[R] Which is the best hardware?

[Prof Brian Ripley]

On Mon, 5 Dec 2005, Kenneth Cabrera wrote:

> Hi R users:
>
> In your opinion and experience, which hardware configuration
> is the best to run R over LINUX ?
>
> With "best" I mean best performance,
> and also cheapest. (about U$ 2.000 the whole basic system:
> mother board+CPUs+RAM+HD)

I presume you don't need a display or keyboard ....

Well, prices depend on where you are and the quality of components, e.g. 
power supplies.

But as I am just buying a new system I have some idea. I would suggest an 
Athlon 64 X2 would be a good choice: that's a 64-bit system with de facto 
two processors which can be bought here with 2Gb RAM at well under your 
price.

> By the way, which LINUX distribution is the best to
> run R with high computing technics (simulation, bayesian, etc) and huge data 
> base?
> and in combination with what kind of (cheap) hardware?

They are all based on the same components: a distribution is just the 
packaging and installation tools.  For performance it seems that systems 
based on gcc3 rather than gcc4 still have a small edge, but I would say 
local expertise is far more important.  (At one point in the committee for 
a large procurement I pointed out that a 10% difference between two 
systems was 2.5 months' of Moore's Law, and that covered the spread of 
benchmark results for all the contenders.  So if you want better 
performance, just wait a few months.)

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595