how to estimate the time of rendering .

[mustan9]

You might be able to do a simple statical calculation based upon a history of renderings.

Capture the render time, face count, render settings, etc.. etc.. to a log file.

Then when estimating render times look up past render times and calculate an average.

It won't always work correctly, but the more data you collect. The better the guess.

[hblan]

good idea , but it need too much time . and it looks this work should be done by discreet . :)

[handiklap]

good idea , but it need too much time . and it looks this work should be done by discreet . :)

I don't think you're considering the incredible variability of a scene with reflection/refraction, particles, caustics, photon emission, hair - the list goes on; that would be a ridiculously complex algorithm just to make a marginally accurate prediction of a single frame render time. Nevermind there are outside variables as well - processor speed, ram speed and capacity, system load, etc.

[mustan9]

that would be a ridiculously complex algorithm just to make a marginally accurate prediction of a single frame render time

It is not a ridiculously complex algorithm. It's the basis of statistical calculations to figure out if a given sample belongs to a given population, or to tell if a sample variable came from a given population.

You could easily create a MaxScript that recorded render activities, and then cataloged those jobs into 10 different populations. Those populations could be 1, 2, 5, 10, 15, 30, 45, 60, 120, and 240 minute renderings. The larger your sample base becomes the better the sample is, and it's possible with statistics to tell if your sample represents a given population.

Once enough data is collected. You can then check the properties a scene has, and using statistics check if that scene belongs to one of the 10 populations. Thus, you figure out how long that scene would take to render.

It's actually very easy to do with statistical mathematics.

[soshiant]

It is not a ridiculously complex algorithm. It's the basis of statistical calculations to figure out if a given sample belongs to a given population, or to tell if a sample variable came from a given population.

You could easily create a MaxScript that recorded render activities, and then cataloged those jobs into 10 different populations. Those populations could be 1, 2, 5, 10, 15, 30, 45, 60, 120, and 240 minute renderings. The larger your sample base becomes the better the sample is, and it's possible with statistics to tell if your sample represents a given population.

Once enough data is collected. You can then check the properties a scene has, and using statistics check if that scene belongs to one of the 10 populations. Thus, you figure out how long that scene would take to render.

It's actually very easy to do with statistical mathematics.


"It is a ridiculously complex algorithm!" is exactly true!!!
If such a thing was sth practical it would become a part of every render engine.
Maxscript knows nothing about how renderer do rendering(ie MR )or when renderer doing optimizations.
you can use statistics on constant variables not on floating variables though.
(it is consums a lot of cpu to calculate such a thing)
imagin renderer is ray tracing a scene with this specs:
a plane
a light with ray traced shadows on
a glass
a camera
animate tha cam to move from left to right.
is the time for every frame constant?
do it again and compare the results.
its enough to say that the calculated time depends on the position of the camera(and other parameters of course but one is enough to proove here!)
it shows that the TIME eaquation is NOT DETERMINISTIC.

[handiklap]

USE STATISTICS!!1one

takes too long

Well, it's pretty much the only way without a lot of heavy calculations that would take a lot longer.

No you're wrong, USE STATISTICS!!1one

So how would you estimate a render time with a population of 0? I believe the problem is with your reading comprehension, not with our understanding of elementary mathematics. I wasn't suggesting that statistics in this context are difficult or complex. I was actually trying to explain that statistics are, in fact, the only efficient way to do so. But kudos to your superiority complex.

hblan - Really, the only efficient way to do this is to do what has been suggested about logging statistics to a file and then comparing your scene to those statistics. With enough samples, you could set it up pretty nicely and even have an idea of how adjustments to particular systems will impact your render times.

[CGTalk Moderation]

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.