MathGroup Archive 2013

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

Search the Archive

Re: AMD Benchmark[]

  • To: mathgroup at smc.vnet.net
  • Subject: [mg130619] Re: AMD Benchmark[]
  • From: Roman Pearce <rpearcea at gmail.com>
  • Date: Fri, 26 Apr 2013 04:27:16 -0400 (EDT)
  • Delivered-to: l-mathgroup@mail-archive0.wolfram.com
  • Delivered-to: l-mathgroup@wolfram.com
  • Delivered-to: mathgroup-outx@smc.vnet.net
  • Delivered-to: mathgroup-newsendx@smc.vnet.net
  • References: <kklfij$7q3$1@smc.vnet.net>

Sorry this is not a direct answer.  We bought the FX-8350 (piledriver) for our lab and we were surprised by how good the performance was compared to the (older) Core i7 2600k.  If you can use 8 cores regularly it's a good cpu at a great price.  The Core i7 3770k is more expensive and faster for single threaded computations.  For parallel workloads the FX-8350 was very good, although we mainly tested the integer units and not the floating point units, which I understand are shared.

I like the FX-8350 for mid-range desktops with plenty of RAM and SSD.  Intel has faster cpus but they are more expensive.

A nasty situation can occur with older software compiled by the Intel compiler.  In some cases by default it does not use optimized code paths on AMD cpus.  I'm not sure what the situation is currently, or whether Mathematica is affected.  Maybe someone can comment.  See:

http://www.agner.org/optimize/blog/



  • Prev by Date: Re: Export to PDF outline issues
  • Next by Date: Re: Speak errors (was Re: audio)
  • Previous by thread: Re: AMD Benchmark[]
  • Next by thread: Re: AMD Benchmark[]