Speed..Speed..Speed

Speed..Speed..Speed

Post by Jim Mose » Wed, 10 Mar 1999 04:00:00



Am currently running a K6-2 300Mhz processor with 128Mb of 100 Mhz
memory and considering
upgrading to a faster board and processor. I am pursuing a project which
will require scads of floating point
calculations on large arrays..the results of which are written back out
to disk..about 500 Mbytes worth.
So I'm looking at the new high end processors..PII/PIII and discover
this PII Xeon chip.

Does anyone have any experience with this chip? It very expensive..about
$900. A dual Xeon board from
SuperMicro with both processors runs over $2000. So how much faster is
this chip?  Intel's web site
says it has a 512K L2 cache (even more expensive versions have 1Mb and
2Mb cache) which runs
at PROCESSOR SPEED .. 400 Mhz in this case. The comparable PII-400 chip
has a smaller(?) L2 cache
which runs at 1/2 chip speed. The comparable AMD K6-2 400 chip has a 1Mb
L2 cache but it is off chip
and accessed by the 100 Mhz bus.

So there you have it. Is the Xeon really worth double the PII price or
triple the K6-2 price?
As far as I can see.. the main difference is the L2 cache speed. Has
anybody seen any benchmarks
on this chip vis-a-via "comparable chips?

Jim M.

 
 
 

Speed..Speed..Speed

Post by Douglas Graha » Wed, 10 Mar 1999 04:00:00


For fpu the PIII outperform for the k6-2 and k6-3 however WIAT FOR THE K7 it
is gonna wipe the floor with anything that intel can produce.  I have it on
good faith that the 500K7 will out perform the Xeon(P3)600

 
 
 

Speed..Speed..Speed

Post by Daniele Bernardin » Wed, 10 Mar 1999 04:00:00



> Am currently running a K6-2 300Mhz processor with 128Mb of 100 Mhz
> memory and considering
> upgrading to a faster board and processor. I am pursuing a project which
> will require scads of floating point
> calculations on large arrays..the results of which are written back out
> to disk..about 500 Mbytes worth.
> So I'm looking at the new high end processors..PII/PIII and discover
> this PII Xeon chip.

> Does anyone have any experience with this chip? It very expensive..about
> $900. A dual Xeon board from
> SuperMicro with both processors runs over $2000. So how much faster is
> this chip?  Intel's web site
> says it has a 512K L2 cache (even more expensive versions have 1Mb and
> 2Mb cache) which runs
> at PROCESSOR SPEED .. 400 Mhz in this case. The comparable PII-400 chip
> has a smaller(?) L2 cache
> which runs at 1/2 chip speed. The comparable AMD K6-2 400 chip has a 1Mb
> L2 cache but it is off chip
> and accessed by the 100 Mhz bus.

> So there you have it. Is the Xeon really worth double the PII price or
> triple the K6-2 price?
> As far as I can see.. the main difference is the L2 cache speed. Has
> anybody seen any benchmarks
> on this chip vis-a-via "comparable chips?

> Jim M.

check out

http://www.tomshardware.com

there you will find any kind of benchmark.
The caches of the various cpus are as follows:

cpu             L1cache         L2cache         L3cache

K6-2            64K             512K-2M 100MHz  none
K6-3            64K             256K cpu speed  512K-2M 100MHz
Celeron         32K             128K cpu speed
PII             32K             512K 1/2 cpu speed
PIII            32K             512K 1/2 cpu speed
PII-XEON        32K             512K-1M cpu speed

my 2 cents:
buy an alpha if you really need speed.

Regards,
Daniele

--
********************************************************************
   Daniele Bernardini    
   Sektion Theoretische Physik, LMU Muenchen
   Theresienstr. 37,  80805 Muenchen DEUTSCHLAND        
   Tel: +49 (89) 23944378

   http://www.ls-wess.physik.uni-muenchen.de/~dani/
********************************************************************

 
 
 

Speed..Speed..Speed

Post by Kevin G. Rhoad » Wed, 10 Mar 1999 04:00:00


Your question is indefinite, it is not obvious whether
your program is compute-bound or i/o-bound.

If these 500Meg files are just written line by line as results
are computed, then their size is probably irrelevant and
your program is likely compute-bound.  Faster processor,
faster cache and more main RAM are what you need then.

