I've been reading about performance benchmarks for PovRay on the G5 and it
seems that there is a wide disparity in results.  Several have posted
benchmark results (using the built-in Benchmark scene) of not much over 1
minute with a dual 2Ghz machine, while others have posted upwards of 20
minutes.  Running PovRay 3.6.0 on my dual 2 Ghz G5 machine, I'm getting
results of about 38 minutes.  I don't see any issues related to CPU usage.
It appears to use both processors equally at about 50%.  Any ideas about
why there is such a difference?

Thanks for any light anyone can shed on the subject.

Dave.

Post a reply to this message

In article <web.4102bb1b89c889b61b6d5cc50@news.povray.org> , "David Brewer"
<dav### [at] brewer-familyorg> wrote:

> I've been reading about performance benchmarks for PovRay on the G5 and it
> seems that there is a wide disparity in results.  Several have posted
> benchmark results (using the built-in Benchmark scene) of not much over 1
> minute with a dual 2Ghz machine,

Wherever you saw those results, they were either for another benchmark or
they were just plain wrong because either someone made a mistake running the
benchmark or intentionally manipulated the results for fun.  Happens all the
time.

> while others have posted upwards of 20
> minutes.  Running PovRay 3.6.0 on my dual 2 Ghz G5 machine, I'm getting
> results of about 38 minutes.

While I would expect something around 30 minutes, 38 minutes does not sound
all that far off.

> I don't see any issues related to CPU usage.
> It appears to use both processors equally at about 50%.

You cannot use CPU usage window to see how much processor time a specific
application consumes.  All you see there is the time-average over a second
or more that each processor was used. Applications will be executed on
either several times a second.  Use the Activity Monitor so see how much an
application uses on each processor.  POV-Ray only uses one thread for
rendering, and hence will only use one CPU.

> Any ideas about why there is such a difference?

There is none.

    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

Thank you, Thorsten.

Dave.

"Thorsten Froehlich" <tho### [at] trfde> wrote:
> In article <web.4102bb1b89c889b61b6d5cc50@news.povray.org> , "David Brewer"
> <dav### [at] brewer-familyorg> wrote:
>
> > I've been reading about performance benchmarks for PovRay on the G5 and it
> > seems that there is a wide disparity in results.  Several have posted
> > benchmark results (using the built-in Benchmark scene) of not much over 1
> > minute with a dual 2Ghz machine,
>
> Wherever you saw those results, they were either for another benchmark or
> they were just plain wrong because either someone made a mistake running the
> benchmark or intentionally manipulated the results for fun.  Happens all the
> time.
>
> > while others have posted upwards of 20
> > minutes.  Running PovRay 3.6.0 on my dual 2 Ghz G5 machine, I'm getting
> > results of about 38 minutes.
>
> While I would expect something around 30 minutes, 38 minutes does not sound
> all that far off.
>
> > I don't see any issues related to CPU usage.
> > It appears to use both processors equally at about 50%.
>
> You cannot use CPU usage window to see how much processor time a specific
> application consumes.  All you see there is the time-average over a second
> or more that each processor was used. Applications will be executed on
> either several times a second.  Use the Activity Monitor so see how much an
> application uses on each processor.  POV-Ray only uses one thread for
> rendering, and hence will only use one CPU.
>
> > Any ideas about why there is such a difference?
>
> There is none.
>
>     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

"Thorsten Froehlich" <tho### [at] trfde> wrote:
> You cannot use CPU usage window to see how much processor time a specific
> application consumes.  All you see there is the time-average over a second
> or more that each processor was used. Applications will be executed on
> either several times a second.  Use the Activity Monitor so see how much an
> application uses on each processor.  POV-Ray only uses one thread for
> rendering, and hence will only use one CPU.


I finally received the 2.5 GHz dual G5 I ordered June 28. So far it looks
pretty good. I've added RAM to total 3 gig.

A quick run of the benchmark (POV-Ray 3.6.0) shows it runs about 7.5
times faster than my previous machine, a 550 MHz PowerBook. Here
are my times, just using the pull-down Render benchmark.

Optimized for G4  25:51
Optimized for G5  25:07

Nothing out of this world, my uncle's (much less expensive :-) AMD 2800+
comes within a couple of minutes of this time, but I really enjoy being able
to do other things while rendering, and I like OS X. I also notice some
amazing numbers for CineBench; if I ever get Cinema 4D it will be very fast.

Running the Activity Monitor shows some strange results as the processors
appear to share the work, one running at 100% for about a second, then
swapping to the other processor 100%. I have no idea how accurate it is;
I set it to update every 0.5 seconds but it appears to display only every
second. My experience at work with SGI multiprocessor code shows great
improvements in speed by locking a process to a processor, but that's a
completely different system.

I wonder if this processor swapping is hurting performance on dual G5's?
(Assuming the Activity Monitor display is accurate.)

-Mark Slone

Post a reply to this message