Virgilio Vasconcelos: Animator - Portfolio, Demo Reel, Animation, Blender...

Oct

Ao ver que as otimizações no render interno do Blender feitas pelo usuário conhecido como Jaguarandi foram migradas para o trunk do Blender 2.5, resolvi fazer um pequeno teste:

Usando o arquivo oferecido pelo conhecido Blender Render Benchmark, decidi comparar as velocidades de render entre a versão atual (2.49b) e a em desenvolvimento do 2.5 (svn revision 23930). Obviamente a velocidade do render no vídeo é comprometida por alguns fatores: não é comum renderizar a mesma cena em paralelo, nem mesmo gravar o processo em qualidade HD. Mesmo assim, achei interessante ver como - mesmo sob tais condições - a versão 2.5 conseguiu renderizar a cena duas vezes antes que a versão 2.49b terminasse.

Em condições normais (ou seja, sem outros processos pesados rodando em paralelo, e com 2 threads dedicadas ao render), o tempo de render na versão 2.49b foi de 01m03s, enquanto a 2.50 terminou em incríveis 25s! :)

Realmente a nova versão do Blender chega para melhorar em todos os aspectos o programa que já é excelente. Muito obrigado e parabéns a todos os desenvolvedores!

E para quem tiver curiosidade, meu sistema é um Core2Duo 2.53 GHz, Ubuntu 9.04 64 bits. Gravado com o RecordMyDesktop e editado com o Blender.

(15) Comentários

19/Oct/2009

Pablo Lizardo disse:

Nice music and video!
It's been done in blender? (Edit: you say so!)
Thanks for the benchmark!!!! 2.5 rules! :D

19/Oct/2009

lucky disse:

haha, nice test!

19/Oct/2009

stephen disse:

my guess is that RecordMyDestkop shared a core with 2.49. try it like this:

time blender -b input.blend -f 1
time blender2.5 -b input.blend -f 1

19/Oct/2009

Virgilio disse:

Hey guys!

Thanks for the feedback!

@stephen: I did this test before I went recording. ;)

19/Oct/2009

Juan Pablo Bouza disse:

Jaja!! Great!

19/Oct/2009

Amarneethi disse:

"...let's be polite and wait for 2.49b to finish its job :) " thats crazy...
Try rendering separately with 2.5 and 2.49. So that we get more accurate results... We dont knw which core the system is using much and which one not... Anyway just a opinion....

19/Oct/2009

stephen disse:

Virgilio --

Thanks for verifying it. I'm sure i'm not alone in how impressive the blender project/contributers are. I am blown away that they do so much and keep the app as responsive as it is.

Thanks for making that film!

20/Oct/2009

Dm7 disse:

Wow! I had to laugh while you throw in "listening to what you have to say" messages, while waiting for both rendering to finish their job! That really cracked me up and it really made video way more fun to watch! Thanks! And thanks for doing it the subtitle way too... since I'm deaf. Very hard nowadays to follow tutorials because of that video trend... T_T so I give you a bonus star for that! Very awesome comparasion, considering that you got recorder running as well!

20/Oct/2009

AMGO disse:

Really wonderfull Thanks Jaguarandi great job, I love Blender more than ever. ja

20/Oct/2009

Reyn disse:

Wow! That was fast! And you even tried rerendering that of the 2.5! Hahaha! That was just awesome! Thanks, man.

(I think I'll have a try at test rendering in 2.5 too)

-Reyn

20/Oct/2009

nikola disse:

I am anxious if making 2.50 be the active window is not influencing the speed - the inactive window could be getting penalties at process priority

20/Oct/2009

Virgilio disse:

Hey guys, thanks for your feedback! :)

@Dm7: I'm very happy you liked it and that my subtitles were helpful! :)

@nikola: actually I did this test a couple of times before I decided to record it to put online. I believe the real test is the one I did without any concurrency: this way Blender 2.5 was even faster. The video is more to illustrate that than a scientific experiment. :)

Cheers

21/Oct/2009

Martín Eschoyez disse:

Virgilio, nothing more to add, but i wanted to say that this is a really good video. NIce music, and i've really LOL to those comments. Thanks you and Jaguarandi!

21/Oct/2009

therealnoz disse:

Very nice. This will help a lot!

28/Apr/2010

agilliorink disse:

Hi...... post good :)