If those 500Meg files are scratch pads that you access over
and over again, then disk i/o speed may be limiting.  If that
is the case, then adding LOTS more RAM and devoting a
big chunk to disk caching can help more than anything else.

Even if you are compute-bound, whether raw processor speed,
cache size or main RAM size is most helpful is not obvious.

Questions:
Have you profiled the code to find out where it is spending time?
Have you tried changing the OS disk cache settings to see what
effect (if any) that has?
Do you know whether your EXE and all allocated memory fits
in available RAM or does it go out into swapfile?

Tell us more and we can make better suggestions.
--
Kevin G. Rhoads, Ph.D. (Linearity is a convenient fiction.)


 
 
 

Speed..Speed..Speed

Post by Jeffrey J. Potof » Wed, 10 Mar 1999 04:00:00



> Am currently running a K6-2 300Mhz processor with 128Mb of 100 Mhz
> memory and considering
> upgrading to a faster board and processor. I am pursuing a project which
> will require scads of floating point
> calculations on large arrays..the results of which are written back out
> to disk..about 500 Mbytes worth.
> So I'm looking at the new high end processors..PII/PIII and discover
> this PII Xeon chip.

> Does anyone have any experience with this chip? It very expensive..about
> $900. A dual Xeon board from
> SuperMicro with both processors runs over $2000. So how much faster is
> this chip?  Intel's web site
> says it has a 512K L2 cache (even more expensive versions have 1Mb and
> 2Mb cache) which runs
> at PROCESSOR SPEED .. 400 Mhz in this case. The comparable PII-400 chip
> has a smaller(?) L2 cache
> which runs at 1/2 chip speed. The comparable AMD K6-2 400 chip has a 1Mb
> L2 cache but it is off chip
> and accessed by the 100 Mhz bus.

> So there you have it. Is the Xeon really worth double the PII price or
> triple the K6-2 price?
> As far as I can see.. the main difference is the L2 cache speed. Has
> anybody seen any benchmarks
> on this chip vis-a-via "comparable chips?

We did some investigation on quad Xeon boxes before going with a
dual 21264 Alpha machine.  From what I saw, the Xeon wasn't really
much faster than the 450 MhZ PII and cost a whole lot more.  If
I had to go with Intel chips, I'd stick with a dual 450 MhZ PII.

If speed is the primary concern, however, you'll get the most out of
an Alpha.  The 21164, 533MhZ NT boxes are going for about $3400
(or substantially less, depending on configuration) w/video,
256mb ram, etc.

Jeff

 
 
 

Speed..Speed..Speed

Post by John Burto » Wed, 10 Mar 1999 04:00:00



> We did some investigation on quad Xeon boxes before going with a
> dual 21264 Alpha machine.  From what I saw, the Xeon wasn't really
> much faster than the 450 MhZ PII and cost a whole lot more.  If
> I had to go with Intel chips, I'd stick with a dual 450 MhZ PII.

> If speed is the primary concern, however, you'll get the most out of
> an Alpha.  The 21164, 533MhZ NT boxes are going for about $3400
> (or substantially less, depending on configuration) w/video,
> 256mb ram, etc.

> Jeff

Jeff,
  I'm sorta in the same boat as the original poster, but perhaps in a
slightly lower price range. I found the price of the Alpha 21164, 533Mhz
box, about the same as a dual 500Mhz PentiumIII (I know, no real
improvement over Pentium II, but the pricing was inline with PentiumII
*if* it ran at that speed). The Alpha had a SpecFP95 of 21.9 & SpecINT95
of 16.6, where a *single* 500mhz  P-III had a SpecFP95 of 14.7 and
SpecINT95 of 20.6. Supposedly, from the Spec benchmarks, a dual P-III
system *should* be better at floating point than the Alpha 21164 533
Mhz. (I'll believe it when I actually see it...;-)

My question for you is, what kind of prices are we talking about for a
dual 21264 alpha machine?

John

--
John Burton, Ph.D.
Senior Associate                 GATS, Inc.  


(757) 873-5920 (voice)           (757) 873-5920 (fax)

 
 
 

Speed..Speed..Speed

Post by Kent Perrie » Wed, 10 Mar 1999 04:00:00


<SNIP>

Quote:> of 16.6, where a *single* 500mhz  P-III had a SpecFP95 of 14.7 and
> SpecINT95 of 20.6. Supposedly, from the Spec benchmarks, a dual P-III
> system *should* be better at floating point than the Alpha 21164 533
> Mhz. (I'll believe it when I actually see it...;-)

If the SMP box linerally scales, which it doesn't.  If your application isn't
multi-threaded, the second processor will be of little use to you....

