Statalist The Stata Listserver


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: Stata/MP on quad cores


From   [email protected] (Vince Wiggins, StataCorp)
To   [email protected]
Subject   Re: st: Stata/MP on quad cores
Date   Fri, 20 Apr 2007 09:50:30 -0500

Stas Kolenikov <[email protected]> asks,

> [...] in terms of parallelizing the workload, is one quad-core
> processor better with Stata/MP than two dual-cores?

It is hard to say, other than recognizing that many quad-core processors are
newer than their dual-core brethren and therefore may be faster and have
larger caches.  

Processor architectures differ in many ways and these ways are not limited to
the difference between dual- and quad-core versions.  Even so, we have not
found many consistent performance difference among processor architectures or
the multi-core vs. multi-processor computers.  We see differences, they just
do not generalize across all commands or even across problem sizes.

Why?  The differing use of fast, on-board cache memory is the main reason that
different architectures (given processor speed) perform differently, and as
problem size varies the effect of cache memory varies.

For a given number of cores or processors, our recommendation is that first
you get the fastest processors you can and that second you get processors with
the largest available on-board cache memory.

> My understanding is that for either configuration, I would need a
> four-thread MP license, and then for two-processor quad-core,
> 8-thread license.

That is right.  Stata/MP licenses are based on the number of processor cores,
whether the cores are all on one processor or spread across many.

 
-- Vince 
   [email protected]

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index