Forum: Poser - OFFICIAL


Subject: Poser 8... First Impressions

ziggie opened this issue on Aug 04, 2009 · 617 posts


Keith posted Thu, 06 August 2009 at 9:38 AM

Quote -  Ahh, this thread is pretty big, so forgive me if these questions have already been addressed, but after scanning all pages I still have some doubts about upgrading:

  1. I've noticed some talk about P8 rendering faster than PPro, even though 8 doesn't have the 64-bit render engine. Considering that this was the main reason I upgraded to Pro, just how much better is P8 supposed to be at it? I just want faster renders; I was unfortunately extremely spoiled by XSI's mental ray implementation, but obviously, I can't really afford XSI for home use.

I did a test last night.  I run a quad-core Vista-64 with 8 GB on the board.

I loaded up a scene I'd done a few weeks ago.  It's an interior set, 4 gen 3 Daz figures with skin textures that included reflection (so they looked like metal statues, lots of curves to bounce around).  Three lights, a spotlight outside pointing in a door to simulate the sun and two interior point lights at the ceiling to simulate internal lighting.  There's glass with refraction/reflection scattered around, some materials (like the floor) with some reflection, lots of places for AO to be working.

Because of the camera angle, much of the reflective and refractive material was in the upper half of the image: this becomes important.

In both programs (Pro and P8), settings were 4 raytrace bounces, 50 irradiance, 5 pixel samples, min shade of .5, bucket size of 32.  The sun was the only light with a shadow, raytraced.

The differences between the programs: gamma correction (2.2) on Pro, in P8 HSV tone mapping with a 2.2 exposure.  The other difference was the lights: in both the sun was set at 140%, constant (old standard Poser), each interior point light at 50%.  In Pro, obviously, they were constant while P8 was set at realistic (ie square) falloff.

I knew this would make it too bright in Pro, but I wanted to check for speed.

Both programs were tested the usual way: render and canceled to make sure the texture maps were loaded, then rendered and timed.

Pro: 16 minutes, 30 seconds
P8:  13 minutes, 40 seconds

Of course, having the light settings as they were made P8 look better, but that was expected.  The real killer was how they were rendered.  In Pro, the image was "half done" according to the progress bar (and two cores idle) at about the 6 minute mark, because the lower half of the image had significantly fewer complicated reflected/refracted things to deal with.

So this is a real-world (ie, type of image I'd likely render for whatever) that was significantly faster, even without the renderer being able to take full advantage of my machine.  I'd suspect that in pretty much any case where the materials and objects aren't equally distributed across the image area, so that no quadrant/half of the image is more "complicated" than any of the others, you'll see that same speed increase just because of the more rational utilization of a multi-core machine.