Quirks that slow down vray...

[Inhance Digital]

Having learned mental ray, brazil, and now (FINALLY) Vray -- which will be the last renderer I learn (because its going to Maya!) -- the script idea is brilliant. I find myself having to troubleshoot other artists GI work and if we had a script they could run before I even attempt to dissect their scene it would save allot of time.

And, well, time is pretty much everything!

[CCS]

So where is it already? It's been a month!

[devin]

Vlado, I'm a relatively new user of Vray (about 6 months) and I've got to say that a script like the one you’re talking about would be invaluable. Vray is a great program but for those of us just starting out it's extremely difficult to know which settings to optimize and 9 times out of 10 I just deal with long render times because I don't have the time to figure it out. This script would be a tremendous time saver to people like me and probably to most of the Vray community I expect. Do you have any idea if you will do this and when it might be available?

[HAWF]

yeah! there's a nice idea!
In sometimes we don't have so much time to do renders and more renders to tweak time vs. quality.

Cheers!

[jreynolds]

I don't recall seeing that thread in the past... Sounds like a great idea!

Jon

[duke2]

Beep. Still an awesome idea.

[joconnell]

Hmm - I wonder if there would be a way of giving you some kind of diagnostic readout to let you know what the noise threshold of the dmc or aa was? I was going to write a post asking Vlado about the relative expensiveness of each type of sampler - for example I was doing a simple AO element and it gave me quicker results to use less AA and vray dirt samples up around 60. Another car scene I worked on which was lit by lots of very large area lights (sometimes being bounced off cards) and vray blend materials with multiple levels of glossy came out better with a lower dmc and higher AA.

As Vlado mentioned there's no real rule that works for every scene but it seems aa is the only real way to clean up a scene - I suppose this makes sense since things like ao, glossy and area light subdiv rays could be shooting all over the place (similar to the wasteful nature of photon mapping) and ultimately aa is going to be dealing with a 2d converged grid of pixels. Is there such a thing as a ranking table of which subdivs are going to be the slowest? For example:

1. Glossy
2. Area Light
3. Depth of field
4. Motion blur

Again no doubt this is so scene subjective to the amount you're using of each and how much of the render frame it takes up. Would it be possible (and I'm not sure it'd even help) to have some kind of diagnostic mode that would give use the current noise threshold that each of the samplers is feeding back to the aa so that we could work out which are giving bad results and causing more aa samples to be used?

For example something like:

Glossy samples threshold - 0.04
Area light samples threshold - 0.01

So if we had our aa clr threshold set to 0.005 the glossies are going to be causing a lot more samples and cause the AA to work harder - some kind of readout might be a way of pointing us towards what's causing the issue. Again I'm not sure whether this is even the way it works or whether having the unified dmc system is going to make it hard to isolate the noisiness of each component of the render?

[sv]

Hmm - I wonder if there would be a way of giving you some kind of diagnostic readout to let you know what the noise threshold of the dmc or aa was? I was going to write a post asking Vlado about the relative expensiveness of each type of sampler - for example I was doing a simple AO element and it gave me quicker results to use less AA and vray dirt samples up around 60. Another car scene I worked on which was lit by lots of very large area lights (sometimes being bounced off cards) and vray blend materials with multiple levels of glossy came out better with a lower dmc and higher AA.

As Vlado mentioned there's no real rule that works for every scene but it seems aa is the only real way to clean up a scene - I suppose this makes sense since things like ao, glossy and area light subdiv rays could be shooting all over the place (similar to the wasteful nature of photon mapping) and ultimately aa is going to be dealing with a 2d converged grid of pixels. Is there such a thing as a ranking table of which subdivs are going to be the slowest? For example:

1. Glossy
2. Area Light
3. Depth of field
4. Motion blur

Again no doubt this is so scene subjective to the amount you're using of each and how much of the render frame it takes up. Would it be possible (and I'm not sure it'd even help) to have some kind of diagnostic mode that would give use the current noise threshold that each of the samplers is feeding back to the aa so that we could work out which are giving bad results and causing more aa samples to be used?

For example something like:

Glossy samples threshold - 0.04
Area light samples threshold - 0.01

So if we had our aa clr threshold set to 0.005 the glossies are going to be causing a lot more samples and cause the AA to work harder - some kind of readout might be a way of pointing us towards what's causing the issue. Again I'm not sure whether this is even the way it works or whether having the unified dmc system is going to make it hard to isolate the noisiness of each component of the render?

yeah this would be cool. Would also like to know what noise level you are actually at - might help with the age old monitor issue as well - i.e. you monitor might be too dark and hidding noise, when you view it on another monitor or another output, it might show some noise you didn't realise was there - would be good to know you have hit a 0.003 threshold, for example, if this is possible...?

[windowlicker]

Stumbled upon this thread when I was looking for troubleshooting some slow renders - and I feel I have to ressurect it.
A script like this would just be great. Even for us who are working proffessionally with vray every day, a script like this would be awesome. It'd save a lot of time, which is quite essential when on a tight deadline. So please, vlado, tell me you're working on this.

[dellis]

I'm all for the script too! It would be really nice to have the suggested settings from the script printed in a separate file or to the listener for comparisons.

[stevesideas]

Yep, the script sounds like a no brainer to me...I think you have the backing of most vray users

[schroeder]

I gotta second this - I can't think of a single thing that I would have done to increase my times the way they have, and Vlado, did this ever see the light of day? If so I'll comb the forums, but if not, my god this would be helpful.

[Fekta.studio]

+1 , do it Vlado...

[simmsimaging]

+1 It would be useful for all users IMO.

b

[Slazzo]

+1 Very usefull script, how or where will be posted in case, just to be ready