OctaneRender 2020 for DAZ Studio [TEST and STABLE] Tweet

[face_off]

At the time I was using Texture Shaded view with cameras switched in between with Perspective.

OK, thanks. For testing, may I suggest recording your OutOfCore settings, and also have Windows Task Manager open to monitor your memory. It would be good to see if the crash occurs when you have a lot of free VRAM and system RAM available, and if it only happens if OutOfCore is enabled or disabled.

Paul

[abayliss]

However, after using this work around, and avoiding the crash, I have also started seeing CUDA 700 errors randomly occurring, requiring me to restart DAZ.

Can you post the specific Octane Log when this happens please?

Paul

I have attached that portion of the log.

[mikmod]

At the time I was using Texture Shaded view with cameras switched in between with Perspective.

OK, thanks. For testing, may I suggest recording your OutOfCore settings, and also have Windows Task Manager open to monitor your memory. It would be good to see if the crash occurs when you have a lot of free VRAM and system RAM available, and if it only happens if OutOfCore is enabled or disabled.

Paul

OoC settings I have set at 28928MB out of 64GB I have. Kind of overkill, but I'm curious about the scenes that exceed measly 8GB VRAM Nvidia puts in, well, high end cards after all. But that scene wasn't exceeding the VRAM capacity, it was about 75% full at the time. Example on the config/status screenshots has a bit lighter mesh, but still quite a bit of data. DAZ updated the beta to 4.12.1.117 in the meantime, and these screens come from it.

AVX-512 issue with Iray still isn't fixed.

I'll attach the screenshot of most custom plugins I have as well.

Ah, when "combined meshes" option is on, in Final Render your "Realistic Hair (Brown)" shader (from your library) becomes white on hair strands made by this relatively new Daz plugin integrated into the program.

[face_off]

OoC settings I have set at 28928MB out of 64GB I have.

I suggest reducing this, or turning it off, to see if it resolves your crashes.

Paul

[rajib]

I have updated the installers at the top of this thread with:

2020.1.0.55
- Compiled with Octane 2020.1

Paul

Thanks Paul. Just to let you know the problem with all version of the 2020 Beta and RC that leads to instant crash on render start of old scene is still there in the final release. The workaround of rendering something first - a cube / plane etc and then loading the old scene and render works. But please fix this.

I also was plagued with the CUDA 700 errors recently with the 2080 Ti RTX but interestingly ever since I started using the Titan RTX I don't get them anymore. I guess it is tied to the OOC memory usage that generates the CUDA errors. With the Titan RTX I never get into the OOC memory because of it's 24GB video ram, so no errors.

Update:
I also don't see the smart unfold. Please enable it. Staying silent even and not enabling even after multiple request have been made is not appreciated.

Update 2 :
Spoke too soon on the CUDA error. Got my first CUDA 700 error whlie using the Titan RTX. It happened during modification of scene object while rendering was going on. I guess it causes some pointer issues. Anyway, now I tend to pause render before I make changes to the scene.

[SiliconAya]

I have another scene the causes more or less the same crash.

Code: Select all

DAZStudio.exe caused ACCESS_VIOLATION in module "C:\Program Files\DAZ 3D\DAZStudio4\octane.dll" at 0033:000000004D52C740, Octane::isGreyscalePass()+5684448 byte(s)


The same workaround of opening the viewport before loading the scene still works.

[rajib]

Hi Paul,

Please take a look at viewtopic.php?f=44&t=72407&p=382265#p382263. I am listing all the issues I am finding with the current latest build of 2020.1. Hope you can resolve them. I will add to the same post if I encounter something else.

Also take a look at viewtopic.php?f=44&t=72407&p=382265#p382265 and see if you can add VOB and other Octane specific features that are missing from the Daz plugin via a placeholder method. I have detailed one possible solution.

Regards,
Rajib

[wader]

I have another scene the causes more or less the same crash.

