[UE 5.8 , 5.8.1, 5.8.2] [Stereo render in ps5] Stereo render in ps5 its wrong

General Issues and Bug Reporting (Unreal Engine)
,
1

Summary

Stereo render in ps5 its wrong, look like a bad uv sphere mapped and no refresh, maybe Fireant plugin problem or UE, work fine in 5.7, look like freeze screen, i got this in console : i get this error in console LogRHI: Warning: Auto Agc::suspendPoint() due to submission timeout. LogRHI: Warning: Agc::suspendPoint() gap exceeded 1.00 second threshold: 4.00 seconds LogSony: Warning: GPU timeout: payload on the [Graphics] queue has not completed after 5 seconds. LogSony: Warning: GPU timeout: payload on the [Compute] queue has not completed after 5 seconds.

What Type of Bug are you experiencing?

Platform Console

Steps to Reproduce

just make vr template project enable fireant and launch in ps5

Expected Result

good stereo render as allways

Observed Result

stereo render its wrong, look like freeze

Affects Versions

5.8

Platform(s)

Windows

Upload an image

ps5 freeze.png
ps5 freeze.png4000×2040 8.77 MB

Additional Notes

i think is a freeze problem, and sometimes work fine, but if restart game usually start freeze again, no problem with no vr app

2 Likes
3

The status of UE-394862 changed to ‘Needs Triage’. We are routing this to the appropriate team for investigation.

1 Like
4

I re-tested on 5.8.2 with a plain non-VR project (no FireAnt, VR disabled) and it works fine — no crash on exit, no GPU timeout on relaunch, no freeze.

  1. Non-VR project: works correctly, no crash, no freeze.
  2. VR project with FireAnt enabled: same freeze/GPU timeout issue as before (Agc::suspendPoint() gap exceeded, GPU timeout... 5 seconds).
1 Like
5

Root cause found — community fix confirmed working

A fellow developer (Ekaergos, via the forum thread) identified the fix:

In Engine/Platforms/PS5/Source/Runtime/AgcRHI/Private/AgcViewport.cpp, line 697, change:

cpp

RHICmdList.ImmediateFlush(EImmediateFlushType::DispatchToRHIThread);

to:

cpp

RHICmdList.ImmediateFlush(EImmediateFlushType::FlushRHIThread);

I’ve tested this on 5.8.2 with VR/FireAnt enabled and it resolves the issue completely — no more crash on exit, no GPU submission timeout on relaunch, no frozen-frame/VR distortion.

Why this matches the observed symptoms: DispatchToRHIThread only queues the flush on the RHI thread without waiting for it to complete, while FlushRHIThread blocks until the RHI thread’s command queue is actually drained. If viewport teardown/present on PS5 relies on a completed flush before releasing or resetting GPU-side state, using Dispatch instead of a real Flush would leave outstanding GPU work at shutdown — matching the Agc::suspendPoint() gap exceeded / GPU timeout warnings seen on the next launch, and the subsequent frozen-frame reprojection artifact in VR.

For reference, in 5.8.2 the issue only reproduces with VR/FireAnt enabled (a plain non-VR project works fine), which is consistent with this being specific to the PS5 viewport present/flush path used by the VR compositor path.

Please consider including this fix in an upcoming patch. Happy to provide additional logs or repro details if useful.

1 Like