Hardware to optimize Stata speed

[Ernestina delPiero]

Hello. I am wondering what hardware specifications (CPU, HDD, RAM, etc.) one should aim to optimize if the goal is to have Stata code run faster. i, for one, do a lot of 'reshape' and 'statsby' command. What specifications would make these commands execute faster?

Thank you.

[Daniel Schaefer]

It's hard to say because it depends somewhat on the underlying implementation. In general linear algebra operations are well suited to optimization for parallel processing. Assuming stata is optimized for parallel processing I guess you might lean towards processers with more cores or greater multithreading capacity. Also note that in their hardware requirements page they recommend a CISC processor over an RISC processor.

You should have no less than 32 gigs of ram, but I'd recommend 64 gigs. You want to avoid using swap space or virtual memory on your hard drive at all costs since this will dramatically slow processing time. I'm not sure if stata consumes the gpu. This thread is helpful in general but doesn't clarify that point. I'd guess the on-board GPU is just fine for most applications, so you may be able to skimp on the graphics card. A solid state drive will load data into memory much faster (although the cable is the real bottleneck, especially relative to the CPU bus), but the downside is that if you are writing data to the hard drive often the hard drive won't live as long as a disk drive, and may be somewhat more prone to failure.