?

Log in

No account? Create an account
 
 
22 May 2011 @ 01:08 pm
Renders  
So, I've started working with LightWave3D again on the new iMac.

It has a few issues which will probably lead me to upgrade it at some point -- it has a tendency to run out of memory for no good reason (presumably because it's a 32-bit executable dating back a number of years -- it tends to run out of memory with large shadow maps and images when it's only using maybe 2GB out of the 16GB I have available. Version 10 is presumably 64-bit, however -- it seems to require Snow Leopard. It's still its cranky old self, too, with the occasional crash -- it's something to be tolerated, since I can't afford/justify buying Maya).

At any rate, since I've sort of lost the originals of some of my older renders (in that I can never re-render them in larger versions or anything, most of these are from around a decade ago), I've decided to make new versions (not exact copies, more new renderings inspired by the old one). The first one I did was "Game" (old one on the left, new one on the right):

 

[Click through for larger versions.]

The old one was rendered on 10 Dec 2000 (or possibly a day or two earlier). Total render time was 2:56:24 on... Maybe a P3-800? ...at 1280x960.

Statistics:

31 objects
31791 points
27768 polygons
1 light
33 surfaces
0 image maps

The new one was rendered today (22 May 2011). Total render time was 2:15:51 on my i7-2600 (3.4GHz, 4 cores hyperthreaded, i.e., 8 cores to the OS) at 5120x3840.

Statistics [updated]:

71 objects
159964 points
130257 polygons
1 light
18 surfaces
0 image maps

The big difference between the images is the use of radiosity on the new picture; both use depth-of-field, volumetrics and a lot of reflection. I suspect if I'd tried to notch up the quality effects to the level used on the newer version on the old computer it would have taken weeks -- assuming it actually had the memory to handle it at all. I suspect a newer version of Lighwave would do slightly better -- while the bulk of the render time was calculating radiosity, which was equally distributed over the cores, it's less efficient rendering polygons and wastes at least 25% (often more) of the CPU with it's not-entirely efficient thread scheduling.

And after all that... I'm not really sure the new version is actually an improvement (and my attempt to put a robotic hand in there just didn't work -- it looked terrible).