Hi, I'm by no means a hardware expert, so please be
gentle...

I'm sure many of us have browsed 
http://www.haveland.com/index.htm?povbench/index.php
and seen the (to me) astonishing 00:07:34 for a single
Pentium Xeon on a Dell Precision 450 running Lunar Linux.
By comparison my old Pentium III coppermine running
SuSE 8.2 and a gcc-optimized PovRay 3.50c took 02:12:59.

Anyway I thought "I gotta get one of those" and forked
out for a brand new box (bespoke built) with 2 P4 Xeons
and 4Gb of core memory, and put Lunar on it.

Generally I'm extremely pleased with it, but dissappointingly
my benchmark (pov 3.6.1 this time) comes nowhere close.
With a gcc optimized build I measured 00:35:30 and after
getting hold of the Intel Compiler and rebuilding with that
I got it down to 00:30:28.

here's some of my /proc/cpuinfo

billh@Tarragon ~ $ cat /proc/cpuinfo
processor       : 0
vendor_id       : GenuineIntel
cpu family      : 15
model           : 2
model name      : Intel(R) Xeon(TM) CPU 3.06GHz
stepping        : 9
cpu MHz         : 3056.924
cache size      : 512 KB
fdiv_bug        : no
hlt_bug         : no
f00f_bug        : no
coma_bug        : no
fpu             : yes
fpu_exception   : yes
cpuid level     : 2
wp              : yes
flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca 
cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe cid
bogomips        : 6029.31

... x4

I'm happy enough with my vast improvement, but when the
time comes to upgrade I'd like to know what I really should
have bought -- though I now suspect I couldn't have afforded
it :-)

Any comments?


-- 
Bill Hails
http://thyme.homelinux.net/

Post a reply to this message

Bill Hails <bil### [at] europeyahoo-inccom> wrote:
> I'm sure many of us have browsed 
> http://www.haveland.com/index.htm?povbench/index.php
> and seen the (to me) astonishing 00:07:34 for a single
> Pentium Xeon on a Dell Precision 450 running Lunar Linux.

  The result is either bogus (do the povbench maintainers really
corroborate the submissions?) or the result of rendering the
benchmark with non-standard parameters (eg. without antialiasing,
with a lower resolution or whatever) thus being invalid.
  There's simply no way that processor can render the benchmark that
fast with genuine benchmark settings.
  I don't believe there exists currently *any* processor which can
singlehandedly render the benchmark that fast. All the other <15min
results in that page are probably bogus as well. I'll believe it
when I see it myself live (with me running the benchmark and checking
the results).

-- 
plane{-x+y,-1pigment{bozo color_map{[0rgb x][1rgb x+y]}turbulence 1}}
sphere{0,2pigment{rgbt 1}interior{media{emission 1density{spherical
density_map{[0rgb 0][.5rgb<1,.5>][1rgb 1]}turbulence.9}}}scale
<1,1,3>hollow}text{ttf"timrom""Warp".1,0translate<-1,-.1,2>}//  - Warp -

Post a reply to this message

Bill Hails wrote:
> Hi, I'm by no means a hardware expert, so please be
> gentle...
> 
> I'm sure many of us have browsed 
> http://www.haveland.com/index.htm?povbench/index.php
> and seen the (to me) astonishing 00:07:34 for a single
> Pentium Xeon on a Dell Precision 450 running Lunar Linux.
> By comparison my old Pentium III coppermine running
> SuSE 8.2 and a gcc-optimized PovRay 3.50c took 02:12:59.
> 

Well - the problem about these benchmark results is that there is no way 
to proove their validity - people could enter any time and no one would 
notice.

You can find some accurate results in the tests Nicolas made recently:

http://pov4grasp.free.fr/articles/fastpov1/

Christoph

-- 
POV-Ray tutorials, include files, Sim-POV,
HCR-Edit and more: http://www.tu-bs.de/~y0013390/
Last updated 23 Sep. 2004 _____./\/^>_*_<^\/\.______

Post a reply to this message

> http://www.haveland.com/index.htm?povbench/index.php
> and seen the (to me) astonishing 00:07:34 for a single
> Pentium Xeon on a Dell Precision 450 running Lunar Linux.

