![]() So you might get, say, 1.8x with two cores, 2.5x with four, and 3x with eight, and so on, with negligible improvements after four cores. ![]() This is a strange claim: most of the commands in Stata exhibit severe diminishing returns to number of processors. ![]() I'm not sure if the gain is as big as you would expect it to be and honestly I think you are underusing Stata MP. I used it on a 24 core server and it made a gigantic difference for me. Stata MP is great when you have the setup to use it. If regress is indeed nearly linear in the cores, then using it on a 24-core server would decrease the time from 24 hours to about one hour, no? From experience, I know these can sometimes take a very long time to run. Let's say he was doing a large fixed effect regression with one million+ observations.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |