On 902244688 seconds since the Beginning of the UNIX epoch
Chris G. Demetriou wrote:
>
>You "need" more than 16MB of memory.  (I seem to recall ross saying that the s
>ystem got nearly, but not quite, to single-user for him on a 16MB system, but 
>even if it does get there it'll spend all its time paging.)

The system actually had 5 days of uptime at 16MB, but that is probably
because I use it for only two purposes:  as my alarm clock (an at job which
invokes mpg123 :) and cracking rc5.  It does truth be told support a couple
of users infrequently logging on.  I only seem to run into problems when
doing compiles.  This is the only thing that starts to give uvm a work out.

Sorry that I didn't notice in the INSTALL notes that I should have 32MB..
I do wonder though, whether the low memory condition exacerbate a uvm
bug, or simply fails in an unpreventable way.  As I mentioned above, it
seems to work really well for quite a while before it crashes.  (Otherwise
I would still be running the snapshot, unable to compile kernels.  :)

>I'd advise 48 or 64 if you can swing it, but, depending on what you're
>doing you may be OK with 32.  (I've got 32 in a machine that I use as
>a web and POP server, and it runs file.)  For a Long Time I did all my
>builds on a 64MB AlphaStation 200, running X, etc.

An aside:
 I tried to put another 64MB into it today, but was thwarted by a couple
 of mistakes that I made.
	o I went to Fry's.
	o I did not check the memory before I left the place.
 They didn't have 32MB 72 pin parity SIMMS, so I ``bought'' two 16MB 72
 pin parity SIMMS.  When I got home, I discovered that what I had was
 two 32MB DIMMS.  Which don't go into the machine in question.

 == Roland
 == http://www.imrryr.org/~elric/