[Leonardo]

... so which one is ~.~ and with one if fry?

[zoppo]

very interesting test. i'd like to see more of that kind of comparison, especially made by different artists.

@leo: fry is on the left side.

[Miles]

fry is on the left side.

Sinistro......

[Frances]

Just FYI, ajnabee used this situation as a case study to produce an essential and comprehensive scene optimization quide for new (and some not-so-new) users. The user broke nearly every rule in the book regarding producing efficient scenes for use with spectral raytracers. The main ones being that there were materials whose weight values were >100 (fryrender does not automatically clip weight values, it is up to the user), emitter geometry in some cases was over 4k polys, and there was emitter geometry intersecting other objects. The study resulted in a clean render in less than 3 hours. The case study is published in pdf form, and anyone with forum access is welcome to check it out.

[Leonardo]

Where is that pdf...

3 hours sounds a lot better

[numerobis]

Just FYI, ajnabee used this situation as a case study to produce an essential and comprehensive scene optimization quide for new (and some not-so-new) users. The user broke nearly every rule in the book regarding producing efficient scenes for use with spectral raytracers. The main ones being that there were materials whose weight values were >100 (fryrender does not automatically clip weight values, it is up to the user), emitter geometry in some cases was over 4k polys, and there was emitter geometry intersecting other objects. The study resulted in a clean render in less than 3 hours. The case study is published in pdf form, and anyone with forum access is welcome to check it out.

Yes, very interesting read.
but as the scene is the same for both engines, it seems that mxw works better with not so highly optimized scenes (intersecting emitters, ...)
and you could optimize the mxw scene too.
also the second 3h render was on a much faster machine than the first test, if i remember correctly...

[Frances]

Yes, very interesting read.
but as the scene is the same for both engines, it seems that mxw works better with not so highly optimized scenes (intersecting emitters, ...)
and you could optimize the mxw scene too.
also the second 3h render was on a much faster machine than the first test, if i remember correctly...

It does beg the question. Once it is out of beta, it is likely to be more forgiving. As you can see, this advice is not new to most maxwell users who have been in since the alpha/beta stages.

The case study 15 hour render was on a Dual Xeon 2.33 ghz. The 2 hour render was on a MacPro Dual Xeon dual-core 2.66. So yes, that is a significant difference in specs. Especially since the Woodcrest does not have a very good architecture. I would like to see both scenes run with the same computer specs. Not sure why that didn't happen.

Edited to add: I looked at the Speedtest table and found a benchmark score for a MacPro Xeon dual dual-core 2.66 (~45 min) and a Dual Xeon 2.8 (~112). I didn't see a 2.33 on the table. Anyway, if the performance of the MacPro was 3x that of the Woodcrest, you could divide the 15 hour render by 3 and get 5, to even the playing field? There is still a significant gain by optimizing. But I do wish that both scenes could be rendered with identical specs. Math is not my thing.

Leo, I'm sure someone can pm you the pdf link.

[Leonardo]

something else to note is that the maxwell scene seems to be free of noise long before those 15h

[Frances]

something else to note is that the maxwell scene seems to be free of noise long before those 15h

I only saw where the maxwell and fryrender scenes both rendered for 15 hours and that mw reached SL 19.58. I can only guess that he did the mw render first and stopped it when he liked it. Then he let fryrender go for the same amount of time and the results were certainly not favorable.

[Leonardo]

Perhaps I'm a little biassed... but I think some of us are asking too much from the frying pan

[zoppo]

The case study is published in pdf form, and anyone with forum access is welcome to check it out.

i'd like to read it. which forum do you mean?

[Frances]

The case study is published in pdf form, and anyone with forum access is welcome to check it out.

i'd like to read it. which forum do you mean?

The official fryrender forum.

[zoppo]

thanks, i got it. very usefull!

[KurtS]

http://www.fryrender.com/sh_images/gall ... 20cars.jpg