[fuso]

Actually you're right Brodie and first of all a big sorry to Tok_Tok for hijacking this thread, it's been blown out of proportion,
I just got carried away when realising that I'm not alone out there, biting my fingernails and pulling out my (mostly grey) hair!

And I certainly agree with Brodie (thanks for reinforcing my point). I also have scenes which seem to be fine and I normally
have a pretty consitant workflow based on quite a few years of experience now. Furthermore, I do spend a lot of time on
optimisation, only to find out that a few issues appear and disappear seemingly random. Maybe Maxwell is a software which
needs a bit of 'healthy chaos' in order to work properly, but it is something which goes against my German principles and habits.
Or am I wrong there?

I would love to share the scene (in fact I'd die for some feedback) but unfortunately that's not possible due to confidentiality
of this and many other projects. So I just have to wait and let the renders cook...

[Half Life]

You may be right there -- I've got my issues with the software as well and they do seem to be a moving target sometimes.

Like I say there is much room for improvement and I'm sure it will improve -- certainly make your voice heard on the wish list and in the bug forums (and here too!) because as a longtime user you have valuable perspective that I can only read about.

I'm glad you said something, but I'm with Brodie -- you've got power to burn (I wish I had half of that), clearly there is something else going on.

Do the benchmarks all look the same for the similar spec machines? Also are you using anything like simulens or saving the output to a large bit-depth image format? Are you using displacement?

In my experience it is a bad idea to cooperatively render a scene on unevenly powered machines -- the weaker machines bring the whole process down... like the old saying "only as strong as the weakest link."

Best,
Jason.

[fuso]

Thanks Jason, I really appreciate your words! Ok let me try to go into as much detail as possible with this particular job:

Local node (JW 220):
[INFO] Num processors: ( logical: 8, physical: 8 )
[INFO] Low Priority Enabled
[INFO] - Num Meshes: 219
[INFO] - Num Triangles: 76553
[INFO] - Num Vertexes: 41011
[INFO] - Num Normals: 35720
[INFO] - Num Materials: 29 (6 of which emitters, two AGS glass, one 'real' glass)
[INFO] - Render version : 2.5.0.0
[INFO] - desired rendering time : 9990
[INFO] - desired sampling level : 19.000000
[INFO] - using 0 threads
[INFO] - using RS1 engine
[INFO] - illumination layers:
[INFO] . . direct layer: true
[INFO] . . indirect layer: true
[INFO] . . direct caustic reflection layer: false
[INFO] . . direct caustic refraction layer: false
[INFO] . . indirect caustic reflection layer: false
[INFO] . . indirect caustic refraction layer: false

The render output is 1800x1200 SINGLE channel 8bit png:

-s:19 -t:9990 -nowait -node -idcpu:4535 -slupdate:10 -v:4 -depth:8 -camera:view03 -ml:off -motionblur:off
-displacement:off -dispersion:off -res:1800x1200 -p:low -channel:alpha,no,16,png -channel:shadow,no,16,png
-channel:material,no,16,png -channel:object,no,16,png -channel:motion,no,16,png -channel:zbuffer,no,16,png
-channel:roughness,no,16,png -channel:fresnel,no,16,png -embedded -zmin:0 -zmax:1 -threads:0 -p:low -rs:1

SL after 1h 45min between all 9 (2 dual quad core, 1 quad core hyperthreaded, 5 quad core) nodes are pretty consistant:
Cooperative SL at that time: 15.554

Render node: JW231 Job ID: 1 sl: 10.0892
Render node: JW232 Job ID: 1 sl: 9.93101
Render node: JW235 Job ID: 1 sl: 9.9784
Render node: JW237 Job ID: 1 sl: 10.0782
Render node: JW238 Job ID: 1 sl: 9.98311
Render node: JW239 Job ID: 1 sl: 10.0687
Render node: JW240 Job ID: 1 sl: 10.9157
Render node: JW222 Job ID: 1 sl: 11.2467
Render node: JW220 Job ID: 1 sl: 11.2419


Hopefully, that gives you an idea why I'm so concerned, under the set-up above this job should be finished after 1-2 hours.