Kent

 
 
 

Speed..Speed..Speed

Post by wizar » Wed, 10 Mar 1999 04:00:00



> Am currently running a K6-2 300Mhz processor with 128Mb of 100 Mhz
> memory and considering
> upgrading to a faster board and processor. I am pursuing a project which
> will require scads of floating point
> calculations on large arrays..the results of which are written back out
> to disk..about 500 Mbytes worth.
> So I'm looking at the new high end processors..PII/PIII and discover
> this PII Xeon chip.

> Does anyone have any experience with this chip? It very expensive..about
> $900. A dual Xeon board from
> SuperMicro with both processors runs over $2000. So how much faster is
> this chip?  Intel's web site
> says it has a 512K L2 cache (even more expensive versions have 1Mb and
> 2Mb cache) which runs
> at PROCESSOR SPEED .. 400 Mhz in this case. The comparable PII-400 chip
> has a smaller(?) L2 cache
> which runs at 1/2 chip speed. The comparable AMD K6-2 400 chip has a 1Mb
> L2 cache but it is off chip
> and accessed by the 100 Mhz bus.

> So there you have it. Is the Xeon really worth double the PII price or
> triple the K6-2 price?
> As far as I can see.. the main difference is the L2 cache speed. Has
> anybody seen any benchmarks
> on this chip vis-a-via "comparable chips?

> Jim M.

Jim;

If you intend to develope and run this software under LINUX then why are
you even looking at a Intel chip.    Simply put NONE of the Intel chips are
floating point standouts.     If your seriously considering chips in the
XEON line, then a Alpha is not out of the price range, nor would a PowerPC
based machine.     You might even be able to find a SPARC in this price
range.     Start looking for Alpha powered units from Microway, ASPEN or
DCG to name just a few.

The nice thing about a totally open system like Linux is the freedom to
choose the hardware of you choice.

Dave

 
 
 

Speed..Speed..Speed

Post by José Rui Faustino de Sous » Thu, 11 Mar 1999 04:00:00



>which
>will require scads of floating point
>calculations

Rule of thumb need floating point use intel... BUT if you have money enough (and
if you are considering dual Xeon  2MB you have) take a look at alpha the new
processor 21264 is really the best money can buy... And if you run linux you can
save a bunvle in software...

Quote:>on large arrays..

Another rule of thumb: the larger the array the more a large cache will pay...
If you can fit the whole array into cache it will compute much faster. And even
worst you can have perverse problems like zero cache being faster tham just a
little... The reason for this is that the array does not fit into cache and
cache misses are ocurring all the time and the program will run slower because
of cache organization overhead...

Quote:>the results of which are written back out
>to disk..about 500 Mbytes worth.

Anyway if you need lots of IO get a SCSI subsystem maybe RAID level zero for
even faster performance (see DPT they have very nice caching controlers)

Quote:>So I'm looking at the new high end processors..PII/PIII and discover
>this PII Xeon chip.

PIII does not pay... Yet maybe with the new chipsets at 133Mhz FSB and faster
clock speeds... The new intructions will take _too_ long to be implemented by
compiler makers...

Quote:>So there you have it. Is the Xeon really worth double the PII price or
>triple the K6-2 price?

Double PII maybe not... But triple K6 (for the application you have in mind)
sure is!

Quote:>As far as I can see.. the main difference is the L2 cache speed. Has
>anybody seen any benchmarks

And floating point, wich is very important to you...

See http://www.tomshardware.com/ they have a lot of benchmarks there...

Best regards
Jos Rui
========================================================
Jos Rui Faustino de Sousa
========================================================

http://homepage.esoterica.pt/~jrfsousa/
telephone://+351-(0)39-444940
SnailMail://rua Carlos Alberto Pinto de Abreu, no30C, 1o
3040 Coimbra Portugal
========================================================

 
 
 

Speed..Speed..Speed

Post by John Burto » Thu, 11 Mar 1999 04:00:00




> >which
> >will require scads of floating point
> >calculations

> Rule of thumb need floating point use intel... BUT if you have money enough (and
> if you are considering dual Xeon  2MB you have) take a look at alpha the new
> processor 21264 is really the best money can buy... And if you run linux you can
> save a bunvle in software...

