Game: AMD Radeon 880M crashes with device lost during startup on nightly-363 #1234

Open
opened 2026-11-16 21:08:43 +00:00 by BeanIssues · 4 comments
Collaborator

Reported on Discord

Longshot version

nightly-363-0cab3548 (source 3b447cf5). Reporter says older builds used to work.

Channel

Nightly.

What happened

Longshot closed unexpectedly during startup after running for 39 seconds. Crash report identifies a graphics-driver failure; the crash record says device lost. Four crash/unclean-end records were recorded in the last day. The reporter asks whether this is related to #1040, which covers a Radeon 860M crash in Kokiri Forest; this report is a startup crash on a different GPU.

Reproduction steps, on-screen behavior, and whether every launch fails are unknown.

Graphics card and driver version

AMD Radeon(TM) 880M Graphics, AMD proprietary driver 26.11.1 (LLPC).

CPU, RAM and operating system

AMD Ryzen AI 9 365 w/ Radeon 880M (20 threads), 30 GB RAM, Windows 11 (build 26300).

Changes from the defaults

Crash report lists quality_preset=6, resolution=4, debug_tools=1, and memory_budget=1; most other listed settings are disabled/default. LONGSHOT_NO_UPDATE_CHECK=1.

Crash record

Longshot crash record 1
version: nightly-363-0cab3548 (source 3b447cf5)
run_id: 1a768200d771585cdb80ba16651486da
when: 20261116T205403Z
uptime_s: 39
reason: device lost
thread: host
module_offset: longshot.exe+0x1c254bab
last_present_age_ms: unknown
log: longshot_log.txt
pid: 13904
started: 20261116T205323Z
signal: 0
signal_code: 0
exception: 0x0
fault_address_class: null
host_backtrace_method: Windows unwind

The crash report attachment says the full logs/longshot_log.txt was not attached. No rc.5 result or log was provided.

Still needed

Full logs/longshot_log.txt, exact reproduction steps and what appears on screen, whether it happens on every launch, whether v3.0.0-rc.5 also crashes (and its log/report if so), and which older builds worked if known.

Attachment

Crash report message.txt from the conversation (contains the crash record above).

Reported on [Discord](https://discord.com/channels/1569849200047427645/1571994344230101012/1571994344230101012) ### Longshot version nightly-363-0cab3548 (source 3b447cf5). Reporter says older builds used to work. ### Channel Nightly. ### What happened Longshot closed unexpectedly during startup after running for 39 seconds. Crash report identifies a graphics-driver failure; the crash record says `device lost`. Four crash/unclean-end records were recorded in the last day. The reporter asks whether this is related to #1040, which covers a Radeon 860M crash in Kokiri Forest; this report is a startup crash on a different GPU. Reproduction steps, on-screen behavior, and whether every launch fails are unknown. ### Graphics card and driver version AMD Radeon(TM) 880M Graphics, AMD proprietary driver 26.11.1 (LLPC). ### CPU, RAM and operating system AMD Ryzen AI 9 365 w/ Radeon 880M (20 threads), 30 GB RAM, Windows 11 (build 26300). ### Changes from the defaults Crash report lists `quality_preset=6`, `resolution=4`, `debug_tools=1`, and `memory_budget=1`; most other listed settings are disabled/default. `LONGSHOT_NO_UPDATE_CHECK=1`. ### Crash record ```text Longshot crash record 1 version: nightly-363-0cab3548 (source 3b447cf5) run_id: 1a768200d771585cdb80ba16651486da when: 20261116T205403Z uptime_s: 39 reason: device lost thread: host module_offset: longshot.exe+0x1c254bab last_present_age_ms: unknown log: longshot_log.txt pid: 13904 started: 20261116T205323Z signal: 0 signal_code: 0 exception: 0x0 fault_address_class: null host_backtrace_method: Windows unwind ``` The crash report attachment says the full `logs/longshot_log.txt` was not attached. No rc.5 result or log was provided. ### Still needed Full `logs/longshot_log.txt`, exact reproduction steps and what appears on screen, whether it happens on every launch, whether v3.0.0-rc.5 also crashes (and its log/report if so), and which older builds worked if known. ### Attachment Crash report `message.txt` from the conversation (contains the crash record above).
Author
Collaborator