Code: Select all

DAZStudio.exe caused ACCESS_VIOLATION in module "C:\Program Files\DAZ 3D\DAZStudio4\octane.dll" at 0033:000000004D52C740, Octane::isGreyscalePass()+5684448 byte(s)


The same workaround of opening the viewport before loading the scene still works.

I had forgotten about this, but the crash situation still happens with a small number of my Daz scenes made in 2018.

Re-saving those scenes does not remove the need to render a different scene in Daz before loading the problem one for crash-free rendering.

It's as if some artifact or value in the scene file is causing the newer Octane to perform an operation it doesn't expect. Unfortunately, I don't have a simple scene to share in that regard for testing, the problem ones each involve multiple purchased assets.

[wader]

Hi Paul / @face_off ,
Not sure where else to report this, but since upgrading to the latest version of Daz Studio ( 4.12.1.117 ), Daz will not shut down after closing the its UI - it can be seen in Task Manager with the same memory allocation and no Processor cycles, sitting indefinitely until force-closed.

Some quick experimenting revealed that removing the Octane plugin relieves this problem, but installing any recent version of Octane pluging - e.g., 2019 Enterprise, 2020 Enterprise - causes Daz to no longer quit properly.

So, while this may be due to a change in the latest Daz product, it's apparently Octane being the only cause that I can find thus far. Not sure how to help provide more information to diagnose the cause, unless you can recreate.

[mikmod]

I have updated the installers at the top of this thread with:

2020.1.0.55
- Compiled with Octane 2020.1

Paul

Thanks Paul. Just to let you know the problem with all version of the 2020 Beta and RC that leads to instant crash on render start of old scene is still there in the final release. The workaround of rendering something first - a cube / plane etc and then loading the old scene and render works. But please fix this.

I also was plagued with the CUDA 700 errors recently with the 2080 Ti RTX but interestingly ever since I started using the Titan RTX I don't get them anymore. I guess it is tied to the OOC memory usage that generates the CUDA errors. With the Titan RTX I never get into the OOC memory because of it's 24GB video ram, so no errors.

Update:
I also don't see the smart unfold. Please enable it. Staying silent even and not enabling even after multiple request have been made is not appreciated.

Update 2 :
Spoke too soon on the CUDA error. Got my first CUDA 700 error whlie using the Titan RTX. It happened during modification of scene object while rendering was going on. I guess it causes some pointer issues. Anyway, now I tend to pause render before I make changes to the scene.

I've got this CUDA 700 error just at the end of preview render (three primitive cubes with some snake-skin textures, 200 samples left out of 1000, path tracing, 1400x1500 res) on RTX 2080, Win 7 x64, 445.87 driver, when I enabled AI upsampling as "Upsampler mode 2x2" from the droplist. Rendering never ended and the whole plugin just hung, when I invoked "Rebuild scene", and Daz Studio with it. All this on "they-made-it-stable-in-the-meantime" DS 4.12.1.117.

I have the feeling Nvidia broke something in the almost whole 442-445 driver lineup, when it comes to Optix/neuray/whatnot, cause even their Iray 2019 was falling back to CPU in the Daz Studio betas, that used such Iray. All cause of sudden "memory allocation errors" in the Daz log during switching textures/shaders/material properties in the scene and using Iray preview at the same time. Daz guys fixed some of these bugs, and the program fell back to CPU mode less frequently, but it's unclear, how much they had to improvise to get around it. Maybe something in the drivers also bugs Octane as well?
I could back then (and still can) game on my card with full VRAM filled up with textures, so the card's memory chips aren't at fault, I think.

Oho:

Code: Select all

18:40:59.978: +++++ Plugin running.
19:23:38.976: Up-sampling error: CUDA_ERROR_LAUNCH_FAILED: unspecified launch failure
19:23:38.977: CUDA error 719 on device 0: unspecified launch failure
19:23:38.977:   -> failed to copy pitched memory from device
19:23:38.977: Render engine failure occured! (WARNING)