Page 1 of 2
Procs comparisons
Posted: Mon Jun 27, 2005 9:29 pm
by def4d
This is a simple test i made to compare rendering speed between all of our computers.
Scene is 800x600 and 20 minutes, 2 threads, turned 3ds off while rendering, and a coffee.
Dell XPS Gen5 P4 HT 3,2 2Mo cache 1Go@533Mhz Sata (of course don't matter our Graphic cards!)
Here is the Max6 scene :
(removed)
Posted: Mon Jun 27, 2005 9:34 pm
by macray
Please let us know the results for all the computers!
Posted: Mon Jun 27, 2005 9:37 pm
by def4d
i mean mine, and yours
Posted: Mon Jun 27, 2005 9:52 pm
by zoppo
no max here ...
Posted: Mon Jun 27, 2005 9:56 pm
by Mihai
measure instead how long it takes to reach a certain sl. Otherwise we'll be judging by our eyes...
Posted: Mon Jun 27, 2005 9:58 pm
by ludenhud
maya user
Posted: Tue Jun 28, 2005 10:21 am
by macray
Cinema user. So post a unified dxf or wavefront-object. I can't use a max-object.
Let's say we calculate up to sample level 8 or 9 then post our times.
Posted: Tue Jun 28, 2005 5:18 pm
by Mihai
Yes, just post the mxs scene, but rendering should be done for a bit more than sl 9, maybe sl 15.
Posted: Tue Jun 28, 2005 5:27 pm
by def4d
Here it is!
(removed)
The target samples is 25, Maximum render time is 20mn, 2 threads.
My computer has reached level 9 for that time and settings, let me see results for yours!
Posted: Tue Jun 28, 2005 5:55 pm
by macray
Your settings are not compareable. I'm rendering on a laptop, so there is definitely no 2nd thread with me.
There are several people, who can't use 2 threads (or 1 is better, and to compare it to 4 threads will be interesting, too.)
So I propose to use a final sample level without given time to see all the different cpu's (including HT, dual, quad) reach the same level and compare about the time they needed.
Posted: Tue Jun 28, 2005 8:22 pm
by macray
I downloaded your mxs and rendered it. I can tell you that it finished after 20 minutes (record!
) and that I reached sl 8. But how long it would endure to reach another level... that's something I can't tell. So you still don't know how much faster your computer is.
You should link a new mxs with the time 1000 min, but a certain sample level to reach before it ends. I think 12-15 are enough. Else its going to take to long on some computers.
More information:
Dell Latitude D800
Pentium M 1.7 GHz
512 Mb Ram
Win XP Prof. SP1
screen resolution 1920x1200
Posted: Tue Jun 28, 2005 8:28 pm
by def4d
This zip contains the same scene, but 72000mn for rendertime, and 50 samples, 4 files for different threads settings : 1, 2, 4, & 8 threads.
(removed)
Posted: Tue Jun 28, 2005 8:30 pm
by Mihai
You don't need to provide different files for different threads or times.....
It's the same mxs file...
The benchmark should be:
How long did it take on your computer to reach n sampling levels?
Just decide what that sampling level should be, say 15.
Posted: Tue Jun 28, 2005 8:40 pm
by def4d
ok i'm preparing another one with 15 samples max, but can you explain for the threads?
Posted: Tue Jun 28, 2005 8:44 pm
by Mihai
you should really set the threads according to how many processors you have in your machine.
I'm not sure though if for Intel procs (with hyperthreading) if you should set it to 4.
Safest is if you have a dual intel machine, just set the threads to 4.
Anyway, when anybody wants to render your mxs, they can set the nr of threads to use themselves, so you can set it to whatever you want and upload it. If you set it at 4, they can specify 1 using the mxcl commands (explanation found in the docs), so it doesn't matter.
They can also specify the sl and time, so it they don't matter either what you set it to when you export the mxs.
The only things you need to set are the camera settings which can't be changed (for now maybe) using mxcl commands.