POV-Ray : Newsgroups : povray.general : Speed of POV 3.02 on lots of stuff Server Time
13 Aug 2024 17:24:21 EDT (-0400)
  Speed of POV 3.02 on lots of stuff (Message 1 to 6 of 6)  
From: Mark Arrasmith
Subject: Speed of POV 3.02 on lots of stuff
Date: 14 Jul 1998 13:56:50
Message: <35ab8dd2.0@news.povray.org>
In the old thread of Improving the Speed of POV-Ray 3.02 on Win95 (which
started on June 15)  there was a small processor dicussion about speed and
POV-Ray.

Typical stuff of . . .
>Funny. I have a Pentium II 233, 64MB System. This is the part
>where I lecture you about the K6's FPU performance, but I won't. :)

Well being tired of this and having access to a few different computers so I
decided to run the speed test . . .

-i skyvase.pov
+v1
-d
+ft
-x
+a0.300
+r3
-q9
-w640
-h480
-mv2.0
-b1000

And these are my numbers, in order of best performance and all running
POV-Ray 3.02 for the respective operating systems.  You can adjust the
approximate system cost to your experience.

Alpha 21164A 533MHz 128Meg Ram
    Operating System:  RH Linux 4.2
    POV: Alpha Linux Command Line Version
    approximate system cost:  $3000
    time:  1 min 22 sec

Alpha 21164PC 533MHz 64Meg Ram
    Operating System:  WinNT 4.0
    POV:  AlphaNT Command Line Version (personal compile)
    approximate system cost:  $1900
    time:  1 min 23 sec

PII 333MHz 64Meg Ram
    Operating System:  Win95
    POV:  Standard Distribution
    approximate system cost:  $1800
    time:  2 min 3 sec

K6 266MHz 64Meg Ram
    Operating System:  Win95
    POV:  Standard Distribution
    approximate system cost:  $1200
    time:  2 min 58 sec

K6 200MHz 128Meg Ram
    Operating System:  RH Linux 4.2
    POV: x86 Linux Command Line Version
    approximate system cost:  $1000
    time:  3 min 9 sec

Pentium 200MHz MMX 64Meg Ram
    Operating System:  WinNT 4.0
    POV:  Standard Distribution
    approximate system cost:  $1000
    time:  4 min 1 sec

I get a few things from this.  One: the Alpha 21164PC really kicks ass.
Two: I'll never buy Intel for rendering, I'll buy Alpha.  And Three: If you
are really interested in rendering and time, use Linux and the commandline
version of POV-Ray.

Just my info.  Please add your times if you want (though sticking to version
3.02).

Mark Arrasmith
arr### [at] mathtwsuedu


Post a reply to this message

From: Doug Partridge
Subject: Re: Speed of POV 3.02 on lots of stuff
Date: 15 Jul 1998 02:15:08
Message: <35AC3ADA.7443@lawson.appcomp.utas.edu.au>
Why the price difference between the alpha running linux and the alpha
running nt?

There was also only one second difference between linux and nt, does
this make linux really worth it?

I usume the prices are us dollars


Doug Partridge
~
~


Post a reply to this message

From: Mark Arrasmith
Subject: Re: Speed of POV 3.02 on lots of stuff
Date: 15 Jul 1998 10:01:24
Message: <35aca824.0@news.povray.org>
Doug Partridge wrote in message
<35A### [at] lawsonappcomputaseduau>...
>Why the price difference between the alpha running linux and the alpha
>running nt?


The Alphas are actually two different processors.  The 21164A (also known as
the LX version) is the "full" version and the 21164PC (also known as the SX
version) is an alpha minus the onboard L2 cache and at most 1meg cache on
the motherboard.  The 21164PC is a much smaller die size and therefore much
cheaper to produce.  So it costs less.

>There was also only one second difference between linux and nt, does
>this make linux really worth it?

It seems Linux makes a big difference only in x86 systems for POV-Ray.

I'm still trying to figure out why the 21164A didn't beat the 21164PC by a
larger margin.  In most software I have written the 21164A is about 5% to
10% faster.

