Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: Re: memory management in Win2000

From (Kevin Turner)
Subject   st: Re: memory management in Win2000
Date   Fri, 11 Jul 2003 10:46:14 -0500

Steven Stillman <> wrote: 

>>This is more of a windows questions than a stata question.  I frequently
>>work on large datasets (150-200m) on a Pent IIII PC w/ Win2000 and 512m ram.
>>I find that it takes a long time (3-5 minutes) for these large datasets to
>>initial open (stata is very fast once the datasets are loaded).  Does anyone
>>have any suggestions on how to get Win2000 to allocate memory quicker to
>>stata (this seems to be what holds things up)?  I was thinking that ala Unix
>>there might be a way to assign a fixed block of ram to stata instead of it
>>being dynamically reallocated every time a dataset is opened.

And Friedrich Huebler <> had this suggestion:

>Do you use Stata 8 with a dataset from an earlier version of Stata?
>Try to convert the dataset to Stata 8 format and see if it takes less
>time to open.

This suggestion is a very good one. Using the latest format of dataset for your
version of Stata is always faster.

Another suggestion:
Due to the fact that it takes so long and that the size of the dataset is
already a large percentage of the physical RAM (200/512), the odds are that
Windows is being forced to use virtual memory. You can verify this by using the
Windows Task Manager (typically you can right-click on the task bar to bring up
the Task Manager; Ctrl-Alt-Del is another way to get to it). On the Task
Manager, there is a tab for Performance and on this tab is a couple of tables
showing memory usage. If your Total Commit Charge memory is higher than your
Total Physical memory, you are in a state where Windows is forced to use
virtual memory. 

You can *probably* witness the problem by viewing the Available Physical Memory
_during_ the load of the Stata dataset -- it should hover around 6-3Mb as
Windows continually swaps memory from Physical to Virtual space. 

	1. Buy more RAM. 
	2. Close all unnecessary applications while you are working with these
	   large datasets.
	3. There are a few more odd tricks that can be done to speed up 
	   virtual memory, but may not fix the entire problem: 

Hope this helps,

*   For searches and help try:

© Copyright 1996–2015 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index