(Fwd) Re: Athlon vs Pentium III

Simen Timian Thoresen simentt at dolphinics.no
Thu Feb 22 00:46:46 PST 2001


Sorry about the double delivery. I did intend to send it to the list, not just 
reply to the sender.

> >   memory bound
> >   CPU bound
> >   disk I/O bound
> >   network I/O bound (in its parallel version)
> >   cache-size sensitive (tied into memory/cpu bottlenecks, usually)
> > 
>  Excuse me. If my application is mpi-povray-alike, or it happens
>  to be mpi-povray, in what situation i will be? what is my best
>  choice in Cost/Performace ratio for now? Dual PentiumIII or
>  Single Athlon? thanks in advanced.

I might be completely off my rocker here, but I would believe mpipov to be
cpu-bound.

The version of mpipov I saw renders a block of lines on each process, and 
there is no mpi communication between setup and final data except 
housekeeping, so the network will probably be bored most of the time.

Each process renders one pixel after another, so unless you have a really 
advanced scene/texture stack, I would guess you don't have to worry much 
about cpu cache. 

Also, unless your scene is very complex, and the segment rendered by each 
node is very large I would think that memory should not be a problem either.

That leaves CPU speed (and, to a lesser degree, memory speed) as the 
bottleneck.

Yours,
-S

--
Why are there penguins embedded in my toaster?

Er det rart?
The gnu RART-project on http://valinor.dolphinics.no:1080/~simentt/rart




More information about the Beowulf mailing list