Files from the Discord report:

Files from the Discord report: - [message.txt](https://git.magicbean.dev/attachments/01ced772-5021-4a02-98ae-c1d11dc17ee3)
Author
Collaborator

From Discord:

Reporter clarified on Discord that the device-lost crash occurs in Kokiri Forest. The full game log and rc.5 result are still missing. Related known fix: PR #489 addresses an AMD desktop Ryzen iGPU ASTC device-lost failure in Kokiri Forest, and is included since v3.0.0-rc.5; the reporter’s Radeon 880M/APU is different hardware, so applicability is unconfirmed.

From [Discord](https://discord.com/channels/1569849200047427645/1571994344230101012/1571994344230101012): Reporter clarified on Discord that the device-lost crash occurs in Kokiri Forest. The full game log and rc.5 result are still missing. Related known fix: PR #489 addresses an AMD desktop Ryzen iGPU ASTC device-lost failure in Kokiri Forest, and is included since v3.0.0-rc.5; the reporter’s Radeon 880M/APU is different hardware, so applicability is unconfirmed.
Author
Collaborator

From Discord:

New Discord attachment provides a full longshot_log.txt from v3.0.0-rc.5 (source 013b1d3a), confirming the crash also occurs on rc.5 in Kokiri Forest. The log header reports output resolution 2048x1152, graphics scale 1; it mounts content from the v3.0.0-rc.4 build. Near the end it records a 450.5 ms slow submit and GPU waits; the provided excerpt does not include a clear device-lost line or full crash record. Attachment: message.txt from Discord message 1571996063949262908. Still unknown: exact reproduction steps, on-screen behavior, whether every attempt fails, the game state/action immediately before failure, and which older builds worked.

From [Discord](https://discord.com/channels/1569849200047427645/1571994344230101012/1571994344230101012): New Discord attachment provides a full `longshot_log.txt` from v3.0.0-rc.5 (source 013b1d3a), confirming the crash also occurs on rc.5 in Kokiri Forest. The log header reports output resolution 2048x1152, graphics scale 1; it mounts content from the v3.0.0-rc.4 build. Near the end it records a 450.5 ms slow submit and GPU waits; the provided excerpt does not include a clear device-lost line or full crash record. Attachment: `message.txt` from Discord message 1571996063949262908. Still unknown: exact reproduction steps, on-screen behavior, whether every attempt fails, the game state/action immediately before failure, and which older builds worked.
Author
Collaborator

From Discord:

reports an 860M device-lost crash in Kokiri Forest (#1040) and suggests testing RX_VK_ASTC_CPU=1, RX_VK_VRAM_MB=6144, and HOS_SHORT_SLEEP_TIMER=1 in longshot_env.txt on this 880M system. The suggestions and rationale are unverified for the 880M. PR #1114 is open; it proposes defaulting the short sleep timer on Windows integrated GPUs for performance, not a confirmed device-lost fix.

From [Discord](https://discord.com/channels/1569849200047427645/1571994344230101012/1572059593280987248): <name> reports an 860M device-lost crash in Kokiri Forest (#1040) and suggests testing `RX_VK_ASTC_CPU=1`, `RX_VK_VRAM_MB=6144`, and `HOS_SHORT_SLEEP_TIMER=1` in `longshot_env.txt` on this 880M system. The suggestions and rationale are unverified for the 880M. PR #1114 is open; it proposes defaulting the short sleep timer on Windows integrated GPUs for performance, not a confirmed device-lost fix.
Sign in to join this conversation.
No milestone
No project
No assignees
1 participant
Notifications
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/Longshot#1234
No description provided.