Distance field lighting related crash

I have the following in my level:

  1. enabled DFGI, HFGI, DFAO and RTSF with default settings and disabled all static lighting.
  2. a movable skylight and directional light.
  3. a 505x505 terrain with 2 height blended layers.
  4. a tree static mesh about ~20k polys, imported with vertex colors, a simple vertex blend material with 6 4k textures, mesh DF resolution set to 1.
  5. 100 of these trees placed with the foliage brush, at x10-x30 uniform scale. (about 4,000 world units)

Now once I enable “affect distance field lighting” in the foliage settings, the driver and subsequently the editor crashes. I thought it was due to VRAM running out, but it lets me place a maximum of about 40 trees, with this enabled, where only 1.8GB of VRAM is consumed. Before I place the trees vram usage is at 750MB. I have substracted the vram amount taken up by the OS (about 400MB). So I still have 2GB to spare. I tried taking the mesh DF resolution way down and decreasing the foliage scale, also extended the TdrDelay from the registry but doesn’t seem to help. Is this normal? What are the best practices to stay within 1gb for distance field texture usage.

Sometimes this happens and I have to restart the editor, I would start a new topic but cant reproduce it and could be related:

Black slowly fills the world but only the part thats in my screen.

I am running 4.11 preview 7, win10x64, my graphics card is R9 290 4GB, 16gb ram, i7 3770K CPU. I have the latest driver (15.30.1025.1001).

EDIT: In case someone is having a similar problem, r.VPLMeshGlobalIllumination=0 stopped the crashes for me.

Hi hurt720 -

Can you post the Machine ID for the crash that is occurring here? Also, please make sure that you are submitting (sending) your crash when it occurs.

Thank You

Eric Ketchum

Hi Eric,
Unfortunately it doesn’t present me with a “send crash” form, but with the following popup: “Video driver crashed and was reset! Make sure your video drivers are up to date. Exiting…”. I understand its the amd driver that’s crashing and may not be the engine’s fault, but I’ve never had problems like these when pushing the gpu in other packages/engines. Here’s my log: link text

EDIT: installed the 16.1.3 hotfix, unfortunately it didn’t make a difference. Is there something else I can send you? There are no critical or error events in the event viewer at the time of the crash.

Hey hurt720,

The error message you are receiving is actually a known and reported issue (UE-21389), and deals directly with the AMD R9 200 and 300 series cards. This issue is not something we can resolve on our end, and we have reached out to AMD to issue a patch to help resolve the problem. Unfortunately, we are now simply waiting on them to respond and issue a driver update for these cards.

We appreciate your patience while this issue gets resolved and hopefully AMD issues a driver update soon that fixes this problem. Let me know if you have further questions or need additional assistance.

Cheers,

Thanks, should have done more research before posting… Sorry to bother you but can you point me to the known issues list for the release branch, because I can’t seem to find it.

No need for apologies. The more reports we get, the more verbose our issue becomes, and the more attention it receives. The known issues is actually not public as it is an internal list in our bug reporting system, but we are working to get this changed so the public can see issues like these and track them.

With that in mind, we appreciate you taking the time to write up the report and let us know if you need further assistance.

Cheers,