[comp.sys.amiga] Another A3000 RAM question

davids@cup.portal.com (David Kenneth Schreiber) (08/10/90)

I've been trying to install an extra meg of RAM in my 3000, but have been
thwarted by (I think) a bad memory chip.  I've got bank 0 filled with the
DIPs that came with the machine;  in bank 1 I've installed the ZIPs.  When
I power up, Kickstart 2.0 loads, the machine is reset, then I get a 
System Error #80000003 at location 07e03234 (well, somewhere in the
07e03000's;  I believe 07e03234 is the correct address).  Can anyone
tell me what socket a bad ZIP would have to be in to cause an error
at that address?  I've tried swapping chips around, but the same error
always shows up at the same place.

The chip types, by the way, are Fujitsu MB81C4256-80PSZ and Mitsubishi
M5M44256AL.  Both are 80ns, page mode, 256x4 RAMs.  The Mitsubish has one
pin missing, though (the technical support person at Krueger said this
was not a problem, that some ZIPs are just made that way;  is he
right?).

Thanks.

-Dave Schreiber at davids@cup.portal.com         "Coffee, Darling?"

P.S.  Kickstart/Workbench 1.3 loads fine, and Memtest V2.0 reports no
errors.

daveh@cbmvax.commodore.com (Dave Haynie) (08/10/90)

In article <32628@cup.portal.com> davids@cup.portal.com (David Kenneth Schreiber) writes:
>I've got bank 0 filled with the DIPs that came with the machine;  in bank 1
>I've installed the ZIPs.  When I power up, Kickstart 2.0 loads, the machine 
>is reset, then I get a System Error #80000003 at location 07e03234 (well, 
>somewhere in the 07e03000's;  I believe 07e03234 is the correct address). Can
>anyone tell me what socket a bad ZIP would have to be in to cause an error
>at that address?  I've tried swapping chips around, but the same error
>always shows up at the same place.

With 256K x 4 chips, it takes 8 chips to make up one 1 megabyte, which is
of course organized as 256K x 32.  From your description, the only thing I
can tell is that something in the second bank of chips, the ZIPs you added,
may be amiss.  The A3000 memory builds down from $07ffffff as you add it.
The base machine comes with 1 Meg located in $07f00000-$07ffffff.  The next
bank is located from $07e00000-$07efffff.  All 8 chips go to make up that
bank, so a problem in any one of them will affect the entire 1 Meg chunk.
Before you panic too much, make absolutely sure that you're putting these
parts in the right sockets; the arrangement of sockets doesn't make much
sense to anyone who didn't have lay out the circuit board.

>The Mitsubish has one pin missing, though (the technical support person at 
>Krueger said this was not a problem, that some ZIPs are just made that way;
>is he right?).

Yes, that's correct.  The 256K x 4 parts have one unused pin, some makers 
provide a useless pin there, others leave the pin off.  That pin is the
extra address pin for the 1 Meg x 4 parts.

>P.S.  Kickstart/Workbench 1.3 loads fine, and Memtest V2.0 reports no
>errors.

Ah, so.  Check the parts your machine came with; the DRAM in the first bank.
If these read something like "44258" rather than "44256", you have a known
problem.  The A3000's memory controller, the RAMSEY chip, has a special
memory mode for support of SCRAM (Static Column DRAMs) memories to support
burst mode.  This special memory mode, however, has the unfortunate effect
of confusing the hell out of normal page-mode memories.  If there are any
page-mode memories in system, even just one, the static column mode must
be disabled for the entire Fast memory system.  This detection has to be
done in software.  The A3000's version of 1.3 doesn't bother with trying to
determine this, it simply leaves the SCRAM modes off.  Under the current
2.0, they are doing some kind of SCRAM check, but it's not very robust.  So
SCRAM memories in the first bank will cause the SCRAM mode to be set, no
matter what's in the latter Fast memory banks.  And I have heard that some
A3000s are now going out with the faster SCRAMs in the first DIP bank.  So
my guess is that you have SCRAMs in your first 1-meg bank.
-- 
Dave Haynie Commodore-Amiga (Amiga 3000) "The Crew That Never Rests"
   {uunet|pyramid|rutgers}!cbmvax!daveh      PLINK: hazy     BIX: hazy
      Get that coffee outta my face, put a Margarita in its place!

davids@cup.portal.com (David Kenneth Schreiber) (08/13/90)

[foo foo fodder fodder]

In article <13750@cbmvax.commodore.com> daveh@cbmvax.commodore.com (Dave Hayni
e) writes:
.In article <32628@cup.portal.com> davids@cup.portal.com (David Kenneth Schrei
ber) writes:
.>I've got bank 0 filled with the DIPs that came with the machine;  in bank 1
.>I've installed the ZIPs.  When I power up, Kickstart 2.0 loads, the machine
.>is reset, then I get a System Error #80000003 at location 07e03234 (well,
.>somewhere in the 07e03000's;  I believe 07e03234 is the correct address). Ca
n
.>anyone tell me what socket a bad ZIP would have to be in to cause an error
.>at that address?  I've tried swapping chips around, but the same error
.>always shows up at the same place.

[Stuff deleted for brevity]

.>P.S.  Kickstart/Workbench 1.3 loads fine, and Memtest V2.0 reports no
.>errors.

.Ah, so.  Check the parts your machine came with; the DRAM in the first bank.
.If these read something like "44258" rather than "44256", you have a known
.problem.  The A3000's memory controller, the RAMSEY chip, has a special
.memory mode for support of SCRAM (Static Column DRAMs) memories to support
.burst mode.  This special memory mode, however, has the unfortunate effect
.of confusing the hell out of normal page-mode memories.  If there are any
.page-mode memories in system, even just one, the static column mode must
.be disabled for the entire Fast memory system.  This detection has to be
.done in software.  The A3000's version of 1.3 doesn't bother with trying to
.determine this, it simply leaves the SCRAM modes off.  Under the current
.2.0, they are doing some kind of SCRAM check, but it's not very robust.  So
.SCRAM memories in the first bank will cause the SCRAM mode to be set, no
.matter what's in the latter Fast memory banks.  And I have heard that some
.A3000s are now going out with the faster SCRAMs in the first DIP bank.  So
.my guess is that you have SCRAMs in your first 1-meg bank.

Yup, turns out that's what it was.  I have now installed 3Mb worth of
static column mode ZIPs and they're working great.  Thanks for the help.

BTW, to all you other 3000 owners out there looking for static column ZIPs,
I bought mine at $8/ZIP (for 24) from a place in Santa Clara, California
called BJS Electronics (their phone number is 408-748-9762 (1400 Coleman
Rd., if you're local)).  Compared to the $6/ZIP I paid for used page mode
ZIPs from Kreuger, I'd say that this place is a _real_ bargain.  I'd
definitly urge anyone looking for A3000 memory to give them a call.

.-- 
.Dave Haynie Commodore-Amiga (Amiga 3000) "The Crew That Never Rests"
.   {uunet|pyramid|rutgers}!cbmvax!daveh      PLINK: hazy     BIX: hazy
.       Get that coffee outta my face, put a Margarita in its place!

-Dave Schreiber at davids@cup.portal.com             "Coffee, Darling?"
Possible future humerous quote:  "A/UX:  Unix done right!"