We are seeing this as well. Shows up in Win64 Test builds for us but goes away when ran in Development or if PIX is attached. Has also not been seen on consoles.
We are on 5.6.1 and, like mentioned above, can see it with lumen disabled. So not a lumen issue.
We also continue to repro regardless of the value of r.TranslucencyVolumeBlur or r.TranslucencyLightingVolume.Temporal
Setting r.TranslucencyLightingVolume.Batch=0 stops the bug for us, though I don’t want to have to run with that disabled as the commit message where this was introduced suggests this change was a huge perf win.
Minor note: In our testing, there is definitely a correlation between the prevalence of the flashing and the window being in focus or not. If alt-tabbed away, we see the flickering much more often.
Thanks for pointing out those other case. Using the repro project from the Niagara smoke I was able to reproduce the flickering by adding a second directional light and translucency volume settings with large grid sizes and am looking into this. Does this setup sound like your repro case - does the flickering only occur for you when multiple directional lights are present?
I just wanted to post a reminder that Epic will be on holiday break starting next week (Dec 22, 2025) and ending Jan 5, 2026 and there will be no responses from Epic on public or private tickets, though you may receive replies from the community on public tickets.
We have also ran into this issue in 5.5/5.6/5.7 and after finding this thread we tried to do a bit of debugging and it appears to be caused if a directional light is set to affect translucent lighting.
We tried a bunch of different things and we found that when we removed *all* shadow casting lights, the issue went away. Than we slowly starting adding them back and we found that adding a single shadow casting spot light re-introduced the issue right away.
So than we loaded up 5.6.1 from EGS and made a sample project to test our theories and low and below in 5.6.1 we get some major flickering. There are a few different cases like it happens worse in PIE, and on editor load.
I have uploaded a bunch of videos of the different cases here:
Thanks for providing the sample project. What GPU and driver version were you using to reproduce the flicker? Unfortunately, I wasn’t able to see the flickering when I opened the sample in UE 5.6.1 on an A6000 in Editor like in the videos so I suspect I may need to swap GPUs to repro this.
We are using 3070 ti as our most common config where we are seeing this problem, as far as the actual drivers go, we are using a variety of drivers, but for example, these videos were captured on a 7950X PC with an nvidia 4070 with 581.8 drivers that our technical art director is using at the moment but i even tried with latest drivers and issue still exists
(You can open the LightFlickerRepro.log inside the shared folder to check the logs to confirm this as we shared it with the logs)
I spent some time trying to repro with the provided project on a 4070 Super with vanilla UE 5.6.1 (and a local build) 581.8 and dual monitors with 1440p resolutions like I saw in the logs, but the only time a saw a flicker was once briefly when ALT+TABing between the editor and another app, I wasn’t able to reproduce any of the flickering seen in the videos when loading the editor, traversing the map, entering/exiting PIE.
I did notice in the repro that the animated spotlight in the blueprint seems to intersect the floor in an odd way, you may also want to verify the flicker isn’t related to this.
[Image Removed]
Is your studio using a shared DDC? If so, does this issue reproduce for you when forcing a local shader compilation with -noshaderddc or -DDC-ForceMemoryCache? You may also need to clear the local PSO cache to recompile those shaders with -clearPSODriverCache.
We’ve been battling this issue on and off for a while, and have seen it in a couple of different projects (both 5.6). As others pointed out, shadow casting lights and other light modulation functions injected into the TLV seem to be the culprit, and manifest as particles cross its boundaries.
A recent modification we did that solved a big number of cases was switching the TLV to temporal filtering, and using a weight of about 0.95, the responsiveness trade-off was very acceptable for our use case, and it completely removed this artifact in a lot of our remaining cases.
If you’re seeing this artifact, give it a try. You can use something like:
I nuked our entire zen cache and it seemed to have gone away so [mention removed] was probably correctly, but I’m a bit confused why we were seeing that issue on a blank project
[mention removed], I think it’d be unlikely that a new project would pull the shader for a blank project, but if the engine version was the same, same shader key, same source etc so that the hashes matched, it’s theoretically possible that the blank project pulled the shaders - in which case, if you still have that DDC somewhere it could provide clues if we identify some hash collision or some shader compilation error that was fixed by recompiling the global translucency shaders.
[mention removed] I was briefly able to replicate the issue with 2 directional lights once a while ago, but haven’t since upgrading drivers/recompiling shaders. Are you compiling the shaders for your package builds fresh every time or are shaders ending up in the package from an existing DDC?
[mention removed] I’m not really sure than, on our main project it seems to kind of mostly dissapear, so but the project we shared was on an empty new project separate from everything so there must be still some issues somewhere
In our local testing it didn’t resolve the issue when disabled, but would be good to hear if it has any affect in your repros. To test, turn off r.TranslucencyLightingVolume.UseTypedUAVLoad while the flickering is happening.
We found a solution that seems to be working so far for our case (our game has a worst-case-scenario for the VSM/TLV interaction: very large open world, sun constantly moving, lots of transparent / translucent particles).
First, I believe at least part of the issue is that the TLV is sampling sections of the VSM that are falling back to coarser pages or in the boundaries of clipmaps, so given the cache churn caused by directional lights such as the sun at certain angles, we have temporal instabilities that we then see as flickering. I wrote a filter that tries to detect these clipmap discrepancies and cushions the change. It works like this:
Keep a history of the last clipmap and shadow term applied to each injected TLV cell.
On light injection, see if the clipmap sampled by the cell is the same as the past one (see FVirtualShadowMapSampleResult.ClipmapOrMipLevel).
If the clipmaps differ, do a lerp between the last applied term and this one. The alpha is a function of the clipmap difference and whether the new sample is darker or lighter (because we often see this as a sudden bright flick, we are faster to adopt darker terms than bright ones)
Save the new term and clipmap.
This produces very good results for our case, but I literally just added it today. I’ll wait for a couple of days for people to try it out more extensively and come back to report
Just adding some more notes from our testing. It’s insanely annoying to repro this so not a ton of new info.
We thought we had some captures in RenderDoc that showed the flicker but I couldn’t capture any more since the morning so I’m now not sure if we even caught a bad frame or it was some other noise in our visuals.
What seems to be consistent however is the problem reacting to poor frame times. When our GPU time is high enough the TLV seems to be incorrect, then dropping the frame time (via screenpercentage) it is correct again. There is a sweet spot between 33 and 16ms where it flickers, for my tests around 28-25ms has been the most violent flashing.
Switching AA method to TAA from TSR fixes the issue.
Chaning r.TSR.AsyncCompute cvar also changes the frequency of the flicker with 3 being the most rapid one (smaller bursts of flicker rather than a slower on/off )
Looks like our alleviation is working fine. I also tried it with the project files shared by Branislav in this thread. Here is a video of the flickering (which I manage to repro on a vanilla distribution of UE 5.7 by placing that big surface at the fringes of the TLV), and the filter as we turn it on and off.
