Game: Crash during Great Deku Tree intro fade to black on nightly-486 (Intel GPU) #1731
Labels
No labels
area/accessibility
area/audio
area/input
area/install
area/menu
area/renderer
area/saves
area/tooling
area/upscaling
beanstalk: approved
beanstalk: failed
beanstalk: needs approval
beanstalk: running
beanstalk: tested
ci: approved
good first issue
gpu/amd
gpu/igpu
gpu/intel
gpu/nvidia
help wanted
kind
bug
kind
chore
kind
docs
kind
feature
kind
performance
kind
question
needs measurements
platform/linux
platform/macos
platform/steam-deck
platform/windows
preview
priority
critical
priority
high
priority
low
regression
risk
high
risk
low
risk
medium
status
blocked
status
confirmed
status
duplicate
status
invalid
status
needs-info
status
wontfix
ci: approved
No milestone
No project
No assignees
2 participants
Notifications
Due date
No due date set.
Dependencies
No dependencies set
Reference
MagicBeanDev/Longshot#1731
Loading…
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Reported on Discord
Longshot version and channel
nightly-486-09a20cd6 (nightly)
What happened and steps
The reporter says Longshot crashes during the title intro cutscene, when Navi is sent to Link by the Great Deku Tree, during the fade to black. They observed VRAM use around 4.5 GB before the crash and believed the VRAM limit was exceeded. Setting
RX_VK_VRAM_MB=11264did not prevent the crash. They prefer the very low frame rate they experienced before over being unable to get past the cutscene.The crash occurs every time, when Navi and the Great Deku Tree begin talking about Link or during the transition into the cutscene. Dropping all resolutions slightly improved the progress: they could hear the music in Link's tree house, but the display remained black and the game still crashed, after FPS dips then settling around 30 FPS. The
longshot_env.txtmodification did not help and the file has since been removed. Exact graphics-setting comparison beyond lowering resolutions is unknown.Where in the game
Great Deku Tree intro cutscene, at Navi's delivery to Link / fade to black and transition to Link's tree house.
GPU and driver
Integrated Intel UHD graphics, driver version 32.0.101.7088. The reporter describes it as having 8 GB VRAM; the log identifies Intel XeSS 2.0.2.
CPU, RAM and operating system
11th Gen Intel Core i3-1115G4 @ 3.00 GHz, 16 GB RAM, Windows 11.
Graphics settings and environment override
Logs show output resolution 2048x1152, graphics scale 1, Intel XeSS 2.0.2, and frame rate cap 30. In the runs with these settings, Longshot reports VRAM around 4.8 GB used of 7.2 GB allowed (e.g. 4938 MB used), not an exceeded budget. Lowering all resolutions slightly advanced playback but did not avoid the black screen/crash. The reporter says
RX_VK_VRAM_MB=11264did not help and has removedlongshot_env.txt. Other settings are unknown.Additional logs
The reporter supplied five more
longshot_log.txtfiles from nightly-486. They document multiple restarts and do not show the same crash trace in the excerpts; one has a logged Intel driver fault while compiling a graphics pipeline:Other excerpts show the game running around 22–30 FPS and VRAM below the reported 7.2 GB allowed budget. These additional logs are attached to the issue.
Original crash log
The initial attached
longshot_log.txtends with an access violation in Intel'sigvk64.dllon the Presentation Thread:Still needed
Effects of changing upscaling or other graphics settings; confirmation of actual available/shared VRAM capacity if relevant.
Related reports
This is not established as the same issue as #668 (RTX 5050 hard hang in Castle Town) or #837 (GTX 1050 Ti VRAM pressure/freezes); this report is an Intel driver access violation during the Great Deku Tree intro. PR #1291 handles Windows commit-limit allocation failures, but these later logs show VRAM below Longshot's allowed budget and do not establish that fix as applicable. PR #1685 concerns NVN device mode getters and does not address this game crash.
Files from the Discord report:
Updated with more from Discord: Added the reported Intel UHD integrated GPU, Intel i3, 16 GB RAM, and Windows 11 details; driver and reproduction details remain unknown.
Updated with more from Discord: Added the exact CPU and Intel graphics driver version supplied by the reporter.
From Discord:
Reporter confirms the crash occurs every time: when Navi and the Great Deku Tree begin talking about Link, or during the transition into their cutscene. They report nearly 30 stable FPS during the intro now, versus about 3 FPS previously, but cannot skip or get past the cutscene. Exact override confirmation and effects of graphics-setting changes remain unknown. Discord follow-up: https://discord.com/channels/1569849200047427645/1572769730903941234/1572930288261140643
From Discord:
Reporter says the
longshot_env.txtmodification did not change the crash and the file is currently removed. This confirms the VRAM override did not help; no new graphics-settings comparison was provided.Updated with more from Discord: Added that lowering all resolutions slightly advances the intro but leaves a black screen and eventual crash, with dips before stabilizing at 30 FPS.
Updated with more from Discord: Added five nightly-486 logs; they show VRAM below budget and include a separate Intel driver pipeline-compilation fault, not a confirmed reproduction of the original access violation.
Files:
From Discord:
Reporter notes that at the moment of the crash, the on-screen VRAM meter's cap drops and displayed usage then exceeds that limit before crashing. They can provide a real-time video if useful. Existing logs sampled earlier show usage below budget, so a recording capturing the meter immediately before failure may clarify this discrepancy.
Claimed: investigating the repeated Intel UHD / driver 101.7088 intro access violation at igvk64.dll+0x87793. Worklist score is S0 × 3.5% reach × 0.7 confidence ≈ 0.20; #1676 reports the same driver offset. I’ll build current main from the local dump and test on RX 9070 XT; Intel-specific reproduction still needs an Intel tester. Prior work: #404 addressed a different over-budget dispatch crash, not established by these logs.