Yvonne Aberg ([email protected]) asked
> I currently have a rather powerful HP Vectra (Pentium 4, 1.8 Ghz, and 2 Gb
> in RAM), but I need to use large data sets (2-3 gb) and I am wondering
> whether one should consider switching to the unix world. I do not have any
> previous experience of working with Stata for Unix so it would be
> interesting to know what other users think about this. Which platform should
> one primarily consider when cost is a major concern? Are Windows-based
> work-stations an alternative?
PCs running on 32-bit Intel hardware, whether using Windows or
Linux, have a theoretical maximum memory allocation for a single
process of 2 GB. In practice, this number is slightly less than
2 GB under Linux, and sometimes as low as 1.4 GB under Windows.
Often when people say they have datasets of a certain size, they
may be talking about raw ASCII files, which are larger than
the same information stored in a Stata dataset. If Yvonne's datasets
in Stata format go up to no more than 1.5 GB, then Linux might be
a better alternative than Windows due to Windows not always being
willing to give applications that much memory.
If Yvonne truly needs to work with Stata datasets that are 2 to 3
GB, then 64-bit hardware must be used to break past the 2 GB limit
of 32-bit machines. Currently 64-bit versions of Stata exist for
Sun Solaris, Silicon Graphics Irix, and Compaq Tru64 Unix (also
known as Digital Unix or the DEC Alpha). Such machines will be
able to handle Yvonne's memory requirements, but their price to
performance ratio is not as good as that of typical 32-bit
Intel or AMD hardware.
If cost is a major concern, it may be impractical to move to
such hardware. In this case, I would recommend carefully
examining the datasets and analysis to be performed and writing
helper do-files that load only the data that is needed at any
given time or that process the data in pieces that will fit within
the 2GB 32-bit limit.
--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/
