Stata/MP

[omerlast]

Hello,

I want to buy Stata/MP but i have no idea about suitable computers which is good for Stata/MP. Is there anybody who use Stata/MP? What are your computer technical specifications? Are you satisfied?

Sincerely

[wbuchanan]

I had Stata 12MP2 running on a six core Dell box running windows 7 as well as on my macbook aluminum pro 17" with a dual core cpu. Recently I upgraded to Stata 13MP8 and use it primarily on a new MacBook Pro with a 2.8Ghz quad core and can safely say that I would find it difficult/painful to go back to working with the IC/SE versions of the software. I typically work with datasets that have 500,000-6,000,000 observations and the performance gains are fairly substantial from what I've experienced. If you tend to work with smaller datasets and/or don't use many computationally intensive functions it may not be quite as apparent, but it is otherwise impressive to say the least.

[Rich Goldstein]

I use Stata 13 MP/4 on a Mac DeskPro and on a MacBook Pro (neither is new); note that I actually have 8 cores on the desktop machine but was not impressed with the difference between MP4 and MP8; the difference between MP4 and MP2 is, however, noticeable in my work

[omerlast]

Thank you wbuchanan and Rich Goldstein. My datasets varies between 12,000-550,000 observations. It takes a lot time, for example, for Pesaran CD test or Shapley decomposition. My book is i3, 4 gb ram, 4 core. My stata is 12/se. It can not even calculate Shapley decomposition, which is based iteration, if independent variable number bigger than 15. It again takes a lot time for -xtgls- if i put all (49) independent variables. I guess i have to buy stata 12/mp? Can we say that if i buy stata 12/mp solve time issue for this computer? Or, is there another problem with my compouter?

[Rich Goldstein]

you should look first at http://blog.stata.com/tag/parallel/ and then click the link to the MP performance report

[ben earnhart]

If it can't do it at *all* then it might be RAM, not cores. In general, IC/SE will chug away, take more time, but eventually get the job done. MP will speed things up, but if it's hitting a wall and completely barfing, I'd seriously consider RAM issues. But, if it's just taking too long, then MP could help. On a Mac, I don't know how to monitor memory usage from outside of Stata, and you can't run memory while in the middle of a single command, but if you can somehow figure out how much RAM is being used, that would help you diagnose what's going on. From your posting, sounds like for the Shapley decomposition (fails to run) the problem might be RAM, whereas xtgls (runs slow) might benefit from MP.

[omerlast]

Thx ben earnhat. I guess me problem is bicuspid; ram and core. Just stata/mp can not solve my problem, i upgrade my computer at the same time.