"Rule of thumb need floating point use Intel" ???? I've always heard
just the opposite...need floating point, DON'T use Intel...

Anyone have any Benchmarks on the 21264 ? Price / Mflop ?

Quote:

> Anyway if you need lots of IO get a SCSI subsystem maybe RAID level zero for
> even faster performance (see DPT they have very nice caching controlers)

Aggree with the SCSI, Have had trouble with DPT 2044UW controller &
Linux... Seems at maximum transfer rate (20Mhz) I get rnadom SCSI Bus
errors and occasional lock ups...the two drivers (EATA & EATA-DMA) under
kernel 2.0.36 have *different* errors, but *neither* work without errors
at full transfer rates...and yes, I have checked cables, devices,
terminations, etc. I've swapped the card with an Adaptec 2940UW and
didn't have any problems. I swapped the individual drives with new
drives, and got the errors on the new drives. I've tried several cables,
including those sold by Adaptec that were designed to work for maximum
transfer rates on an UW bus...only consitent thing was DPT controller +
20Mhz SCSI Xfer rate = SCSI Bus errors (timeouts, device busy, etc...)

Quote:> >So I'm looking at the new high end processors..PII/PIII and discover
> >this PII Xeon chip.

> PIII does not pay... Yet maybe with the new chipsets at 133Mhz FSB and faster
> clock speeds... The new intructions will take _too_ long to be implemented by
> compiler makers...

True, the P-III does not have any new features that will cause any
preformance increase, but the manufacturers I looked at are selling
systems with them that are pricewise the same as a PII at the same
speed... (i.e. a P-II 400 system is about $300 more than a P-II 350, and
a P-III 450 is $300 more than a P-II 400, and a P-III 500 is $300 more
than a P-III 450...) so it seems they are being priced as high speed
P-II systems...

Quote:

> >As far as I can see.. the main difference is the L2 cache speed. Has
> >anybody seen any benchmarks

> And floating point, wich is very important to you...

Ummm...where did you get your info? I haven't been able to find *any*
benchmarks that say a 400 Mhz Xeon is any better at Floating point
operations than a 400Mhz P-II... In fact they have the *same* core (FPU
included) and the primary difference between the Xeon and the P-II (and
P-III) is that the L2 cache on the Xeon runs at *cpu* speed, where the
L2 cace on the other run at 1/2 cpu speed... so, data access is somewhat
faster, but the floating point operations is not significantly
different...

Quote:

> See http://www.tomshardware.com/ they have a lot of benchmarks there...

Also take a look at the Spec homepage (http://www.spec.org) for
comparisons with *lots* of systems...

John

--
John Burton, Ph.D.
Senior Associate                 GATS, Inc.  


(757) 873-5920 (voice)           (757) 873-5920 (fax)

 
 
 

Speed..Speed..Speed

Post by Richard Payn » Thu, 11 Mar 1999 04:00:00


Quote:>If you intend to develope and run this software under LINUX then why are
>you even looking at a Intel chip.    Simply put NONE of the Intel chips are
>floating point standouts.     If your seriously considering chips in the
>XEON line, then a Alpha is not out of the price range, nor would a PowerPC
>based machine.     You might even be able to find a SPARC in this price
>range.     Start looking for Alpha powered units from Microway, ASPEN or
>DCG to name just a few.

>The nice thing about a totally open system like Linux is the freedom to
>choose the hardware of you choice.

Agreed. If you are looking for Alpha vendors we have a list of them
available,
including those that do/do not support Linux.

http://www.alphalinux.org/hardware/vendors.shtml

--
Rich Payne
(Speaking for myself, not my employer)
payner at timken dot com

Looking for Alpha-Linux info?
http://www.alphalinux.org

 
 
 

Speed..Speed..Speed

Post by Jeffrey J. Potof » Thu, 11 Mar 1999 04:00:00




> > Am currently running a K6-2 300Mhz processor with 128Mb of 100 Mhz
> > memory and considering
> > upgrading to a faster board and processor. I am pursuing a project which
> > will require scads of floating point
> > calculations on large arrays..the results of which are written back out
> > to disk..about 500 Mbytes worth.
> > So I'm looking at the new high end processors..PII/PIII and discover
> > this PII Xeon chip.

