Page 2 of 3

Posted: Thu Mar 24, 2005 6:05 pm
by lluistar
I think the bug affects dual processor computers.

Maxwell counts the time that every processor needs to render a scene. So a quad computer will display four times the real time it needs to render a scene.

Does it happen to the dual Xeons out there?

Posted: Fri Mar 25, 2005 12:50 am
by victor
Yes, the Maxwell clock is wrong in Mac OSX, so this is probably the reason of all this topic. Sorry guys for spending time trying to understand it.

Posted: Fri Mar 25, 2005 4:43 am
by lluistar

Yes, 4 threads are faster than 2 threads and 2 threads are faster than 1 thread.

But at a loss of quality. even if all reached 10 samples the noise in the 4 threads is more visible than just 1 thread.

Posted: Fri Mar 25, 2005 5:02 am
by andronikos916
Voxelisation uses only 1 cpu like in Win XP.


Posted: Fri Mar 25, 2005 9:06 am
by N&B
Well, like I said earlier in this post, I'm doing stupid performance tests to occupy myself, waiting for the plugins for 3D Mac apps to be released :lol: :oops:

by the way, where are mx-monitor and mx-manager for network rendering? I only found mxcl in the mac package.

And yes, the 4 threads image has more noise, but as it's faster than 1 th, maybe level 11 can be reached in the same time. So wich one would be the noisiest?

EDIT: And the answer is......YES!

with 4 threads, M~R reached level 11 in 33min 46s (stopwatch time)!

with 1 threads, M~R reached level 10 in 39min 10s (stopwatch time)

Posted: Fri Mar 25, 2005 1:39 pm
by N&B

Posted: Fri Mar 25, 2005 4:05 pm
by lluistar
looks good. Nice test

Posted: Fri Mar 25, 2005 6:06 pm
by The Pixel Artist
Very interesting thread (topic that is).

Just a quick test for us Windows version users, it looks like it works for us as well.


Tested it to 8 samples, first with 2 threads (I have a hyperthreaded P4 550), and then setting threads to 4. I couldn't tell any difference other than noise patterns, but it appears the noise is the same. Point being the 4 threaded version finished 40 sec. faster. It may have made a much bigger difference time wise at a higher samples value. Time value does work properly on the PC version. This is very interesting and weird.

Posted: Sat Mar 26, 2005 10:13 pm
by MetinSeven_com
I've also noticed the difference between real time and indicated Maxwell render time. This is especially apparent when you set the mxcl.exe process to a lower priority and start actively using other software while rendering. This means the Maxwell time indicates the actual intrinsic time that qualifies as true rendertime and is not merely a clock starting to tick from the moment the render starts.



Posted: Sat Mar 26, 2005 10:30 pm
by tom
I agree Metin, low switch is usefull when rendering for fun or working behind but if we're doing test we must run nothing more than maxwell at the same time even we're running it at normal. Testers, please pay attention to this.

Posted: Sun Mar 27, 2005 6:49 pm
by N&B
yep, during all tests only Maxwell was running, taking 85-90% cpu usage (due to unix background tasks and multi tasking management of OSX)

Posted: Mon Mar 28, 2005 10:29 am
by mgroeteke
i cannot test maxwell, because the installation is simply not running, but this 'thread' thing is well known by lightwave users. on double processor G5 powermacs 4 or 8 render threads are normally faster than 2.


Posted: Mon Mar 28, 2005 11:16 am
by N&B
thanks for the tests guys.

It seems that G5 processors aren't performing so well compared to amd and intel ones....
Serious lack of optimization, probably...
I saw that a single amd64 3500+ better performs than a dual G5 2ghz

I'm wondering if buying some amd64 for rendering wouldn't provide the best bang for the bucks.
Do you know if it's possible to take control of linux render nodes from OSX?

Posted: Mon Mar 28, 2005 11:42 am
by lluistar
if you do a top command in the terminal while rendering a maxwell scene gives you:

rendering @ 1 thread >> 3 threads (1 render + 1 display + 1 terminal)
rendering @ 2 thread >> 4 threads (2 render + 1 display + 1 terminal)
rendering @ 3 thread >> 5 threads (3 render + 1 display + 1 terminal)
rendering @ 4 thread >> 6 threads (4 render + 1 display + 1 terminal)
rendering @ 8 thread >> 6 threads (4 render + 1 display + 1 terminal)

I don't think that above 4 threads you get any improvement.

Besides the G5 performance compared to intel or ADM, this is the first release of the ALPHA stage for Mac. Let's hope for better optimization.

Command Line

Posted: Tue Apr 26, 2005 10:50 am
by Nigel Baker
Hello there,

Can someone tell me where is this command line written.
Is it the Terminal Window or the command the "cl options" in the LW8 Maxwell render Dialogue Box.

Still confused.