POV-Ray : Newsgroups : povray.general : Ver. 3.6 speed Server Time
3 Aug 2024 02:23:38 EDT (-0400)
  Ver. 3.6 speed (Message 1 to 10 of 13)  
Goto Latest 10 Messages Next 3 Messages >>>
From: Groucho
Subject: Ver. 3.6 speed
Date: 21 Jun 2004 17:34:24
Message: <40d75460@news.povray.org>
Comparing the render time of the same model with versions 3.5 and 3.6 I
always get that 3.6 takes around 30% more time to render than 3.5.

I'm using the Windows version on a P4 2.4GHz with 512MB and I always tested
with files that doesn't need to page memory.

There is anything to tune? Is it normal?

Thx

Groucho


Post a reply to this message

From: Apache
Subject: Re: Ver. 3.6 speed
Date: 21 Jun 2004 19:18:06
Message: <40d76cae$1@news.povray.org>
Could you be a bit more specific?

What kind of scene is it?


Post a reply to this message

From: Groucho
Subject: Re: Ver. 3.6 speed
Date: 22 Jun 2004 03:15:46
Message: <40d7dca2@news.povray.org>
It happens with all type of scenes but I'll give you an example. It uses
transparencies, only one simple light and radiosity.

----------------------------------------------------------------------------
-------
global_settings {
  assumed_gamma 1.0
  radiosity{}
}

camera {
  location <0,0,-3>
  look_at <0,0,0>
}

light_source {
  <0,0,0>
  rgb 1
}

#declare r=seed(2);
#declare i=0;

#while (i<100)
 #declare x1=2*rand(r)-1;
 #declare x2=2*rand(r)-1;
 #declare x3=2*rand(r)-1;
 #declare r1=sqrt(x1*x1+x2*x2+x3*x3);
 #if (r1<1)
  sphere  { <x1,x2,x3> rand(r)*0.2
   pigment {rgbt <rand(r),rand(r),rand(r),rand(r)>}
   finish { phong 0.2
    phong_size 10
    ambient 0.05
   }
  interior {ior 2}
  }
 #end
 #declare i=i+1;
#end
----------------------------------------------------------------------------
--------------------
Cheers

Groucho.


"Apache" <apa### [at] hotmailcom> wrote in message
news:40d76cae$1@news.povray.org...
> Could you be a bit more specific?
>
> What kind of scene is it?
>
>


Post a reply to this message

From: ABX
Subject: Re: Ver. 3.6 speed
Date: 22 Jun 2004 03:46:28
Message: <v4ofd0tp33f9t6vphsbd7449jchlhhim65@4ax.com>
On Mon, 21 Jun 2004 23:34:46 +0200, "Groucho" <gro### [at] tripodcom> wrote:
> Comparing the render time of the same model with versions 3.5 and 3.6 I
> always get that 3.6 takes around 30% more time to render than 3.5.
>
> I'm using the Windows version on a P4 2.4GHz with 512MB and I always tested
> with files that doesn't need to page memory.

Well, "files that doesn't need to page memory" can be something you can think
happen but it can be otherwise. I can easily write scene in a few lines which
needs more memory that all computer I have ever seen had. Size of the scene
file doesn't determine memory usage during rendering.

In the scene you have about 100 spheres all with some transparency. Every
surface on the ray path increases memory usage. Spheres can be overlaped and
bounding hierarchy (something related to raytracing algorithm) can be harder
to walk.

> There is anything to tune? Is it normal?

I would do as follow:
1. Turn off transparency and compare speed between versions
2. Turn off overlaping spheres and compare between versions
3. Determine dependency between number of spheres and slowdown.
None of above could be of much help for you but it could give us some overview
what could be problematic. Be sure you use 3.5 and 3.6 under the same
conditions of your OS and be sure that difference is really large. Note that
talking abou difference of 30% in case of 10 second long render is somehow
misleading because there is so many OS related activity that it influence
rendering time. It could be worth to compary some longer renders.

ABX


Post a reply to this message

From: povray
Subject: Re: Ver. 3.6 speed
Date: 22 Jun 2004 11:08:30
Message: <40d83285.3381650@localhost>
On Mon, 21 Jun 2004 23:34:46 +0200, "Groucho" <gro### [at] tripodcom>
wrote:

>Comparing the render time of the same model with versions 3.5 and 3.6 I
>always get that 3.6 takes around 30% more time to render than 3.5.
>

I have not done a rigorous test (ie - me with a topwatch
rendering one scene with each version) but I think sometimes
I get a slowdown also ... not all the time though.

>I'm using the Windows version on a P4 2.4GHz with 512MB and I always tested
>with files that doesn't need to page memory.
>

I think it's that pesky p4.  I have a p4 and they are slower
than an AMD chip running at the same clockspeed.  My guess is
that p4s have a sucky floating point implementation.

version 3.5 had some sort of routine for optimization on a
p4.  It used to display a line of text every time it
rendered that said something like "SSE2 detected - using
SSE2 optimizations".  I don't see this line of text when
rendering with 3.6.

>There is anything to tune? Is it normal?
>

I wonder this also.

>Thx
>
>Groucho
>
>

-- 
to all the companies who wait until a large user base becomes
dependant on their freeware, then shafting said happy campers with
mandatory payment for continued usage. I spit on your grave.


