Corona Renderer 4 for 3ds Max - Daily Builds Discussion

[Christa Noel]

    Fixed saturation -1 not resulting in grayscale color (both in VFB and ColorCorrectTex)
    Fixed error when loading LightMix CONF files

great! thanks

[arqrenderz]

I just saw this line in the Trello for Corona 4 bugfixes: "Fix CoronaProxy bad performance, memory usage in some situations"

Does "in some situations" have anything to do with loading proxies over 10Gb Ethernet? We've been troubleshooting a slow to open max file where removing the proxies (held on a server via 10GBe) seems to remedy the problem.

Sorry to bump this but I feel that it got snowed under by other posts.


What we saw is the more proxys you have, the ram situation is worse and worse, EVEN with proxys hidden on a new open scene, corona proxys will eat the ram, so the option is to not have proxys and a 4 gb max file, or a 200mb max file with a ton of ram.
(the proxys eat ram even when cache on ram is off)

[bhiggins]

Good morning,

We are getting the error "ending sampling mask to slaves timed out. Try to increase synchronization time or decrease maximum pixels per transfer" when rendering with DR. We believe this is causing corona to crash and has been a pain to deal with for the last couple months.

Is there something we are missing or can do to mitigate this issue? We've tried reducing max pixels per transfer but are still experiencing crashes.

We are on the current 5/3 daily build. This issue has plagued us from Corona 3 hotfix 2.

I've submitted a minidump from the crashed instance of 3dsmax that was attempting to render to the farm.

[ihabkal]

I've had this issue for the latest 2 builds, when i render using DR, the frame buffer Beauty pass will go black randomly after a few minutes, then if I cycle between the elements and back to beauty it will show up again, then go black after some minutes. When render ends it is randomly either saving or not saving the beauty (black layer) depending on if it shows or not at the time. In some cases the other elements also would have the alpha wrong as in a big blag square of the element missing or added. Anyone having same issue? could it be a memory leak when it receives data from the render nodes?

+1 exactly the same issue. Haven’t had any loss when saving out (to Cxr) but the black VFB during DR is identical. I agree that it happens with the DR machines send their data back to the workstation - I played with how many passes it takes between sending the data back and got less frequent black screen. Flicking back and forth to other elements gives the same behaviour too (returning VFB back to normal).

Hi,

Would you both be able to provide me with your scenes? I am currently trying to reproduce this, but so far I have been unsucessful.

Thanks,

Rowan

Sorry for keeping on this subject, I tried an older build and no success, am trying the save to CXR workaround. it could be EXR issue. I have 9 renders to deliver in the next 2 days and uncertainty if the pc will save a good image or not is a spirit killer. I updated to the latest Nvidia drivers and am now running latest build since changing to early April build did not solve it.

[danio1011]

I've had this issue for the latest 2 builds, when i render using DR, the frame buffer Beauty pass will go black randomly after a few minutes, then if I cycle between the elements and back to beauty it will show up again, then go black after some minutes. When render ends it is randomly either saving or not saving the beauty (black layer) depending on if it shows or not at the time. In some cases the other elements also would have the alpha wrong as in a big blag square of the element missing or added. Anyone having same issue? could it be a memory leak when it receives data from the render nodes?

+1 exactly the same issue. Haven’t had any loss when saving out (to Cxr) but the black VFB during DR is identical. I agree that it happens with the DR machines send their data back to the workstation - I played with how many passes it takes between sending the data back and got less frequent black screen. Flicking back and forth to other elements gives the same behaviour too (returning VFB back to normal).

Hi,

Would you both be able to provide me with your scenes? I am currently trying to reproduce this, but so far I have been unsucessful.

Thanks,

Rowan

Sorry for keeping on this subject, I tried an older build and no success, am trying the save to CXR workaround. it could be EXR issue. I have 9 renders to deliver in the next 2 days and uncertainty if the pc will save a good image or not is a spirit killer. I updated to the latest Nvidia drivers and am now running latest build since changing to early April build did not solve it.

This is a shot in the dark, but was the older build you tried before they introduced the bloom and glare progress bar was introduced into the dailies?