Best,
Jost

[Half Life]

Thank-you, a couple questions:

[INFO] . . direct caustic reflection layer: false
[INFO] . . direct caustic refraction layer: false
[INFO] . . indirect caustic reflection layer: false
[INFO] . . indirect caustic refraction layer: false (why have real glass and disable these?)

Are you using a very narrow depth of field? Do you have any materials with abnormally high bump settings?

But other than this I cannot see anything wrong with the settings -- so that leads me to believe that the problem would have to be geometry/material based... but it would have to be really bad to slow it down that much and all the usual suspects are covered so that to me means that there is some major issue with the emitter geometry/setup or a bug.

I'm assuming the emitter geometry is all very low poly, no intersections, etc?

Best,
Jason.

[fuso]

Hi Jason, once again thanks for looking into this. I've disabled the caustics because I wanted quick 'tidy' renders with not too
many artefacts caused all over the building and pavement in front. I also still have the impression that it slows down the render
process but I might as well be wrong there.

Here are the camera settings:
F-Stop: 5.6
Shutter Speed: 420
ISO: 100
Lens: 35mm
Field Of View: 52.0776
Target Distance: 20m
No SimuLens

As for the bump map values, I didn't find anything higher than 5 and they're all pretty average in contrast. The geometry is not
really heavy and has been used and optimised many times. No intersections of emitter objects with other geometry and they're
as low-poly as they can possibly be. The only 'fancy' stuff I can find is one coating on that 'real glass' I mentioned, and I'm
using full IOR on one brushed stainless steel material. Naughty.

Cheers

[Half Life]

Yep, I'm at a loss -- the IOR on the metal should have next to no render time impact (unless you are using germanium) and while the coating on the glass doesn't render correctly (at least I think so, which is one of my issues) it should not really be that much of an issue unless it is backlit.

Here's something you can try: Do a clay render/ material override -- this is a quick way to test if it is material related or not... if it renders much faster then replace all materials with the default but the emitters and render again. If it is still good then only try one additional material at a time. This way you may be able to narrow it down to the culprit.

BTW dispersion is enabled with IOR materials and you have dispersion disabled in the render settings... could a conflict like that cause a bug?

Best,
Jason.

[Tok_Tok]

At the end of the day I come back to the forum with a cup of coffee to respond to Jason's post. But then I see that someone else has been
posting some much that I could use this thread as wallpaper for my room.. with some apologies woven into it for high jacking this threat (I think I'll underline those before sticking it to the wall)

hahah don't worry guys, my question was answered and I thank Jason very much for the info!

[Tok_Tok]

Oke, I had time to re-render the material with an orange color at 224. It is a bit better but the image still has noise after 22 hours of rendering
at SL 22. Next I'll try Jason's carpaint, and if that doesn't change anything I'll go with a coating

[Tok_Tok]

Well Jason's carpaint worked very well! The image is free of noise and has a very nice overall look, if i say so myself

For the color i went with a 242 orange and that turned out very well. The only thing is that the paint looks a lot darker then i expected.
But i'm happy with this result, thank you for providing the material Jason!

[fuso]

Hey Tok_Tok, that's looks REALLY nice!! And it's a nice post to conclude this temporarily hijacked thread! So, where do we get
those RGB carpaint materials? I'd love to have a go as well... Oh and another big thanks to Jason for all his efforts of course!

[Half Life]

Here's a temporary link, I can only leave this up for about a week or so:

Link removed

This contains all my materials from this thread:

http://www.maxwellrender.com/forum/view ... 49&t=34388

This one zip file is about 200 mb and all the materials therein are the most recent updated versions (for Maxwell 2.5) -- I recommend anybody who uses my materials to download these and replace all previous versions you may have on your system as these will be superior in many respects.

The reason I can only leave these up for a short while is they will be included with the workfiles in my next video series release.

Best,
Jason.

[fuso]

Thanks so much Jason, that's very kind and generous of you. I'm looking forward to explore that package!!

Cheers from London!

[Half Life]

You are welcome I meant to post these to the forum earlier but it slipped my mind until you asked about it.

Best,
Jason.

[zdeno]

Santa Claus is early this year I see

Merry Christmas everyone! especially for You Half Life !

[m-Que]

Thanks a lot, Jason!