21164A
    onchip L2 and 2 meg cache on motherboard
    128 Meg ram
    Ultra Wide SCSI hard drie
    64 bit Linux

21164PC
    no L2 and 1 meg cache on motherboard
    64 Meg ram
    E.IDE hard drive
    32bit Windows NT

All that should give the 21164A an edge.  The only plus for the 21164PC is
that it can access the cache on the motherboard faster (I think 2
nanoseconds if I remember right).  I need to look into the intervals of POV
more to really what is going on.  That and I'm tired of the division by
zero's (which is an alpha version problem).

Also on Linux vs. NT I've noticed that Linux has more consistant runs.  I
was making an animation under NT where the parse took 45 seconds for the
first frame.  After that the parse time kept getting faster and faster.  At
about the 100'th frame the parse time was down to 19 seconds!  It stayed
there after that.  Under Linux it was around 20 seconds  through all the
frames.  NT has alot of odd behavior that way.

>I usume the prices are us dollars
Yep.  Sorry, about that.

mark


Post a reply to this message

From: Mark Arrasmith
Subject: Re: Speed of POV 3.02 on lots of stuff
Date: 15 Jul 1998 11:47:48
Message: <35acc114.0@news.povray.org>
Well this paragraph doesn't make too much sense!

>
>All that should give the 21164A an edge.  The only plus for the
>21164PC is that it can access the cache on the motherboard
>faster (I think 2 nanoseconds if I remember right).  I need to look
>into the intervals of POV more to really what is going on.
>That and I'm tired of the division by zero's (which is an alpha
>version problem).
>


Corrected version:

All that should give the 21164A the edge.  The only plus for the 21164PC is
that it can access the motherboard cache at 7.5ns (versus the 21164A which
is at 9ns).  I need to look into the internals of POV to understand what is
going on.  And to try and figure out why POV-Ray on the Alpha gives division
by zero and dies during certain renders.

mark


Post a reply to this message

From: Uwe Post
Subject: Re: Speed of POV 3.02 on lots of stuff
Date: 16 Jul 1998 10:35:08
Message: <35AE01FD.BAD5A921@ct.heise.delete-this.de>
Hi pov-benchmark-freaks,

You can find some other Pov-Ray CPU benchmarks on my homepage (allthough
only x86 and not the skyvase...)

http://www.pingweb.de/liga/ping_up/priv/bench/bench.htm

bye

  Uwe


Post a reply to this message

From: Jim Kress
Subject: Re: Speed of POV 3.02 on lots of stuff
Date: 9 Aug 1998 16:44:16
Message: <35cdfc10.0@news.povray.org>
Some additional numbers for the skyvase benchmark presented by Mark:

Xeon 400 MHz with 512MB RAM (Dell Precisions 610 WS)
Operating System; NT Workstation v4 SP3

POV: 3.1b4
time: 1min 53 sec

POV: 3.02 (standard release)
time: 1min 45 sec

POV: 3.01 PII optimized VC++ v5 compile by Steven Pigeon
time: 1 min 34 sec

The 3.02 time is about 5% better that one would expect (based on cpu speed
scaling of the Alpha results)

Assuming no significant differences betwen 3.01 and 3.02, the PII optimized
results are about 15% better than one would expect.


Mark Arrasmith wrote in message <35ab8dd2.0@news.povray.org>...
>In the old thread of Improving the Speed of POV-Ray 3.02 on Win95 (which
>started on June 15)  there was a small processor dicussion about speed and
>POV-Ray.


---- <snip> ----

> Alpha 21164A 533MHz 128Meg Ram
>    Operating System:  RH Linux 4.2
>    POV: Alpha Linux Command Line Version
>    approximate system cost:  $3000
>    time:  1 min 22 sec
>
>Alpha 21164PC 533MHz 64Meg Ram
>    Operating System:  WinNT 4.0
>    POV:  AlphaNT Command Line Version (personal compile)
>    approximate system cost:  $1900
>    time:  1 min 23 sec
>
>PII 333MHz 64Meg Ram
>    Operating System:  Win95
>    POV:  Standard Distribution
>    approximate system cost:  $1800

---- <snip> ---

>
>Just my info.  Please add your times if you want (though sticking to
version
>3.02).
>
>Mark Arrasmith
>arr### [at] mathtwsuedu
>
>
>
>
>


Post a reply to this message

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