 |
|
|
|
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Just installed POV on my newly reanimated Mac G5 running Debian Wheezy.
Ran the benchmark and compared it against my install on my AMD Phenom II X6
running Mint 16
Mac G5 (Late 2005): Debian Wheezy -->
Render Time:
Photon Time: 0 hours 0 minutes 6 seconds (6.479 seconds)
using 7 thread(s) with 7.400 CPU-seconds total
Radiosity Time: No radiosity
Trace Time: 0 hours 1 minutes 49 seconds (109.803 seconds)
using 4 thread(s) with 408.838 CPU-seconds total
povray: removing /tmp/pov24343.ini
povray: removing /tmp/pov24343.pov
POV-Ray finished
AMD Phenom II X6: Mint 16 "Petra" -->
Render Time:
Photon Time: 0 hours 0 minutes 2 seconds (2.248 seconds)
using 9 thread(s) with 2.591 CPU-seconds total
Radiosity Time: No radiosity
Trace Time: 0 hours 3 minutes 29 seconds (209.132 seconds)
using 6 thread(s) with 1232.369 CPU-seconds total
povray: removing /tmp/pov4747.ini
povray: removing /tmp/pov4747.pov
POV-Ray finished
Note: Parse time on the newer machine was faster by almost a whole second, but
Trace time was a bit more than twice as long.
Any ideas why this might be?
Regards,
A.D.B.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Am 04.07.2014 03:34, schrieb Anthony D. Baye:
> Note: Parse time on the newer machine was faster by almost a whole second, but
> Trace time was a bit more than twice as long.
>
> Any ideas why this might be?
Maybe because the G5 uses an entirely different processor architecture? ;-)
The IBM PowerPC architecture originates in the high performance
computing sector, where high precision floating point number crunching
is daily business, and the typical occupation of a CPU is to do just that.
The Intel x64 architecture originates in the personal computing sector,
where daily business means office work, internet browsing and gaming,
and the typical occupation of a CPU is to wait for other system
components (most notably memory).
As a matter of fact, POV-Ray is a very unconventional piece of software
to run on a personal computer, in that it keeps the CPU core itself
extremely busy, as indicated not by the CPU meter (which cannot
distinguish between cores busy computing and cores busy waiting for
memory access to complete), but by CPU temperature. If you want to push
your computer's cooling system to its limits, POV-Ray is the ideal tool.
On poorly designed systems, the POV-Ray performance bottleneck may
actually be not the CPU itself, but the cooling system, which may cause
smart CPUs to enter thermal throttling. (And in such a case you better
hope the CPU /is/ smart; otherwise, permanent damage is a real possibility.)
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
clipka <ano### [at] anonymous org> wrote:
> On poorly designed systems, the POV-Ray performance bottleneck may
> actually be not the CPU itself, but the cooling system, which may cause
> smart CPUs to enter thermal throttling. (And in such a case you better
> hope the CPU /is/ smart; otherwise, permanent damage is a real possibility.)
I can endorse this.
My last laptop developed a broken fan early on and kept shutting itself down
limit and stop PovRay from overheating the poor wee machine.
Stephen
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
"Anthony D. Baye" <Sha### [at] spamnomorehotmailcom> wrote:
> Just installed POV on my newly reanimated Mac G5 running Debian Wheezy.
>
> Ran the benchmark and compared it against my install on my AMD Phenom II X6
> running Mint 16
>
> Mac G5 (Late 2005): Debian Wheezy -->
> Render Time:
> Photon Time: 0 hours 0 minutes 6 seconds (6.479 seconds)
> using 7 thread(s) with 7.400 CPU-seconds total
> Radiosity Time: No radiosity
> Trace Time: 0 hours 1 minutes 49 seconds (109.803 seconds)
> using 4 thread(s) with 408.838 CPU-seconds total
> povray: removing /tmp/pov24343.ini
> povray: removing /tmp/pov24343.pov
> POV-Ray finished
>
> AMD Phenom II X6: Mint 16 "Petra" -->
> Render Time:
> Photon Time: 0 hours 0 minutes 2 seconds (2.248 seconds)
> using 9 thread(s) with 2.591 CPU-seconds total
> Radiosity Time: No radiosity
> Trace Time: 0 hours 3 minutes 29 seconds (209.132 seconds)
> using 6 thread(s) with 1232.369 CPU-seconds total
> povray: removing /tmp/pov4747.ini
> povray: removing /tmp/pov4747.pov
> POV-Ray finished
>
> Note: Parse time on the newer machine was faster by almost a whole second, but
> Trace time was a bit more than twice as long.
>
> Any ideas why this might be?
>
> Regards,
> A.D.B.
What are the exact CPU specs for each computer (ie which model? how fast is it?
are you overclocking?). It looks like from the Phenom II render times that it's
no more than a 1090T? The PowerPC time looks very suspect too. Perhaps you could
render the benchmark and save the image output.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
clipka <ano### [at] anonymousorg> wrote:
> Am 04.07.2014 03:34, schrieb Anthony D. Baye:
>
> > Note: Parse time on the newer machine was faster by almost a whole second, but
> > Trace time was a bit more than twice as long.
> >
> > Any ideas why this might be?
