Mathematica 9 is now available
Services & Resources / Wolfram Forums
-----
 /
MathGroup Archive
2005
*January
*February
*March
*April
*May
*June
*July
*August
*September
*October
*November
*December
*Archive Index
*Ask about this page
*Print this page
*Give us feedback
*Sign up for the Wolfram Insider

MathGroup Archive 2005

[Date Index] [Thread Index] [Author Index]

Search the Archive

Mathematica 5.1.1 benchmark

  • To: mathgroup at smc.vnet.net
  • Subject: [mg58282] Mathematica 5.1.1 benchmark
  • From: Ronald Bruck <bruck at math.usc.edu>
  • Date: Sat, 25 Jun 2005 01:56:51 -0400 (EDT)
  • Sender: owner-wri-mathgroup at wolfram.com

A few months ago I bought a Compaq R3470US laptop (which uses an AMD64
3400+ CPU).  I installed 64-bit Windows and Linux x86_64 and was
pleasantly surprised to observe that my University's site license
included the 64-bit Windows version of Mathematica.

The laptop so exceeded my expectations, especially in its performance
with the Gnu Multi-Precision Library (the GMP website,
<http://swox.se/gmp>, explains why the 64-bit wordsize makes it FOUR
TIMES as fast as my 3-GHz P4), that I invested in an FX-55, currently 
the fastest AMD64 CPU.  And I upgraded my personal license for
Mathematica so I could run it in Linux x86_64, which the University
site license doesn't cover for personal use.

My purpose today is to advise you that if you need a FAST machine, the
FX-55 is the way to go.  I used Karl Unterkofler's benchmark, with the
following results for five independent runs.  (Karl's website contains
the notebook to run, and what the individual numbers measure.  Note
that to avoid caching effects, you MUST quit Mathematica after each
run, or at least start with a fresh kernel; I quit.)

> Mean(Times) = {2.58361, 2.13628, 3.06134, 2.07988, 3.37049, 
>   2.97875, 2.84997, 4.5761, 14.0921, 6.06048, 26.7081, 5.57835, 
>   4.19636, 5.9099, 7.19871}
> 
> StandardDeviation(Times) = {0.0205513, 0.00181412, 0.00178885,
> 0.00683184, 
>   0.001, 0.0142684, 0.0322963, 0.0269155, 0.0222868, 0.011327, 
>   0.0342155, 0.0427091, 0.00796869, 0.00704982, 0.0282381}
> 
> Benchmark mean = 5.22122, StandardDeviation = 0.00960831

That's with respect to a current reference of a 1GHz G4 processor.  The
previous fastest speed reported on Karl's benchmark site,

   <http://www2.staff.fh-vorarlberg.ac.at/%7Eku/karl/timings50.html>

was 4.23, on an FX-53.  (The 13 fastest machines are ALL Athlons.  The
fastest Mac G5 was a dual 2.5GHz G5, with a benchmark of 2.819.)

Technically, my FX-55 is a 2.8GHz machine.  However, I have overclocked
it slightly, and am running the memory at 205MHz (that's 410MHz, DDR). 
It SHOULD do 433MHz DDR, and the CPU is only running at 114 degrees F,
so there's still room for improvement.  These measurements are probably
equivalent to a stock FX-57, the successor to the FX-55, which may
already be available.

Many of my calculations (not just Mathematica) run for 24 to 48 hours. 
The availability of speedups like this is fantastic news to me.  64-bit
ROCKS.

As for Wolfram:  congratulations on staying ahead of the curve in OS's. 
I watched the Mathematica presentation at the Apple World Developer's
Conference, and I look forward to running Mathematica on Apple Intel. 
I just hope there's also an Apple AMD :-)

The 3.06-GHz P4?  My 4-year-old granddaughter inherited it.  But INTEL
will undoubtedly leapfrog AMD again.  This is a win-win situation for
consumers.

--Ron Bruck


  • Prev by Date: Re: Re: Asking questions
  • Next by Date: Re: General 3-state stochastic matrix (again)
  • Previous by thread: Re: Solve yields unreal solutions, NSolve yields real solutions
  • Next by thread: Problems with some Graphics`Polyhedra` functions