[comp.sys.att] WANTED: 3b1 ram expansion

wilber@alice.UUCP (08/17/88)

<In article <591@proxftl.UUCP>, aaron@proxftl.UUCP writes:
<> I have a 3b1 (2mb ram, 67mb HD), and I really need another 2 megs or so. Does
<> anyone know where I can get this or have some for sale?
<
<I'm not _entirely_ certain, but a knowledgable person (someone who
<used to be an _official_ repair person for the 3b1) told me that
<the MMU in the 3b1 could only handle 2.5 megs of real memory.
<Anyone know anything about this?
<--------
<	Christopher J. Calabrese
<	AT&T Bell Laboratories
<	ulysses!cjc

You can have up to 4Mb of real memory on a 3b1.  In practice the limit is
usually 3.5 Mb, since most people would rather get a combo board with 1.5 Mb
and two serial ports than a RAM board with 2 Mb and no serial ports.  Because
of a limitation of the MMU, the maximum user memory *per process* is 2.5 Mb.
At least that is how I understand the situation.

Bob Wilber

fst@mcgp1.UUCP (Skip Tavakkolian) (08/20/88)

In article <8118@alice.UUCP>, wilber@alice.UUCP writes:
><In article <591@proxftl.UUCP>, aaron@proxftl.UUCP writes:
><> I have a 3b1 (2mb ram, 67mb HD), and I really need another 2 megs or so.
><> Does anyone know where I can get this or have some for sale?
><the MMU in the 3b1 could only handle 2.5 megs of real memory.
><Anyone know anything about this?
><--------
><	Christopher J. Calabrese
><	AT&T Bell Laboratories
><	ulysses!cjc
> You can have up to 4Mb of real memory on a 3b1.  In practice the limit is
> usually 3.5 Mb, since most people would rather get a combo board with 1.5 Mb
> and two serial ports than a RAM board with 2 Mb and no serial ports.  Because
> of a limitation of the MMU, the maximum user memory *per process* is 2.5 Mb.
> At least that is how I understand the situation.
> Bob Wilber


There is an old BYTE review you may want to look for (this was for the original
7300 several years back).  Here is what I recollect. The MMU in 3b1/7300 has
24 address lines (I am not sure if it is 68451s).  The difference in 7300 and
the 3b1 (mother board designs) is in the number of address lines which are
ignored (out of the possible 2^24 = 16777216).  That, I believe, is the only
difference in 7300 and 3b1. So that 7300 MMU can address no more than 2 MB
and the 3b1 no more that 4 MB.

This however should not make any difference in the per *user* process address
space in a virtual memory system (demand page allows sizes smaller than a
whole process to be swapped in and out of memory).

Maximum per user process address limit is what the 68010 can address.

Hope this is all true :-)

Sincerely
-- 
Fariborz ``Skip'' Tavakkolian
UUCP	...!uw-beaver!tikal!mcgp1!fst

UNIX is a registered trademark of AT&T

ditto@cbmvax.UUCP (Michael "Ford" Ditto) (08/21/88)

In article <1525@mcgp1.UUCP> fst@mcgp1.UUCP (Skip Tavakkolian) writes:
>There is an old BYTE review you may want to look for (this was for the original
>7300 several years back).  Here is what I recollect. The MMU in 3b1/7300 has
>24 address lines (I am not sure if it is 68451s).  The difference in 7300 and
>the 3b1 (mother board designs) is in the number of address lines which are
>ignored (out of the possible 2^24 = 16777216).  That, I believe, is the only
>difference in 7300 and 3b1. So that 7300 MMU can address no more than 2 MB
>and the 3b1 no more that 4 MB.

No, the only difference between the 7300 and 3b1 motherboards (aside from
some pcb relayout) is the amount of ram installed).  The other differences
are in power supply rating, case style and disk size.

>This however should not make any difference in the per *user* process address
>space in a virtual memory system (demand page allows sizes smaller than a
>whole process to be swapped in and out of memory).
>
>Maximum per user process address limit is what the 68010 can address.

The amount of physical RAM addressable by the MMU is 4MB.  The MMU is a
custom random-logic MMU based around a 1K by 16bit ram array.  The top
two 68010 address lines are used for selecting I/O space and other things
not related to RAM.  That leaves 22 bits for the virtual address, giving
a per-process virtual address space of 4 Meg.  Some of this address space
is used for shared memory, the shared library, and for overlaying the
kernel when it is running.  2.5 Meg (from 0x080000 to 0x3fffff) is available
for user code and data.

>Hope this is all true :-)

Well, it's mostly true with respect to 68000's in general, but the Unix PC
is a bit weird.
-- 
					-=] Ford [=-

	.		.		(In Real Life: Mike Ditto)
.	    :	       ,		ford@kenobi.cts.com
This space under construction,		...!ucsd!elgar!ford
pardon our dust.			ditto@cbmvax.commodore.com

alex@umbc3.UMD.EDU (Alex S. Crain) (08/22/88)

	This summerizes the results of the last discussion we had on this, and
is all true. (What a statment :-))

	The 7300/3b1 are identical with regard to expansion limits, they differ
in the amount of memory that the come with.

	The 7300/3b1 is allowed 4meg memory. 2meg on the mother board (the 
7300 needs a mmu support to go from 512k->2meg, but it is available), and *one*
expansion card, which can be as large as 2meg, and must be in 512k increments.

	The 7300/3b1 has a 2.5meg limit on user addressable memory per process.
this is because of the way memory is mapped, the addressable user area is
bounded by the kernal (low memory) and the shared library/dynamic kernal memory
(high memory). 

-- 
					:alex.
					Systems Programmer
nerwin!alex@umbc3.umd.edu		UMBC
alex@umbc3.umd.edu