>
> Maybe because the G5 uses an entirely different processor architecture? ;-)
>
> The IBM PowerPC architecture originates in the high performance
> computing sector, where high precision floating point number crunching
> is daily business, and the typical occupation of a CPU is to do just that.
>
> The Intel x64 architecture originates in the personal computing sector,
> where daily business means office work, internet browsing and gaming,
> and the typical occupation of a CPU is to wait for other system
> components (most notably memory).
Here are benchmarks from 2005 running Povray 3.6, PowerPC 970 @ 2.5 GHz vs.
Opteron 240 (2.4 GHz). The Opteron from back then easily beats the 970.
http://www.anandtech.com/show/1702/5
There's something not quite right with OP's setup.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Am 04.07.2014 17:22, schrieb jhu:
> clipka <ano### [at] anonymousorg> wrote:
>> Am 04.07.2014 03:34, schrieb Anthony D. Baye:
>>
>>> Note: Parse time on the newer machine was faster by almost a whole second, but
>>> Trace time was a bit more than twice as long.
>>>
>>> Any ideas why this might be?
>>
>> Maybe because the G5 uses an entirely different processor architecture? ;-)
>>
>> The IBM PowerPC architecture originates in the high performance
>> computing sector, where high precision floating point number crunching
>> is daily business, and the typical occupation of a CPU is to do just that.
>>
>> The Intel x64 architecture originates in the personal computing sector,
>> where daily business means office work, internet browsing and gaming,
>> and the typical occupation of a CPU is to wait for other system
>> components (most notably memory).
>
> Here are benchmarks from 2005 running Povray 3.6, PowerPC 970 @ 2.5 GHz vs.
> Opteron 240 (2.4 GHz). The Opteron from back then easily beats the 970.
> http://www.anandtech.com/show/1702/5
That was POV-Ray 3.6, i.e. single-thread performance (and also a
different benchmark scene).
> There's something not quite right with OP's setup.
Maybe the thermal throttling I mentioned?
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
clipka <ano### [at] anonymousorg> wrote:
> That was POV-Ray 3.6, i.e. single-thread performance (and also a
> different benchmark scene).
I've found good correlation between Povray 3.6 benchmark and per core
performance on Povray 3.7 benchmark. PowerPC 970 should be slower, clock for
clock, core for core, compared with Phenom II (which is already faster than K8).
> > There's something not quite right with OP's setup.
>
> Maybe the thermal throttling I mentioned?
No, not thermal throttling. My Phenom II 1090T @ 3.2 GHz (which is stock)
finishes the benchmark in 3 minutes 9 seconds (Ubntu 14.04, gcc 4.8 <- this is
the fastest OS + compiler setup I've been able to find for non-Intel
processors). It's running at 3.2 GHz for all cores the whole time based on the
output of /proc/cpuinfo
I think the Mac is misconfigured somehow. There's absolutely no way a quad-core
PPC 970 is faster than a hex core Phenom II that's also clocked higher.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
I've given this more thought. The most likely explanation is that the Mac is
rendering the benchmark scene at a lower resolution than it should be. The
default resolution for Povray 3.7 benchmark is 512x512. For Povray 3.6 it was
384x384. So go check benchmark.ini and see what the height and width are set to.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
"jhu" <nomail@nomail> wrote:
> clipka <ano### [at] anonymousorg> wrote:
> > That was POV-Ray 3.6, i.e. single-thread performance (and also a
> > different benchmark scene).
>
> I've found good correlation between Povray 3.6 benchmark and per core
> performance on Povray 3.7 benchmark. PowerPC 970 should be slower, clock for
> clock, core for core, compared with Phenom II (which is already faster than K8).
>
> > > There's something not quite right with OP's setup.
> >
> > Maybe the thermal throttling I mentioned?
>
> No, not thermal throttling. My Phenom II 1090T @ 3.2 GHz (which is stock)
> finishes the benchmark in 3 minutes 9 seconds (Ubntu 14.04, gcc 4.8 <- this is
> the fastest OS + compiler setup I've been able to find for non-Intel
> processors). It's running at 3.2 GHz for all cores the whole time based on the
> output of /proc/cpuinfo
>
> I think the Mac is misconfigured somehow. There's absolutely no way a quad-core
> PPC 970 is faster than a hex core Phenom II that's also clocked higher.
This is my thought exactly. The G5 is a 2.5Ghz Quad core, the Phenom II is the
1090T Black Edition it's supposed to be clocked to 3.01 Ghz but the specs are
listed as 3.2 - 3.6.
I haven't fiddled with anything on the mac side of things other than installing
a non-standard OS which uses open firmware. If there's anything configured
strangely, I didn't do it.
I was discussing this with a friend of mine last night, and his theory was that
the G5 is using SIMD, which the Phenom II doesn't seem to have. I suppose that
-could- make a difference with number crunching...
Regards,
A.D.B
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
"jhu" <nomail@nomail> wrote:
> I've given this more thought. The most likely explanation is that the Mac is
> rendering the benchmark scene at a lower resolution than it should be. The
> default resolution for Povray 3.7 benchmark is 512x512. For Povray 3.6 it was
> 384x384. So go check benchmark.ini and see what the height and width are set to.
I'm not using 3.6 on either machine; the install is a fresh build from the git
repository.
on the off chance, I did check benchmark.ini for both machines, and the render
size is 512x512.
Regards,
A.D.B.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
