POV-Ray : Newsgroups : povray.beta-test : SMP beta 16 Positive Response threads question Server Time
28 Jul 2024 20:20:39 EDT (-0400)
  SMP beta 16 Positive Response threads question (Message 1 to 3 of 3)  
From: Andycadd
Subject: SMP beta 16 Positive Response threads question
Date: 21 Oct 2006 09:20:00
Message: <web.453a1e42838b6a38fedbe4b90@news.povray.org>
I use a dual 2.4 Xeon Hyperthreading 400mhz FSB

SSE is 15% faster consistently
2 threads is 1-1/2% faster than 4

Is this normal?

Is there a circumstance when SSE is not faster?
The output png files look identical and have the same size.

Is there a circumstance when 4 threads is faster than 2?
The output png files look identical and have the same size.


Thanks for all your work.


Post a reply to this message

From: Chris Cason
Subject: Re: SMP beta 16 Positive Response threads question
Date: 21 Oct 2006 11:28:58
Message: <453a3cba$1@news.povray.org>
Andycadd wrote:
> I use a dual 2.4 Xeon Hyperthreading 400mhz FSB
> 
> SSE is 15% faster consistently
> 2 threads is 1-1/2% faster than 4
> 
> Is this normal?
> 
> Is there a circumstance when SSE is not faster?
> The output png files look identical and have the same size.
> 
> Is there a circumstance when 4 threads is faster than 2?
> The output png files look identical and have the same size.

How are you timing this? Are you looking at the CPU time or the elapsed time?

Be aware that Windows lies about CPU time used when hyperthreading is
enabled, and there's little we can do about it right now. It will typically
show something like double what it should show, provided all CPU's (real and
virtual) are used.


Post a reply to this message

From: Andycadd
Subject: Re: SMP beta 16 Positive Response threads question
Date: 21 Oct 2006 18:00:00
Message: <web.453a98244e249c10fedbe4b90@news.povray.org>
Chris Cason <del### [at] deletethistoopovrayorg> wrote:
> Andycadd wrote:
> > I use a dual 2.4 Xeon Hyperthreading 400mhz FSB
> >
> > SSE is 15% faster consistently
> > 2 threads is 1-1/2% faster than 4
> >
> > Is this normal?
> >
> > Is there a circumstance when SSE is not faster?
> > The output png files look identical and have the same size.
> >
> > Is there a circumstance when 4 threads is faster than 2?
> > The output png files look identical and have the same size.
>
> How are you timing this? Are you looking at the CPU time or the elapsed time?


Elapsed times

weird that 2 threads is faster than 4

I got the same SSE improvement in beta 15
but those were all 4 threads (default) I think

no other stuff running   100% priority


>
> Be aware that Windows lies about CPU time used when hyperthreading is
> enabled, and there's little we can do about it right now. It will typically
> show something like double what it should show, provided all CPU's (real and
> virtual) are used.


Post a reply to this message

Copyright 2003-2023 Persistence of Vision Raytracer Pty. Ltd.