[R] 8 fast or 4 very fast cores?

[Clint Bowman]

I'm in a similar situation and am looking seriously at a pair of E5-2643v3 
(6 cores each-hyperthreaded).

Clint Bowman			INTERNET:	clint at ecy.wa.gov
Air Quality Modeler		INTERNET:	clint at math.utah.edu
Department of Ecology		VOICE:		(360) 407-6815
PO Box 47600			FAX:		(360) 407-7534
Olympia, WA 98504-7600

         USPS:           PO Box 47600, Olympia, WA 98504-7600
         Parcels:        300 Desmond Drive, Lacey, WA 98503-1274

On Mon, 15 Sep 2014, Prof Brian Ripley wrote:

> On 15/09/2014 11:21, Ben Bolker wrote:
>>  Leif Ruckman <Leif <at> Ruckman.se> writes:
>> 
>> > 
>> >  I am going to buy a new computer ( Dell workstation T5810 - Windows 8)
>> >  to work with simulatons in R.
>> > 
>> >  Now I am asked what kind of processor I like and I was given two 
>> >  choices.
>> > 
>> >  1. Intel Xeon E5-1620 v3 - 4 cores 3.7 GHz Turbo
>> >  2. Intel Xeon E5-2640 v3 - 8 cores 2.6 GHz Turbo
>> > 
>> >  I don't know what is better in simulations studies in R, a few very fast
>> >  cores or many cores at normal speed.
>> 
>>
>>     It's **very** hard to answer such general questions reliably, but I'll
>>  take a guess and say that if you're doing simulation studies you're likely
>>  to be doing tasks that are easily distributable (e.g. many random
>>  realizations of the same simulation and/or realizations for many
>>  different sets of parameter values) and so the more-cores option
>>  will be a good idea.
>>
>>     But it's possible that what you mean by "simulation studies" is
>>  different.
>>
>>     If you can do some benchmarking of your problems on an existing
>>  machine that would probably be a good idea.
>
> Unfortunately unless it is of very similar architecture that may not help 
> much.
>
> Three issues hard to scale from are the 'Turbo', the hyperthreading of modern 
> Xeons and the cache sizes.  Now, I happen to have machines with multiple 
> E5-24x0 and E5-26x0 Xeons: both do hyperthreading well, so you would have 8 
> or 16 virtual CPUs and they will give you say 50% increase in throughput if 
> all the virtual cores are used.  But you cannot scale up from using just one 
> process on one core.
>
> I find it hard to think of tasks where option 1) would have more throughput, 
> but if most of the time you are not running things in parallel then the 
> higher speed on a single task is a consideration.
>
>>
>>     Ben Bolker
>>
>>  ______________________________________________
>>  R-help at r-project.org mailing list
>>  https://stat.ethz.ch/mailman/listinfo/r-help
>>  PLEASE do read the posting guide
>>  http://www.R-project.org/posting-guide.html
>>  and provide commented, minimal, self-contained, reproducible code.
>> 
>
>
> -- 
> Brian D. Ripley,                  ripley at stats.ox.ac.uk
> Emeritus Professor of Applied Statistics, University of Oxford
> 1 South Parks Road, Oxford OX1 3TG, UK
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>
>