> > Does anyone have any experience with this chip? It very expensive..about
> > $900. A dual Xeon board from
> > SuperMicro with both processors runs over $2000. So how much faster is
> > this chip?  Intel's web site
> > says it has a 512K L2 cache (even more expensive versions have 1Mb and
> > 2Mb cache) which runs
> > at PROCESSOR SPEED .. 400 Mhz in this case. The comparable PII-400 chip
> > has a smaller(?) L2 cache
> > which runs at 1/2 chip speed. The comparable AMD K6-2 400 chip has a 1Mb
> > L2 cache but it is off chip
> > and accessed by the 100 Mhz bus.

> > So there you have it. Is the Xeon really worth double the PII price or
> > triple the K6-2 price?
> > As far as I can see.. the main difference is the L2 cache speed. Has
> > anybody seen any benchmarks
> > on this chip vis-a-via "comparable chips?

> > Jim M.

> Jim;

> If you intend to develope and run this software under LINUX then why are
> you even looking at a Intel chip.    Simply put NONE of the Intel chips are
> floating point standouts.

They aren't standouts, but they are cheap and get the job done.  I favor
the Alpha CPU for FP intensive work, but Intel chips are a cost effective
solution in many cases.

Quote:> If your seriously considering chips in the
> XEON line, then a Alpha is not out of the price range, nor would a PowerPC
> based machine.     You might even be able to find a SPARC in this price
> range.

Yeah, but why ?  Intel is cheap and will have a much better price/performance
ratio than any Sun platform.  These days it really seems that it's down to
Intel or Alpha if you are doing computational work.

Quote:> Start looking for Alpha powered units from Microway, ASPEN or
> DCG to name just a few.

> The nice thing about a totally open system like Linux is the freedom to
> choose the hardware of you choice.

Totally open, but the compilers aren't so hot.

Jeff

 
 
 

Speed..Speed..Speed

Post by Michael Ros » Thu, 11 Mar 1999 04:00:00





> > >which
> > >will require scads of floating point
> > >calculations

> > Rule of thumb need floating point use intel... BUT if you have money enough (and
> > if you are considering dual Xeon  2MB you have) take a look at alpha the new
> > processor 21264 is really the best money can buy... And if you run linux you can
> > save a bunvle in software...

> "Rule of thumb need floating point use Intel" ???? I've always heard
> just the opposite...need floating point, DON'T use Intel...

> Anyone have any Benchmarks on the 21264 ? Price / Mflop ?

> > Anyway if you need lots of IO get a SCSI subsystem maybe RAID level zero for
> > even faster performance (see DPT they have very nice caching controlers)

> Aggree with the SCSI, Have had trouble with DPT 2044UW controller &
> Linux... Seems at maximum transfer rate (20Mhz) I get rnadom SCSI Bus
> errors and occasional lock ups...the two drivers (EATA & EATA-DMA) under
> kernel 2.0.36 have *different* errors, but *neither* work without errors
> at full transfer rates...and yes, I have checked cables, devices,
> terminations, etc. I've swapped the card with an Adaptec 2940UW and
> didn't have any problems. I swapped the individual drives with new
> drives, and got the errors on the new drives. I've tried several cables,
> including those sold by Adaptec that were designed to work for maximum
> transfer rates on an UW bus...only consitent thing was DPT controller +
> 20Mhz SCSI Xfer rate = SCSI Bus errors (timeouts, device busy, etc...)

> > >So I'm looking at the new high end processors..PII/PIII and discover
> > >this PII Xeon chip.

> > PIII does not pay... Yet maybe with the new chipsets at 133Mhz FSB and faster
> > clock speeds... The new intructions will take _too_ long to be implemented by
> > compiler makers...

> True, the P-III does not have any new features that will cause any
> preformance increase, but the manufacturers I looked at are selling
> systems with them that are pricewise the same as a PII at the same
> speed... (i.e. a P-II 400 system is about $300 more than a P-II 350, and
> a P-III 450 is $300 more than a P-II 400, and a P-III 500 is $300 more
> than a P-III 450...) so it seems they are being priced as high speed
> P-II systems...

> > >As far as I can see.. the main difference is the L2 cache speed. Has
> > >anybody seen any benchmarks

> > And floating point, wich is very important to you...

