Why is 1.0074 so much slower?

I upgraded to SR1 (1.0074) and now my scene processing time is exorbitant. In the original release it would take at most 10 seconds to process a model before the initial passes would begin…now it is taking 10-20 minutes before the initial passes begin. Is anyone else noticing this?

I was also initially getting dictionary attribute errors when I start sketchup, but realized it was because I had skindigo installed as well. Now that I removed indigo, do you think a reinstall of Vray would speed things back up or is that a non-issue?

It’s frustrating because I love the new release for the way it handles materials and clipmaps, but all the time I save in setup, I lose in processing.

Re: Why is 1.0074 so much slower?

A bug in 1.00.74 makes Light Cache very slow. Untick Batch Render and you should be back to normal again.

Re: Why is 1.0074 so much slower?

Oh, hang on, do you mean the processing before the render window appear?

Re: Why is 1.0074 so much slower?

Yes, before the render window appears at all. I have batch render unchecked and low thread priority unchecked. Not sure what’s going on. I’m thinking it’s either the fact that I’m working on a model that was originally setup using the original release, and thus I have linked materials present, or it’s due to installing SR1 while I had indigo present. Not really sure…just a guess and a good place to start.

On a side note, I noticed that if batch render is checked you can speed up the light cache pass by moving the scroll slider of the Vray dialog box up and down. It’s weird, but it works. Maybe that fact will give the programmers a clue about the bug and also help speed up caching when rendering over a network.

Re: Why is 1.0074 so much slower?

Ok, then I’m not sure what it is.

I thought it was the Light Cache issue. It’s mentioned in a couple of threads here. You can also click on the window’s titlebar and it will run at normal speed. Very odd. But fixed in the next build.

Re: Why is 1.0074 so much slower?

Earthmover,

I completely despaired the first time I ran v.1.00.74- I was working on a 20mb model of about 2km2 urban area and the previous parse time of 3 to 4 minutes became more like 20 minutes. ASGvis explained that this was due to v.1.00.74 improved parsing (i.e. it doesn’t screw up materials like v.1.00 did, which also means groups don’t disappear from renders), but it didn’t feel like much of an improvement to me.

Since then I’ve found out that the ridiculous parse times are due not to geometry, but rather because v.1.00.74 takes forever to analyse materials which are applied to groups or components. If you try this test- do a “save as” of your model and then delete all your materials from your SU material editor so everything is just the default texture you’ll see that the parse time will be similar to v.1.00 (actually less as there’s no materials). Unfortunately this means that the only way to ensure that old skp/VfSU models will parse in a reasonable time is to go through the group heirarchy of your model and apply the default material to all the groups so you can establish which have their materials applied directly to faces and which have materials applied to the groups. For this reason I have my default set as a very bright yellow for front faces and bright blue for back faces so these colours stand out against more “normal” textures in models. When you come across groups which have turned to the default colour, go into them and reapply the textures directly to the geometry and once they’re all done your model should parse pretty quickly. Unfortunately it’s imperative that you do this with every single group and components in a model as I have a fairly simple skp/VfSU model of the Barcelona Pavilion which was taking 15 minutes to parse until I discovered it had just ONE COMPONENT which had material applied to it and this was causing the parse delay.

I really wish ASGvis had released a mini-tutorial/help file with v.1.00.74 as there’s so many little things which I’ve had to relearn after using v.1.00 and so many bugs and workarounds which I’ve had to scrabble around for on these forums.

Re: Why is 1.0074 so much slower?

That’s interesting findings.
I felt that even the original version was slow when materials was applied to group and components, so I stopped doing that a long time ago. Unfortunately we have some old models at the office which does. Actually had one last friday, it had about 800.000 faces (! - Yes, insane. Very detailed with high poly components..) It would simply not render. That was with the original.

With the new version I’ve only tried to render new models, where all material are applied to geometry. So I haven’t actually thought of it. But it’s useful info to know. Especially when I upgrade at the office.

You could ask over at the sketchup forum if anyone could make a ruby that removes material from groups and components. That is provided though that you can find the components that’s now missing materials. Alternative, the ruby could change the material to a very bright colour so it’s more easily seen.

