Newsgroups: comp.graphics.rendering.raytracing,comp.parallel.pvm
From: Thomas Willhalm <Thomas.Willhalm@uni-konstanz.de>
Subject: Re: PVM PoV-Ray speed
Organization: University of Constance, Germany
Date: Thu, 17 Jul 1997 09:48:35 +0200
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Message-ID: <33CDCE53.944@uni-konstanz.de>

Drake Diedrich wrote:
> 
> In article <slrn5spjoc.8tp.thomas@trust.aztec-design.co.uk>,
> Thomas Molesworth <thomas@bass.almac.co.uk> wrote:
> >Has anyone had any experience running PVM PoV-Ray on x86 based systems? I
> >am running it on a network of 3 Linux (=UN*X clone) systems, Pentium-133,
> >486dx2-66 and 486dx33, but I find that the speed is slower than rendering on
> >the single Pentium system. Is this right?
> 
>    The pvmpov tiling seems to interfere with one of the POV3 speedups
> (light- or vista- buffer).  Look carefully at the statistics section
> in your output files, and you'll notice pvmpov is doing 4 times as
> much work as regular povray.  Instead of using square tiles, use tiles
> the full height of your picture, and only 5 or 10 pixels wide.
> Something like +NW10 +NH480 works well on a 640x480 render.

This is not true in my case: If I render the skyvase with +NW10 +NH480
our system needs 39-49 seconds, while rendering with +NW32 +NH32 takes
30-34 seconds.

Bye.

ThW

-- 
Thomas.Willhalm@uni-konstanz.de
http://www.informatik.uni-konstanz.de/~willhalm/
Tschieses lavs ju