Daily builds version 3

[ihabkal]

something very depressing going on here, I think it's embree related, my Threadripper is doing 60 to 70% cpu load tops. I used to always see 100%, tried stress test with CPUZ for 100% temps and cooling are good.

[ihabkal]

something very depressing going on here, I think it's embree related, my Threadripper is doing 60 to 70% cpu load tops. I used to always see 100%, tried stress test with CPUZ for 100% temps and cooling are good.

I think we're back on track, I re-opened and tried and played with the developer settings, somehow i got back to 100%, and from 14 hrs to 2 hrs estimated render time. almost 2:AM here got to get some rest but hopefully it was just a settings fluke.

[bluebox]

So the current daily is stable for now and can be used in production ?

[Nekrobul]

So the current daily is stable for now and can be used in production ?

I am not shure now, i have encountered crash on scene load but the scene is extremly heavy. I would not risk before the masage from devs.

[Juraj]

It is extremely stable for me and has been for past three weeks and I am using Optix non-stop. But I do not use Forest-Pack currently in any scene.

[romullus]

Hmm, i'm getting totally different result. Here's statistics from my quick test with 7% noise limit as stopping condition:

AI denoiser: 5 passes | 5,62% | 36s
Corona denoiser: 15 passes | 5,78% | 2m38s

What is the actual perceived image quality? Do they look the same? Because if so, that's an insane productivity boost...

Actually, i didn't compare image quality, but i have no doubt that images doesn't look the same. Otherwise it really would be insane :]

[bluebox]

Thank you guys. I'm really eager to try Optix but am a bit afraid of any bugs that might occur. Is the current daily backwards compatible with 2.0 ?

[romullus]

I have encounterd very strange issue. The scene created with 30-08 build is contineously crashing with latest build on start.

Same here - installed build 09-14, got crash on load of very first scene, rolled back to 08-16.

[Bzuco]

From changelog:

Improved light sampling in case the geometry is very close to a spherical light.

Can this help eliminate noise near light which was switched off in lightmix?....issue from this topic ?

[Nekrobul]

I have encounterd very strange issue. The scene created with 30-08 build is contineously crashing with latest build on start.

Same here - installed build 09-14, got crash on load of very first scene, rolled back to 08-16.

Do you have corona scatters in this scene?

[romullus]

Do you have corona scatters in this scene?

Nope, i think there were no scatters in that scene.

Hmm, i'm getting totally different result. Here's statistics from my quick test with 7% noise limit as stopping condition:

AI denoiser: 5 passes | 5,62% | 36s
Corona denoiser: 15 passes | 5,78% | 2m38s

It might be that it's just scene specific thing. I tried to replicate it in a more "normal" scene and i got pretty similar results between Corona and AI denoisers.

Hmm, i'm getting totally different result. Here's statistics from my quick test with 7% noise limit as stopping condition:

AI denoiser: 5 passes | 5,62% | 36s
Corona denoiser: 15 passes | 5,78% | 2m38s

What is the actual perceived image quality? Do they look the same? Because if so, that's an insane productivity boost...

Actually, i didn't compare image quality, but i have no doubt that images doesn't look the same. Otherwise it really would be insane :]

Well, eventually i couldn't resist to run that test again and surprisingly, except aliasing, there weren't much difference between the two images. Crazy, but then again, the scene is quite specific.

[Dung (Ivan)]

Hi, search lan on distribute render not working, could not find any computer

Can confirm (DB 18-08-30). Additionally ram information is (again) not displayed in DR tab. Had firewall(s) down, even used the installer to test - no change.

Can you try DB 14.9.2018, please? I can't seem to reproduce either of the issues. Are you on wifi-lan?

[shortcirkuit]

Hi Devs

I noticed with the latest build that if you have 2 instances of max open, one of them is able to find render nodes (slaves) when you add them on the DR list while the other max instance doesnt find anything..... this is a first.

PS mesh by component - amazing addition :)

[fraine7]

I'm currently rendering out 1100 frames of an animation sequence and have noticed some odd behaviour with my render nodes using 3.0 dailies.

Using Corona 2.0, my nodes download the xrefs once as the first frame begins to render. As each frame progresses they parse the scene and join in when ready - this is what I was expecting when I tried the latest 3.0 daily (but it's probably the same activity as in the previous dailies - I will need to check).

Using 3.0 daily from 14-09-18, the nodes load the xrefs on frame 1 which takes about 3 minutes, they then start rendering. On frame 2 and every frame after that, they are downloading the xrefs again from scratch when they already have all the scene data, I'm assuming this should be stored in memory until the render is complete or cancelled - correct me if I'm wrong here btw.

My renders only take 2 minutes per frame on my new ThreadRipper 2990WX, if my nodes could join in without having to unnecessarily download xrefs on every single frame then they would be able to join in and speed up my rendering further.

My question is, is this normal/intended behaviour for xrefs to need a complete redownload on every frame, nothing is changing at all - same xrefs throughout the sequence. 2.0 handles everything well but 3.0 is problematic

[Frood]

Can you try DB 14.9.2018, please? I can't seem to reproduce either of the issues. Are you on wifi-lan?

Same issue, standard ethernet LAN. Additonally I upgraded the LicenseServer and it cannot be found by the nodes - fatal :) Somehow it looks to me like broken subnet broadcasts.


Good Luck