On Mon, 27 Sep 1999 14:48:22 +0300, Margus Ramst <mar### [at] peakeduee> wrote:

>Actually, POV depends pretty much on raw MHz. AMD's FPU might be a factor, but I
>don't think the classic Pentium has a much better unit.

  The earlier Pentium fpu was much better than both the units on the 486DX

486DX2-66 and my Skyvase.pov rendering time (640w x 480h, +a0.3) went from
32 minutes down to 15 minutes, which is more than the clock speed
improvement should show. AMD has greatly improved floating point performance
of its later K6 and K7 processors but I don't know how they compare with the
latest Pentium IIIs using POV-Ray.

  POV-Ray does work the fpu quite thoroughly. This is why most common
magazine benchmarks often don't apply if one is looking for the fastest
machine to run POV-Ray. They often use a test suite of programs that do not
perform as intensive floating point math as POV-Ray does. The converse if
also true - a machine that runs POV-Ray fast does not necessarily have the
good video performance that is required of fast 3d game play. Most test
suites take video performance into account but POV-Ray doesn't care if you
have a video card installed.

  The POVBench site at http://www.haveland.com/povbench/index.htm is a
database of user statistics on various machines. It might just be me but I
have a difficult time trying to make sense of so many entries. Many of these
personal computers are home-built, which may affect how fast they can render
Skyvase. Quite a few entries show that the submitted results use custom
compiled versions of POV-Ray. Without knowing exactly what their POV-Ray
source contained these results cannot be directly compared with others. I
mean, why would someone running Windows choose to compile their own
executable, unless they've modified the source?

-- 
Alan
http://www.povray.org - Home of the Persistence of Vision Ray Tracer

Post a reply to this message