MAXtoA GPU still errors out on second render attempt

[ACCCC]

Steps to reproduce:

1. Empty Scene

2. Create a Physical Material + Bitmap Lookup node with a random texture

3. And/Or add a HDRI Environment and a environment map

4. Run either Activeshade or Renderview IPR then close Activeshade or stop the IPR

5. Re-run Activeshade/IPR

I've discussed a similar issue with Mads a couple of months back where the background of the HDRI just disappeared and i think he created a ticket for this and it got fixed fairly quickly.

I'm not sure if this error on second render was in the bug report back then but this still errors out 100% of the time for me, sometimes it even creates a critical cuda error and i have to restart max, i'm not sure if there is already a ticket for this.

Any ETA for this? It's kinda annoying since those OSL maps are kinda important for lookdev via GPU.

If this only happens on my machine and yours renders fine let me know and I'll add some debug logs.

[madsd]

I can confirm this with a newer build.
Will pop a note about it now.

[madsd]

Just tried to ping the texture loader with a minimum code node.

Returns same issue.

[madsd]

This was working yesterday...
Why is it broken now.
Will check.

Just a FYI, I updated to the latest Studio drivers yesterday.

456.71

So might investigate to roll back on that driver number.

[ACCCC]

Thanks for the help, did you run this on 4.1.2.1 or a newer beta build?

I have only encountered this with maps that take a file as a input so far, i'm curious if this affects all OSL shading operations or only those that use a external file.

[ACCCC]

I'm currently running 456.38 and MaxToA 4.1.2.1 and i'm pretty sure i've encountered this issue with 4.1.0.71 in combination with 456.38 and 452.06.

Are you sure this was working on your machine? What GPU are you using?

I'm considering to DDU my machine.

[madsd]

Its not driver related, I can see they got a ticket for something related already.
Dev responded elsewhere, so on it.

[madsd]

Also, its been fairly long time since I loaded any textures and ran the GPU, so it could have hung without me detecting it while working on OSL shaders.

Just been working full procedural for a while, so never hit this bug.

[madsd]

Slightly newer builds tested here.

Yeah I would not worry about any non texture fetching stuff.
Ive run the OSL on GPU daily and runs OK. No drop outs, but only because I work 100% procedural.

[ACCCC]

Yeah my workflow relies heavily on image textures so in order to render GPU i'm currently restricted to either use the old Autodesk Bitmap node which doesn't really offer a UVW input or Arnold Image which doesn't work in viewport.

[DeclanRussell · Accepted solution]

Hey! Thanks for reporting. We have a fix for this coming up in the next release.

[ACCCC]

Hey, yeah Stephen already mentioned this a couple of days ago.

Thanks alot! Keep up the good work.