Post a reply to this message

From: povray
Subject: Re: Ver. 3.6 speed
Date: 23 Jun 2004 13:13:36
Message: <40d985e8.2353137@localhost>
On Tue, 22 Jun 2004 13:30:04 GMT, pov### [at] bestwebnetalmost (povray)
wrote:

>
>I have not done a rigorous test (ie - me with a topwatch
>rendering one scene with each version) but I think sometimes
>I get a slowdown also ... not all the time though.
>

OK.  I've since done some testing.  Good thing I did because
that statement I made (above) is just *not* correct.

I didn't use a stopwatch but instead used the Total time
displayed by povray.  I selected some files and rendered each
file in v3.5 and v3.6 and noted the times.  All rendered at
320x240, aa 0.3 ...

file                        pov 3.5 total   pov 3.6 total   note
----                        ----            ----            ----
dice_attempt_01.pov         5 min 32 secs   2 min 47 secs   rad + pho
function_pigment_speedtest  38 secs         34 secs
krokar_01                   51 secs         49 secs         rad
sweet_01                    1 min 9 secs    1 min 7 secs

(the above needs a normal usenet (ie fixed width) font to view)

In all four renders, 3.6 was faster.  The most dramatic performance
increase was when photons and radiousity were both in use.

So even on my sucky P4 CPU, v3.6 is faster.  Nice.  :)


-- 
to all the companies who wait until a large user base becomes
dependant on their freeware, then shafting said happy campers with
mandatory payment for continued usage. I spit on your grave.


Post a reply to this message

From: Thorsten Froehlich
Subject: Re: Ver. 3.6 speed
Date: 23 Jun 2004 13:33:04
Message: <40d9bed0$1@news.povray.org>
In article <40d83285.3381650@localhost> , povray@bestweb.net.almost (povray)
wrote:

> version 3.5 had some sort of routine for optimization on a
> p4.  It used to display a line of text every time it
> rendered that said something like "SSE2 detected - using
> SSE2 optimizations".  I don't see this line of text when
> rendering with 3.6.

Just because you don't see something like that does not imply it is not
there.

    Thorsten

____________________________________________________
Thorsten Froehlich, Duisburg, Germany
e-mail: tho### [at] trfde

Visit POV-Ray on the web: http://mac.povray.org


Post a reply to this message

From: povray
Subject: Re: Ver. 3.6 speed
Date: 23 Jun 2004 22:16:51
Message: <40da38df.23983165@localhost>
On Wed, 23 Jun 2004 19:33:04 +0200, "Thorsten Froehlich"
<tho### [at] trfde> wrote:

>In article <40d83285.3381650@localhost> , povray@bestweb.net.almost (povray)
>wrote:
>
>> version 3.5 had some sort of routine for optimization on a
>> p4.  It used to display a line of text every time it
>> rendered that said something like "SSE2 detected - using
>> SSE2 optimizations".  I don't see this line of text when
>> rendering with 3.6.
>
>Just because you don't see something like that does not imply it is not
>there.
>

Yes!  too true!  My experiments (after that post)
revealed that!  Just color me embarassed.  :o



-- 
to all the companies who wait until a large user base becomes
dependant on their freeware, then shafting said happy campers with
mandatory payment for continued usage. I spit on your grave.


Post a reply to this message

From: Groucho
Subject: Re: Ver. 3.6 speed
Date: 24 Jun 2004 08:05:05
Message: <40dac371@news.povray.org>
Watching carefully at your tests, I have understood that the reason why I
get worst results with 3.6 must be out of the pov file. So it must be some
parameter.

The only parameter that I'm using is the antialiasing. And here it is:

ver 3.5 +a0.3: 30 secs
ver 3.6 +a0.3: 30 secs
ver 3.5 +a0.01: 64 secs
ver 3.6 +a0.01: 93 secs

So, antialing management in version 3.6 takes more time than in 3.5.
Does anybody know if at least it works better. I cannot see any difference
in the result.

Cheers,
Goucho

"Groucho" <gro### [at] tripodcom> wrote in message
news:40d75460@news.povray.org...
> Comparing the render time of the same model with versions 3.5 and 3.6 I
> always get that 3.6 takes around 30% more time to render than 3.5.
>
> I'm using the Windows version on a P4 2.4GHz with 512MB and I always
tested
> with files that doesn't need to page memory.
>
> There is anything to tune? Is it normal?
>
> Thx
>
> Groucho
>
>


Post a reply to this message

From: Thorsten Froehlich
Subject: Re: Ver. 3.6 speed
Date: 24 Jun 2004 09:37:44
Message: <40dad928$1@news.povray.org>
In article <40dac371@news.povray.org> , "Groucho" <gro### [at] tripodcom>
wrote:

> So, antialing management in version 3.6 takes more time than in 3.5.

No, we fixed bugs related to antialiasing.  It does not take more time
because of any of the changes, it now takes as long as it should always
have.

    Thorsten

____________________________________________________
Thorsten Froehlich, Duisburg, Germany
e-mail: tho### [at] trfde

Visit POV-Ray on the web: http://mac.povray.org


Post a reply to this message

Goto Latest 10 Messages Next 3 Messages >>>

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