Subject: Re: Memory location mapping/marking
To: Andrew Gillham <gillham@vaultron.com>
From: W. Suika Roberts <ssfr@unm.edu>
List: port-i386
Date: 10/05/2000 15:50:11
On Mon, Sep 25, 2000 at 02:21:10PM -0400, Andrew Gillham wrote:
> W . Suika Roberts writes:

>> Because she is stable at 16 MB, and exhibited similar behavior under
>> linux I'm pretty sure she has a ticklish spot just past 16MB.

>> I'm hoping that someone has a simple way to:
>> 1) find this ticklish spot
>> and
>> 2) tell the system not to touch it

> First try 'memtest86' (search on google.com) to verify your memory
> is ok.  Secondly do you have any device that is an ISA busmaster?
> (I doubt it, but..)  It could be some kind of bouncebuffer issue.

The only one that might be a busmaster is the VLSI PCMCIA controller -- it is
an intel clone that supports DMA for the floppy disk.

memtest86 says everything is fine at 15 ms refresh rate, and that the 16 MB
card is flakey as all get out at 150 ms.

I'd guess that the first 16 MB refresh fast enough to maintain stability under
normal usage, and that PCMCIA access somehow slows down the refresh, so that
all 20 MB aren't refreshed before something tries to read lossy data.
(the system is stable with all 20 MB, so long as I avoid using PCMCIA)

I hadn't realize that software can affect the refresh rate on the memory, so
is there a way to increase the refresh rate?

It's not like the current situation is intolerable, but that extra four meg
could be useful.

	Suika