> Ummm...where did you get your info? I haven't been able to find *any*
> benchmarks that say a 400 Mhz Xeon is any better at Floating point
> operations than a 400Mhz P-II... In fact they have the *same* core (FPU
> included) and the primary difference between the Xeon and the P-II (and
> P-III) is that the L2 cache on the Xeon runs at *cpu* speed, where the
> L2 cace on the other run at 1/2 cpu speed... so, data access is somewhat
> faster, but the floating point operations is not significantly
> different...

> > See http://www.tomshardware.com/ they have a lot of benchmarks there...

> Also take a look at the Spec homepage (http://www.spec.org) for
> comparisons with *lots* of systems...

> John

> --
> John Burton, Ph.D.
> Senior Associate                 GATS, Inc.


> (757) 873-5920 (voice)           (757) 873-5920 (fax)

Again, your mileage may vary. If your application is able to make
use of the Pentium III's new instructions for single precision f.p.,
and your application has significant time spent in loops that are
able to take advantage of the two at a time single f.p. capability,
then your app may see a significant perf. increase. The standard 80x87
instructions are not significantly different from Pentium II, that much
is correct.

--
__

Mike Ross

## Remove 1 on com above to reply, no spam
MicroComputer Software Labs
EY2-03
5350 N.E. Elam Young Parkway
Intel Corp.
Hillsboro, Oregon 97124

For information about Intel FORTRAN and VTUNE:

http://developer.intel.com/design/perftool/ifl24/ifl24wht.htm
http://developer.intel.com/design/perftool/VTUNE/fortenv.htm

For information about EPC FORTRAN 90:
http://www.epc.com

" I don't speak for Intel. All opinions are solely those of the author."

 
 
 

Speed..Speed..Speed

Post by Michael Ros » Thu, 11 Mar 1999 04:00:00




> <SNIP>

> > of 16.6, where a *single* 500mhz  P-III had a SpecFP95 of 14.7 and
> > SpecINT95 of 20.6. Supposedly, from the Spec benchmarks, a dual P-III
> > system *should* be better at floating point than the Alpha 21164 533
> > Mhz. (I'll believe it when I actually see it...;-)

> If the SMP box linerally scales, which it doesn't.  If your application isn't
> multi-threaded, the second processor will be of little use to you....

> Kent

I disagree. I routinely use a dual processor system for my work, and
find that having one processor essentially dedicated to the OS, and
the other running the app definitely helps. IMHO.
--
__

Mike Ross

## Remove 1 on com above to reply, no spam
MicroComputer Software Labs
EY2-03
5350 N.E. Elam Young Parkway
Intel Corp.
Hillsboro, Oregon 97124

For information about Intel FORTRAN and VTUNE:

http://developer.intel.com/design/perftool/ifl24/ifl24wht.htm
http://developer.intel.com/design/perftool/VTUNE/fortenv.htm

For information about EPC FORTRAN 90:
http://www.epc.com
" I don't speak for Intel. Opinions are solely those of the author."

 
 
 

Speed..Speed..Speed

Post by José Rui Faustino de Sous » Thu, 11 Mar 1999 04:00:00



Quote:>"Rule of thumb need floating point use Intel" ???? I've always heard
>just the opposite...need floating point, DON'T use Intel...

I was not clear it should say: if you need x86 compatibility...

Quote:>SCSI Bus errors (timeouts, device busy, etc...)

Life is dangerous... I would never say that you wont have any problems...

But lets face it most people dont...

Quote:>Ummm...where did you get your info? I haven't been able to find *any*
>benchmarks that say a 400 Mhz Xeon is any better at Floating point
>operations than a 400Mhz P-II... In fact they have the *same* core (FPU
>included) and the primary difference between the Xeon and the P-II (and
>P-III) is that the L2 cache on the Xeon runs at *cpu* speed, where the
>L2 cace on the other run at 1/2 cpu speed... so, data access is somewhat
>faster, but the floating point operations is not significantly
>different...

Again I was comparing Intel to all the others (x86 compatible)...

========================================================
Jos Rui Faustino de Sousa
========================================================

http://homepage.esoterica.pt/~jrfsousa/
telephone://+351-(0)39-444940
SnailMail://rua Carlos Alberto Pinto de Abreu, no30C, 1o
3040 Coimbra Portugal
========================================================