Subject: Re: IP20 memory suckage
To: Gerald heinig <Gerald.Heinig@ngi.de>
From: Steve Rumble <rumble@ephemeral.org>
List: port-sgimips
Date: 02/06/2004 10:36:06
On Fri, Feb 06, 2004 at 09:38:02AM +0100, Gerald heinig wrote:
> That doesn't seem to work. I took out one SIMM and it complained on
> bootup about needing a fully-populated bank. I can't remember the exact
> wording, but it implied that you had to have at least 4 SIMMS installed.

As Alban said, each bank {A,B,C} needs to have four identical
simms, or none at all. The banks should be aligned like:
        --------   -------- A
        --------   -------- A
        --------   -------- B
        --------   -------- B
        --------   -------- C
        --------   -------- C

If memory serves correctly, that is. The order may be reversed,
but you get the idea. Hopefully there's some marking on the board
itself.

From the sound of it, you only have one bank populated, in which
case this step won't be terribly helpful.

> Correct. There's nothing to port: it compiled out of the box.
> And detected no errors at all. However, it only found 20 megs of free
> memory in a 64 meg machine. I reckon the system is detecting 4 16 meg
> SIMMs, of which 2 are broken, so I effectively only have 32 megs. That's
> the only scenario that makes sense.

hinv and dmesg only report 32 as well?

> Amazing. SGI using standard parts?!? The horror! :)

The R4k indigo, indy and indigo2 all use the same memory, I
believe. The R3k, 4d/2x and 4d/3x machines used some wonderful
proprietary sticks. I haven't a clue what older machines made use
of.

You may be able to appropriate some sticks from an old 386 or
486. 

If you can't locate any memory, I can mail you a bank from my
indigo when I'm back home.

Cheers,
Steve