Game: Frame pacing becomes uneven with frame generation in Jabu-Jabu's Belly #1568
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#1568
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-480-d1a307cb (source b4a964e0), nightly.
What happened and steps
Frame pacing is uneven in a room in InsideJabuJabuBelly (child), with consistent low 1% lows despite the room appearing simple. The reporter compared otherwise identical settings and found that disabling frame generation smooths frametime; enabling frame generation at 2x brings the uneven frametime back. Frequency beyond “consistent” is unknown. The screenshot shows Link in a dim cavern room facing a barred tunnel.
Where in the game
InsideJabuJabuBelly (child), Jabu-Jabu's Belly.
Graphics card and driver version
NVIDIA GeForce RTX 5080, NVIDIA 620.70.06.
CPU, RAM and operating system
AMD Ryzen 7 9800X3D (16 threads), 30 GB RAM, CachyOS Linux 7.3.3-1-cachyos, GNOME/Wayland.
Graphics settings and comparison
The supplied report lists frame generation 2x, NVIDIA Reflex On + Boost, fullscreen, HDR Native and other changes from defaults; DLSS with native/DLAA rendering at 3840x2160, Docked profile, frame pacing On. The reporter confirms all settings were identical between the comparison except frame generation. Disabling frame generation smooths frametime; repeatability/count of comparison runs is unknown.
Performance measurements
The reporter says the uploaded performance data followed the profiling guidelines and cannot provide more data. The log confirms
HOS_FPS_LOG=1,HOS_THREAD_TIMES=5,RX_VK_PASS_TIMING=1, andRX_VK_SLOW_LOG=50were enabled. However, the supplied log does not contain captured thread-time or GPU pass-timing samples identifying the affected room; the report records the active room as InsideJabuJabuBelly, but not a 60–90 second timed capture there. Please inspect the attached logs for any usable measurements before further diagnosis.Logs and screenshot
The original report folder contains
report.txt,longshot_log.txt,longshot_log.1.txt, andpicture.png; the screenshot has no visible error text. (Attachments copied from the Discord report.)Still unknown
Whether frame generation's effect is repeatable across multiple A/B runs, how frequently the issue occurs in ordinary play, and measured frametime/thread/GPU timings during the affected scene. The reporter says they cannot provide additional performance data.
Files from the Discord report:
Picking up the validation setup for the Linux/Wayland report. I’ll check the available evidence and scope a matched game build, scene and framegen Off versus 2× comparison with a suitable tester. The separate local Windows RTX 4070 capture will cover the outstanding FIFO latency question.
I checked both attached Linux logs. They contain 2× frame-generation activity and GPU pass samples after entry into Jabu-Jabu’s Belly. The logs are from different nightlies and do not include a labeled FG-off run in the same affected room, so they cannot isolate the cadence change.
The next check is a matched Off versus 2× run on one pinned build and the same room, camera route and settings, preferably RTX 50-series on Wayland. Record the effective present mode, base FPS and evidence that generated frames were actually submitted. A 60–90 s capture per arm would give us a comparable sample.