth, last program f2999864d78ae1d3): a pass of its own ... vk: slow submit: 1265.7 ms: texture lookups 4.5 uploads 4.3 dispatch 1258.3 queue submit 7.8 radv/amdgpu: The CS has been cancelled because the context is lost. This context is innocent. vk: device lost: reading the fault info (RX_VK_CHECKPOINTS=1 for markers) vi: 11531 frames presented (0.4 fps over the last 12.2 s, interval 1) ``` Kernel: ``` amdgpu: ring gfx_0.0.0 timeout amdgpu: GPU reset begin! amdgpu: GPU reset succeeded ``` Without `SKIP_PROG`, the same 12288 clear previously showed `last program ffb53ed102dd9908` with ~3400+ draws — same transition, same end state. ### Related issues - #817 / #1000 — Steam Deck `gfx_0.0.0 timeout` (culling task shader `2c56dc1006942f74`); workaround shipped in v2.0.11 as `RX_VK_SKIP_CODE`, but **does not cover this intro transition hang** - #996 / #708 — AMD device lost family (mostly Windows; same `device lost` / GPU hang pattern) Happy to retest with `RX_VK_CHECKPOINTS=1` (or any experimental build) if that helps name the dispatch that hangs after the 12288 clear. ### Before you send it - [x] I'm on the newest release, or the problem is still there on it. - [x] I searched the open issues for this problem. - [x] I'm not attaching game files, keys, hookshot.exe, the .bin caches or saves.

LucianoMSS commented

2026-11-09 21:36:20 +00:00

(Migrated from github.com)

Attached evidence

Sanitized Deck log excerpt (version / env / options / skip+crash matches / last 250 lines):

Creating gist hookshot-steamdeck-intro-transition-crash.txt
✓ Created public gist hookshot-steamdeck-intro-transition-crash.txt
https://gist.github.com/LucianoMSS/d7cb4148e85c7dc501f11ebd86892973

Clarification on timing

Cutscene playback is fine. The hang is specifically at the transition into gameplay after the cutscene ends (first 3D load), not during the movie itself.

### Attached evidence Sanitized Deck log excerpt (version / env / options / skip+crash matches / last 250 lines): - Creating gist hookshot-steamdeck-intro-transition-crash.txt ✓ Created public gist hookshot-steamdeck-intro-transition-crash.txt https://gist.github.com/LucianoMSS/d7cb4148e85c7dc501f11ebd86892973 ### Clarification on timing Cutscene playback is fine. The hang is specifically at the **transition into gameplay after the cutscene ends** (first 3D load), not during the movie itself.

NotRexed commented

2026-11-09 21:46:50 +00:00

(Migrated from github.com)

Thanks for the detailed timeline (the 12288x12288 depth/shadow target at the cutscene-to-gameplay transition). Linking this to the Linux and Steam Deck tracker #1000, where the same Steam Deck crash after the intro is collected (#1061, #1258).

❤️ 1

NotRexed commented

2026-11-09 22:52:08 +00:00

(Migrated from github.com)

Thanks for the report. To keep each problem in one place, this is tracked in #1000 (Linux and Steam Deck: GPU hangs, crashes and low FPS), which links your details. Closing this as a duplicate; please follow #1000 for updates.

Commenting is not possible because the repository is archived.