[eldo]

my first wish here,..

what i want i a constantly updated rendertime prediction..

for that you would just have to reached SL 4 in order to get an equation out of the values. i guess rendertime-equation is quadratic.. so we need at least 3 conditions in order to get the equation..

f(SL) a SL^2 + b SL +c

SL = Samplinglevel
f(SL) = rendertime [min]

if we have for example:

f(1) = 8 (i know this is very inaccurate)

f(5) = 50

f(7) = 116


then we get the following:

a + b + c = 8

25a + 5b +c = 50

49a + 7b +c = 116

whereas c is as far as i remember 0.. haven´t checked that yet.. too tired so also forgive me if there are any mistakes on this.. im sick for over a week now

then we can build a system of equations out of these 3 equations.. and can solve that in order to get an equation that is correct for the current rendering..

so we can in round terms predict rendertimes for certain Samplinglevels and also vice-versa..


just an idea

[tom]

nice trick...
here's my contribution to this also..
http://www.maxwellrender.com/forum/view ... 5&start=15

so the rendertime maybe calculated after some sampling...

[Mihai]

My brain hurts......

Wouldn't the most accurate method be to first render a tiny preview of the scene to get the time ratio between samples, and that would give you the most accurate prediction? Because I think this time ratio can vary a lot between scenes.