	I'm sorry to say I don't believe in any of these benchmark results
(at least those below the 20 minutes barreer on a single processor).  Even
in those that give some sort of notes on how they prepared the binary and
ran the benchmark.

> Generally I'm extremely pleased with it, but dissappointingly
> my benchmark (pov 3.6.1 this time) comes nowhere close.
> With a gcc optimized build I measured 00:35:30 and after
> getting hold of the Intel Compiler and rebuilding with that
> I got it down to 00:30:28.

> model name      : Intel(R) Xeon(TM) CPU 3.06GHz

	Your results agree pretty well with what I got myself for
the Intel Xeon @ 3.06 GHz:

http://pov4grasp.free.fr/articles/fastpov1/xeon_3.06.png

	GCC 3.4.x gives at best 00:32:30, and ICC 8.1 about 00:28:10 in
my tests -- so you may have a look at which compiler flags I used if you
want to gain another two minutes  :-)

	- NC

Post a reply to this message

Nicolas Calimet wrote:

> Your results agree pretty well with what I got myself for
> the Intel Xeon @ 3.06 GHz:
> 
> http://pov4grasp.free.fr/articles/fastpov1/xeon_3.06.png
> 
> GCC 3.4.x gives at best 00:32:30, and ICC 8.1 about 00:28:10 in
> my tests -- so you may have a look at which compiler flags I used if you
> want to gain another two minutes  :-)
> 
> - NC

I can't seem to close that 2 minutes gap :-), the same options
you used make no appreciable difference, my best time is 00:30:03.
I guess it's down to kernel build and/or libraries.

Thanks for your comments and especially for that report, very
informative.

-- 
Bill Hails
http://thyme.homelinux.net/

Post a reply to this message

> Nicolas Calimet wrote:
> 
> 
>>Your results agree pretty well with what I got myself for
>>the Intel Xeon @ 3.06 GHz:
>>
>>http://pov4grasp.free.fr/articles/fastpov1/xeon_3.06.png
>>
>>GCC 3.4.x gives at best 00:32:30, and ICC 8.1 about 00:28:10 in
>>my tests -- so you may have a look at which compiler flags I used if you
>>want to gain another two minutes  :-)
>>
>>- NC
> 
> 
> I can't seem to close that 2 minutes gap :-), the same options
> you used make no appreciable difference, my best time is 00:30:03.
> I guess it's down to kernel build and/or libraries.
> 
> Thanks for your comments and especially for that report, very
> informative.
> 

As you are speaking about librairies, something noticable is that 
staticaly linked libraries offers better performances ... perhaps this 
can explain your 2 minutes ?

Post a reply to this message

news.povray.org wrote:


>> Nicolas Calimet wrote:
>> 
>> 
>>>Your results agree pretty well with what I got myself for
>>>the Intel Xeon @ 3.06 GHz:
>>>
>>>http://pov4grasp.free.fr/articles/fastpov1/xeon_3.06.png
>>>
>>>GCC 3.4.x gives at best 00:32:30, and ICC 8.1 about 00:28:10 in
>>>my tests -- so you may have a look at which compiler flags I used if you
>>>want to gain another two minutes  :-)
>>>
>>>- NC
>> 
>> 
>> I can't seem to close that 2 minutes gap :-), the same options
>> you used make no appreciable difference, my best time is 00:30:03.
>> I guess it's down to kernel build and/or libraries.
>> 
>> Thanks for your comments and especially for that report, very
>> informative.
>> 
> 
> As you are speaking about librairies, something noticable is that
> staticaly linked libraries offers better performances ... perhaps this
> can explain your 2 minutes ?

Hmm, maybe, but I thought dynamic linking only had a performance hit
while an executable was starting up, I wasn't aware of any penalty
once an executable is up and running.

I was thinking more about optimizations that might have been done
better while building the kernel and libc, libpng etc, though since
my whole system was gcc-compiled with switches optimised for my
hardware (Lunar builds itself from source), I would have expected
a slightly better result than the published benchmark, if anything.

-- 
Bill Hails
http://thyme.homelinux.net/

Post a reply to this message