Buzz Burhans ([email protected]) asked about Stata for Linux64 running
on AMD's new 64-bit chips:
> If speed is an issue (for instance, running -gllamm- on small datasets for
> models with more than 2 random effects), can anyone advise me on:
>
> 1. How much improvement can I expect running the AMD64 and -gllamm- as above?
It is hard to tell you exactly how much improvement you might see without
a comparison of two different machines given a specific problem.
I can tell you that in testing here at Stata, we have found that an
Opteron 246 running at 2 GHz can run our entire suite of certification
scripts in approximately 30% less time than a Pentium 4 running at 2.8 GHz.
We are, to say the least, very impressed with the performance of the
Opteron.
> 2. Any observations/ suggestions/possibilities on specific sources of
> inexpensive AMD64 systems as described above?
If you want to use a lot of memory, you won't find a truly inexpensive
system. You can buy a base Opteron system for under $2000. However,
when you add, say, 8 GB of RAM to it, you will bump that price way up.
For example four 2 GB chips will likely set you back somewhere in the
neighborhood of $3000, not including the rest of the computer.
> Is this combination (the new Stata for these machines) a worthwhile
> investment for those of us running PCs, will I get a substantial
> improvement in performance...i.e. speed?
If you are comfortable and happy working in a Unix/Linux environment,
the answer is yes. I am not sure that I would recommend that a Windows
user switch to Linux64 on an Opteron just for the speed improvement--it
might be too much of a culture shock for them.
If Buzz has some specific -gllamm- speed comparisons he would like to see,
we would be happy to run a 'quick' comparison for him on an Opteron and
on an Intel-based computer and report the results to the list. (Buzz,
email me if you want to set something like this up.)
--Alan
([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/