Re: Why is 1.0074 so much slower?

That makes a lot of sense as I have quite a few nested components in my model. Would bombing the model to explode everything help speed up parsing times, or would materials then need reapplied to the exploded faces? I did a few tests with a clean file and just some simple ungrouped shapes, and the parsing time was fairly instantaneous. Also, within the model that’s giving me the slow time, I went and renamed all the material that were “linked” by version 1.00 and this seemed to cut my parsing time in half.

Interesting suggestion for a ruby. That would be useful when dealing with a lot of existing components.

Re: Why is 1.0074 so much slower?

If you haven’t used the SketchUcation forums yet, Developers' Forum | sketchucation , it’s a really great resource.

Re: Why is 1.0074 so much slower?

There are a number of issues that existed with the parsing in the original release. There was issues with components, groups, and materials that caused geometry to disappear, materials to not show up, or texture coordinates to be screwed up. Those were significant issues and had to be resolved. There were also other issues that had to be resolved in terms of scaling, that were required in order to have the correct response from photometric lights.

So, we made changes to the parsing process to resolve those issues. This meant that we had to do more with the scene, not less. Components were the biggest problem in this regard, because SU offered us very little to get the information that we need from those components. This gets compounded with components within components.

There’s no doubt that its slower than it was in the original release, but we we said that there were improvements, we were referencing the changes and improvements that were made to the process. We don’t like that the process takes as long as it does, but we have our hands tied. SU doesn’t give us as many tools as we require to have the data that we need from the scene. SU’s programming interface is also EXTREMELY slow. We have, and still are, looking for ways to make the process quicker, but we’re fighting a very steep uphill battle. My advice would be to minimize the number of components and nested components as possible.

Re: Why is 1.0074 so much slower?

So components in general makes the processing slower? Or just components with materials applied to them.

Using components is an very important piece of SU modelling in regards to getting a good workflow.

Re: Why is 1.0074 so much slower?

Great news guys, I posted a request on SCF for any ruby scripts that might help with this major problem and Matthieu Noblet very kindly wrote a ruby which automates one stage of the very long, tedious process of replacing group materials with geometry materials.

POST EDITED BY AUTHOR: This script is superceded by the Remove-CG Materials script which can be found in the following thread:

Re: Why is 1.0074 so much slower?

That was quick. That’s why I love that forum.

Re: Why is 1.0074 so much slower?

Amazing! Thanks so much for requesting that. I will try it now and see if it helps.

Re: Why is 1.0074 so much slower?

Is this just components? Or groups as well?

Re: Why is 1.0074 so much slower?

Hy!
I’ve been having some troubles too with the rendering times.
Just to give you all the data about my problems:
- i use v-ray settings allready made, since i am no genius, and i don’t really understand how all those settings work. I only change the focal lenght, to get a wider view of the rooms i’m creating
- i apply materials to the geometry, but i also have components inside components. can this cause the major delaying time?
- i choose to use components instead of groups because there have been cases before when groups disapeared from the scene
- i’ve tried all the settings i could find for interior scenes, and the problem seams to be the long “building light cache” time (very very very long time)
- i’ve been using the new rectangular lights (with the “watts” options), hoping to get closer to the real life parameters. What do you recomend?
- another strange thing…in scense with few materials (like a bathroom - 4-6 materials) things are fine, but in scene a little more complex (like a small livingroom 10-15 material)things just slow down so much that i had to use some settings that were dedicated for the exterios scenes (i think)just to get some renderings in time.
- i do small pictures, like 800X600
- i don’t have great expectations from my scenes, i just want it to work like it did before. the rest is up to me.

PS. I’ve been reading posts on this forum, but i don’t seem to find the answers i need. I hope to get some good, short advices.

Re: Why is 1.0074 so much slower?

There could be any number of things to why things takes time. I think your best bet would be to make a post with specific questions to your specific models. With screenshots and info on your settings.

Also, the latest version is now 1.05.30. Is that the one you use, or do you use 1.00.74 which this thread refers to?