[ihabkal]

thanks for the tip I am starting to think it's the high quality filtering, I turned it off and it seems to not turn black randomly anymore. it seems to be it. I rendered all I needed for my work with CXR so I am ok except for the photoshop all nighter I might have to do to please a client.
CXR is great if we can have an adobe photoshop reader so we can open that one file and have all layers just like EXR, but it takes an extra 10 to 15 minutes for me to export the layers and reassemble them. other than that it seems to be much better in system responsiveness but the file is 7 to 9 times bigger than EXR so my renders are up to 5.6GB per view now from a few hundred MB.

[rowmanns]

thanks for the tip I am starting to think it's the high quality filtering, I turned it off and it seems to not turn black randomly anymore. it seems to be it. I rendered all I needed for my work with CXR so I am ok except for the photoshop all nighter I might have to do to please a client.
CXR is great if we can have an adobe photoshop reader so we can open that one file and have all layers just like EXR, but it takes an extra 10 to 15 minutes for me to export the layers and reassemble them. other than that it seems to be much better in system responsiveness but the file is 7 to 9 times bigger than EXR so my renders are up to 5.6GB per view now from a few hundred MB.

Hi,

When rendering with the high quality filtering did you by chance have the AI denoiser enabled?

If you could provide your scene I will be able to investigate this further.

Thanks,

Rowan

[Jpjapers]

Very very disappointed to see that the PBR material has been taken out of corona V4 on the trello.
Hopefully it means that some uber shader is on the way in the future. Or that the team underestimated how much work the great new caustics needed.
Realistically the PBR shader was all i cared about in this new release as it brings (in my opinion) the most time saving benefit out of all the features requested by not having to build workarounds and being able to work efficiently with pbr shaders.

[Christa Noel]

hi devs, is there a maxscript function by corona which do "create a coronaCam from active perspective view"?
i know that it is doable manually but who knows there is something undocumented like this:

Code: [Select]

coronaRenderer.coronaFp.coronaCamFromView()

[mike288]

hi devs, is there a maxscript function by corona which do "create a coronaCam from active perspective view"?
i know that it is doable manually but who knows there is something undocumented like this:

Code: [Select]

coronaRenderer.coronaFp.coronaCamFromView()

Hi, there is not. But it can be done from the toolbar.

[Dippndots]

Hi,

Sorry we missed this one earlier. We would genernally expect that the system caches the files intelligently, but this might not always be the case.

So in order for us to investigate, it would be really helpful for us to get a copy of your scene and assets if possible?

Instructions on how to do this are in my signiture, please don't forget to let me know the name of the file which you have uploaded.

Thanks,

Rowan

Hey Rowan, thanks. As is tradition, the problem seems to have gone away after we tried a couple more things. If it returns, I'll be sure to make a test scene and asset folder and send it to you.

[fraine7]

Wondering if anyone else has experienced a full system lockup like I have several times with the latest daily build?

Plenty of RAM available and scenes are not very detailed but in 3 different scenes I’ve had windows become completely unresponsive after cycling through render elements in the frame buffer. Not using caustics and in 2 of the scenes I wasn’t even rendering at the time of the lockup, it’s like the frame buffer has tried to catch-up with a change and then got a little bit lost, best way I can explain it for now.

I will do some further testing before sending any scene files but curious to know if this has happened to others?

[JulioCayetano]

Not sure if it's related,  but I have also suffered a couple of full system freezes. First time, I had to re-start the PC. It was behaving correctly and then bam, everything unresponsive. The second time, only the windows desktop got frozen, but the apps still running. Both times I was using Corona with the first Daily that came with Caustics, but they weren't active.

[PROH]

Hi. Just tested the latest DB (may 10) which support multiple suns & environment maps. I want to use it for sun/shadow studies, and therefore used several daylight systems set to different times.

It works as expected with the suns, but I can't get it to work with the skylight maps. I've tried using a single skylight map for all suns, and I've tried using multiple skylight maps (one for each sun), but the problem is the same: the skylight map doesn't follow the suns time as it should.

I can trigger it to change by hiding the suns not used in the layer maneger, but I can't get it to work from the lightmixer alone.

Any idea how to do this, or is it a bug?

[Dom74]

it would be fantastic if the suns could not emit shadows.