I did a little test rendering with a scene containing about 53 meg (peak
memory used) of mostly meshes & spheres dotted along splines... (many
splines.)

There are 3 versions of the main scene file used:  The original on the p3
laptop, and a 2nd version which has no functional changes but many of the
original's non-functional lines like "#declare dummy_var = -1;" modified
with different float values to get it to work on the core2duo machine. (See
note)  The 3rd version is identical to the 2nd except for the name so that I
could render two instances at once.  All #includes are identical.

All are 800x600 +a0.3, none using 64bit versions:
   TIME   Ver          Proc
A) 51:28  3.6          pentium 3m 1ghz  (no real surprise here)
B) 6:35   3.7 b18      E6700 @3.2
C) 5:38   3.6          E6700 @3.2
D) 6:38   3.7 b18      E6700 @3.2 re-rendered
E) 6:06   3.7 b18 sse2 E6700 @3.2
F) 5:07   3.7 b18 sse2 E6700 @3.2 with +bm2
G)16:19   3.7 b18 sse2 P4 3.6 w/HT (prescott) with +bm2
H)13:32   3.6          P4 3.6 w/HT (prescott)

The following two renders were "something.pov" and "Copy of something.pov"
started in two separate instances of 3.6 within 2 seconds of each other.:

I) 5:26   3.6          E6700 @3.2
J) 6:20   3.6          E6700 @3.2

Done again:
I) 6:33   3.6          E6700 @3.2
J) 5:49   3.6          E6700 @3.2


A two-instance rendering with 3.7:
K)13:01   3.7 b18 sse2 E6700 @3.2 with +bm2
L)13:05   3.7 b18 sse2 E6700 @3.2 with +bm2

My overall conclusion is that I have one of those scenes which 3.7 doesn't
handle as quickly as 3.6.  It's interesting how much more striking the
difference between 3.6 & 3.7 is on the core2 than the prescott.  I haven't
tried 3.7 on the p3.



Note: There seems to be something in both 3.6 & 3.7 beta which kicks in an
error "Parse Error: Identifier expected, incomplete function call or spline
call found instead."  The line in question may look something like #local
Spline_A = Array_of_Splines[Ctr]; or be a line calling a macro using a
spline identifier as one of the parameters.  It will work many times over
in a loop, but may fail at some repeatable time during the overall parsing.
Repeatable IF it's on the same machine.  Adding something like #local
dummy_var = 97981316123; either inside or outside the loop seems to affect
whether the render will get through parsing.  I'm always happy to see the
Render Cleanup Warning about the experimental status of splines because
that means it's made it through.

Post a reply to this message