Material Editor performance issues

[Juraj]

OK, uninstalled it all and installed 7.2.

ZERO issues, everything is 100times (even more actually faster), converstion takes 1/10 of second, everything is real-time and fluent.

So what gives ? Like how could this go for whole month ? Is no one here using 2015  and Daily build to do anything ?

[daniel.reutersward]

I´m using 2015 with sp3 and daily builds.

I´m currently using build 2015-01-18 and it has gotten like 5% better since before.

However since I started using daily builds the material editor performance has slowed down a lot! From 100% speed to about 15%...

3ds max also freezes a and is extremely slow when using slate material editor with a lot of both simple and more complex materials, almost unusable. Sometimes I have to wait up to 10 minutes before 3ds max responds again when fixing materials.

[Juraj]

I am going to test now the rendering speed, since the mat editor showed such drastic difference (100times faster is understatment), I have a feeling the slow rendering times I got in recent past were all due to this as well.

[pokoy]

My reports date back to October and since it was limited to scenes with high material count I might have been only one of a few who noticed at all. Now it may get some traction if it occurs on max 2015 without any content in the scene.

[racoonart]

I didn't notice anything in simple scenes but there is indeed something wrong.  Simple scenes are fast and fine, bigger scenes take considerable time to load the mtl editor slots (2, 3 secs each). But it's definitely not "heavier" materials. If I try to edit some complex materials it takes 2 or 3 secs to update but it takes exactly the same amount of time to edit a fresh CoronaMtl or StandardMtl. It's like it tries to re-evaluate all materials on all objects and updates everytime I change something.

I just made a scene with 10k boxes and 1 CoronaMtl, it shows this behavior. Scene (2014) is attached. Default scanline updates mtls instantly, as soon as you switch to Corona it's slow.

(funfact: something like that happend with the "new" slate material/map browser for a long time, and probably still does to some extend. It took 20 minutes to make a new material when you had 40k boxes with 1 mtl instance applied to all in the scene. Old mtl browser did it instantly)

[Juraj]

I will look if some computer still has 2014 installed and test if that exhibited similar behaviour as 2015.

But 2015 + 7.2 = Fantastic

[racoonart]

I tested both 2014 and 2015 (fresh, only corona) and it behaves similar.
Try the box scene I attached in my previous post :)

[Juraj]

I think it's redundant, since I have the issue even without any box at all and no materials.

[pokoy]

Guys, thanks for giving this some attention. I have been hitting a wall on this for a while already.

Now, it's strange how much our tests differ. In my case it's clearly the number of present materials, not geometry or object count. I wonder if it has to do with CPUs and threading and in general (I'm on a Dual Xeon machine) but it's good to have some broader testing now.

To emphasize that it's a daily build thing only, I didn't have this issue back then when using Alpha 7, but didn't see the problem when I went with the daily builds back in September. It got gradually worse when IR was introduced in the dailies and threading was generally a problem with IR.

[Juraj]

I checked only workstation we work on (4930k), but now that you mention it, I will check Xeons.

Edit: OK, the Dual-Xeon (but also Ivy-Bridge based v2), has none of those issues in freshly opened scene. Daily build.

What the hell gives... ?

[racoonart]

Ondra needs a scene where he can see what happens, it happens with 10k boxes, no one else provided a scene so far, so.. no, I don't think it's redundant.

[Juraj]

Ondra needs a scene where he can see what happens, it happens with 10k boxes, no one else provided a scene so far, so.. no, I don't think it's redundant.

Sorry for misunderstanding, I only meant me :- ). Since I have the issue without any particular scene at all. And now, it differs per CPU.

But I will try your scene now on Xeon.

[Juraj]

Well.

Your scene (10k) makes everything rather slow on every computer, irregardless of Daily build or 7.2. (from 4930k to dual-xeons)
Even scanline material behaves slower and lags, but it's different behaviour (not that super slow) than what I had with daily build.

Edit: It's slow even when switching to Vray or Scanline. I think your scene mostly proves 3dsMax goes to crawl when "selecting" too many objects (or even having them?). The moment I deselect those 10k boxes, I don't have any as many issues with any renderers (apart from daily build, which is allways, no matter what conditions).

Example:

Dual-Xeon Daily build: no issue anywhere in clean scene, can't even properly open the scene, just freezes forever, or opens...occasionally before going to freeze. Even the viewport has issue (but I am going through remote control).
4930k Daily build : issues persists in clean scene as well (but only while Corona is active) , 10k scene has similar issues as Xeon above
4930k 7.2 : in clean scene no issues, in 10k scene still slow (with all renderes), but not even close to issues of above.

[racoonart]

Yes, it's always slower (it has to update viewport colors for example), but this here is different:
- Create a teapot, isolate selection (so no boxes have to be redrawn in vp)
- try changing a property of the corona material AND one of the material right of it (specular amount for example)
- now switch Material editor renderer to Scanline and try again. Scanline will update the non-assigned material instantly, Corona will be as slow as with the assigned one.

[pokoy]

I checked only workstation we work on (4930k), but now that you mention it, I will check Xeons.

Edit: OK, the Dual-Xeon (but also Ivy-Bridge based v2), has none of those issues in freshly opened scene. Daily build.

What the hell gives... ?

Ha! That's odd but might help to get an idea what's causing it.
My setup is a dual 2650v2.