View Full Version : Render Speed comparison Help needed

06-25-2003, 04:08 PM
With all the speculation about G this and P that I decided to do a few render tests on some of the big boys at my company.

Is there any bench marking/render test file that I could use. On a site I could download or on the LW cd.



06-25-2003, 04:59 PM
yep... LW cd content directory/scene/benchmark folder.
some say they are dated and not "real world" but thats all there is.

06-25-2003, 05:50 PM
yeah, found it after visiting that benchmark site.

guess it's enough to compare speeds

onlything extra they could put in would be hypervoxels and particles


06-25-2003, 05:58 PM
There are 10 benchmark scenes in the Lightwave content. I'd start there. Oh, and here are my results, just opened the scenes and pressed F9:

DOF.lws: 18.2 sec.
Radiosity_ReflectiveThings.lws: 73.5 sec.
Radiosity_Things.lws: 7.3 sec.
Raytrace.lws: 162.7 sec.
ShinyArtBalls.lws: 386.8 sec.
Skull_Head_Newest.lws: 239.3 sec.
Textures.lws: 10.1 sec.
Tracer-No-Radiosity.lws: 665.7 sec.
Tracer-Radiosity.lws: 1062.6 sec.
ZBufferSort.lws: 13.2 sec.

System specs:
Quicksilver G4 2x800 MHz
1.1 GB RAM
OS 10.2.6
LW 7.5
Segment Memory Limit: 32

I think you need to make sure you compare apples to apples. Do the same thing(s) to the scenes when you test, be it preview, F9, or whatever. Make sure it's standardizable, (like F9's render time display).

06-25-2003, 10:40 PM
damn saw this too late. I'll run the renders when I get back to work. I'll try the dual gigs and dual 1.5 gigs

Hmm... does any one know the story on segment size? I setup mine to 256MBs. what should it be?

Oh how many threads do you setup for dual G4s

thanks for the info.


06-26-2003, 02:06 AM
don't change the segment memory for the benchmarks, that would be an innaccurate comparison.

Segment memory should be just enough to render in one segment, no more. In some cases you may need to set it lower, for example really high res images for print, where there isn't enough memory for the whole image.

Thread settings depend on the machine, the scene you're rendering, and what version LW - do test renders for each scene to see what's fastest.

I used to use 8 threads 95% of the time, somewhere along the way 2 became faster, I don't know when it changed - anybody notice that?