In regards to Watts: 60W V-Ray light doesn’t equal to a lightbulb or lamp of 60W. Instead it’s actual Watts emitted, which is much lower. When I tried to approximate existing lights I had more luck getting the lumens info and matching that unit in V-Ray.

Re: Why is 1.0074 so much slower?

Teya,

I’ll try to answer your questions/bullet points in order:

- Did you know you can change the focal length in SU itself without having to adjust it in VfSU’s camera options? Just select SU’s zoom tool and type in the new focal length (suffix it with “mm” to enter standard camera focal length figures or “deg” to enter Field of View figures). That way you’ll get much closer to WYSIWYG output between SU and VfSU.

- Components inside components do cause parsing slowdown (parsing= the process of VfSU “reading” the skp file), but don’t let that stop you using them. Components are so fundamental to good SU modelling that they are too valuable to give up on and there are ways around the slowdown! (see my final bullet point)

- As a rule, never use components instead of groups, the more components you have in an SU model the slower VfSU’s parsing will be. Just use components and groups as you normally would in SU: groups for geometrical objects which are only used once and components for geometrical objects which will be repeated thoughout the model and which may need to be edited simultaneously later. The “disappearing group bug” in VfSU can be fixed quite easily by following the advice in my final bullet point and sometimes by simply saving, closing and reopening the SU file.

- There’s a rather complex method of calculating the optimum Light Cache subdivision parameters as they relate to the size of your actual render, but as a very rough rule of thumb I use 500 for most screen size (1400x800) renders and up to 1000 for higher quality or larger images. Anything over 1000 will start causing very slow LC build. That’s really a dummies’ guide to LC, search the forums for Dalomars detailed explanation of how LC is calculated. More importantly untick “Batch Render” under the “Global Switch” menu- it causes a very long delay in the LC build.

- No problem with using rect light and watts (though in the long term it might help to become more accustomed to using lumens as they’re the industry standard and are based on how the human eye actually perceives light rather than the rather abstract one joule per second Watt unit, which doesn’t actually describe light, but power. As a rule of thumb, a 100W bulb gives out approx 1200 Lumens.

http://en.wikipedia.org/wiki/Lumen\_(unit)

- This is the most important one! Matthieu Noblet created a free SU ruby script called “Remove C-G” which will take any material applied to the outside of groups and components and apply it directly to the geometry/faces. If you’re wondering what the hell I’m talking about think of car components which have materials applied to tyres, headlamps, bumper, etc, but which have SU’s default material applied to the bodywork so you can apply a material to the component and all the bodywork will be painted he new colour. The trouble with this SU method of working (although useful in SU) is that it makes it extremely difficult (read: slow) for VfSU to read the skp file. This is where Matthieu’s ruby script comes to the rescue. After running it you usually won’t actually see any difference to your skp file, but there will no longer be any default materials within your groups and components as all materials will be applied directly to geometry. Before running it some of my models had parse times of 20+ minutes, now all but the largest model will parse in less than 30 seconds. Be warned that the first time you run it in very large SU files it can take a long time to work and SU will hang while it does, but don’t worry, when it’s finished SU will spring back to life again. I use it many times a day on all my VfSU and can honestly say it’s the one most important thing you can do to dramatically speed up VfSU total render times.

Matthieu’s “Remove C-G” script can be found below.

I really should nag someone about making that topic sticky as it’s a must-have for VfSU users IMO.

Hope this helps,
Jackson

Re: Why is 1.0074 so much slower?

First of all, Thank you so much for your answers!
I will try to follow your advices and make things work properly.
As you can probably tell, english is not my native language, so it’s hard for me sometimes to follow all the post on the forum to find the solutions for my problems, but i do my best!
Thanks again!

Re: Why is 1.0074 so much slower?

Do yiou actually notice a speed difference between using groups vs components?
I would find this very odd as internally, groups are defines exactly like components. The only difference is a flag that tells they are groups so whenever you edit a group copy it’s made unique. In ruby they are accessed from the same collection of ‘definitions’ as components. And when you copy a group around without editing it you see it has instances just like components.