Material Editor performance issues

[daniel.reutersward]

My 3ds max also freezes(besides slow material preview rendering), for about 10-15 minutes, when I have the slate material opened without using it. I´m also on a dual xeon setup, dual 2697 v3.

[Juraj]

Deadclown, what CPU and 3dsMax(&servicepack) are you using ?

[pokoy]

My test with the 10kBoxes.max:

MatEd Scanline = under 1 sec.
MatEd Corona = 13 sec.

Still faster than my 60 sec wait for my production scenes.

DeadClown, could you script a setup where you create 300 unique CoronaMtl (with unique names maybe) and assign them to the boxes, but in a way so that every box gets a material from the 300 mtls pool? I'd like to compare that setup one against the one you already posted.

[racoonart]

Deadclown, what CPU and 3dsMax(&servicepack) are you using ?

Dual Xeon E5520, max 2014 with sp5, max 2015 with sp3

DeadClown, could you script a setup where you create 300 unique CoronaMtl (with unique names maybe) and assign them to the boxes, but in a way so that every box gets a material from the 300 mtls pool? I'd like to compare that setup one against the one you already posted.

Code: [Select]

(
numBoxes = 10000

mtls = for i = 1 to 300 collect CoronaMtl name:("TestCrn_" + i as string)

cnt = 1
for i = 1 to numBoxes do(
if cnt == 301 then cnt = 1
mb = box()
mb.material = mtls[cnt]
cnt += 1
)

)

[pokoy]

Thanks for the script!

My tests with the script:

10k boxes, 300 materials:
MatEd Scanline - under 1 sec.
Mat Ed Corona - ~14 sec

10k boxes, 600 materials:
MatEd Scanline - under 1 sec.
Mat Ed Corona - ~15 sec

Benchmark done manually, so it is not accurate.

It seems the number of materials has a small impact only.
I'm still not seeing the issue I had with the production scene, so I guess material complexity plays a big role in the slowdown, too.

[Ondra]

Fixed. It was a matter of 5 minutes with the right scene.

For the love of god, why are you reporting it only now?! I mean, reporting specific times and providing nice scene for testing, not saying "I think material editor is now slower" This could have been fixed month ago

[pokoy]

Fixed. It was a matter of 5 minutes with the right scene.

For the love of god, why are you reporting it only now?! I mean, reporting specific times and providing nice scene for testing, not saying "I think material editor is now slower" This could have been fixed month ago

My bad :)

My take is that it has to do with all kinds of reports going into one thread and being overlooked fairly quickly without being picked up by the community. I'm afraid that if no one else would've stumbled upon the same issue it would've remained overlooked regardless of how well I would have backed it up.

Thanks for fixing this, can't wait to test the fix.

[Ondra]

yes, that is possible. This only highlights 2 very important rules:

1) EACH BUG REPORT IN SEPARATE THREAD

2) ALWAYS INCLUDE SCENE

;)

[Juraj]

I don't hold my breath yet, but looking forward to fixed daily to check. I don't think the scene illustrated the issue so well because I had it without any geo at all, at one type of CPU, and not the other.

The fact that issues aren't reported lead me to that perhaps not enough testers are doing anything much and most people still use 7.2 for commercial work ? Or perhaps the breadth of platforms isn't that broad (3dsMax versions, Windows versions, CPU types,etc..).

[pokoy]

I get it!

Honestly, I'm happy if it's resolved. Also would like to test the fix on some heavy scenes.

[Ondra]

updated daily is up, you can try it out.

[racoonart]

I don't know how many daily-build testers have scenes which are that kind of "big". I didn't notice until I read the thread and made the box scene. Smaller scenes (what I usually do these days with Corona) worked fine.

[pokoy]

I confirm that the issue is solved with the newest build (Jan 19 2015 17:19:26).
Both the 10kBoxes scene and some of my production scenes now work like they used to before.

Keymaster, you made a man happy today. Thanks!

[Juraj]

Why is it resolved ? Because it solved single case ?

It's faster, but not even close as fast as 7.2, which I just ugpraded to latest build. I attach screenshot on what happens in freshly opened (No geometry at all), scene. 50 second freeze became 20 second freeze. 7.2 did it under one second.

The materials themselves do provide feedback much faster, without such noticeable delay, but I wouldn't call it ideal.

I would much rather this to keep as consistent existing thread.

I was able to reproduce the slowness on 3 different 4930k i7 builds. Dual-Xeon 2680v2 doesn't exhibit it, and runs smoothly. They both run Windows 8.1, 3dsMax 2015 SP 2 and latest daily build.

[pokoy]

For me the MatEd displays materials instantly now, even on the heavy scenes it's quite fast.

Now I wonder if max 2015 is part of the problem. Can you test the same build with 2014?