[comp.sys.next] 4MB SIMMs from Chip Merchant: 4MBx9 yes 4MBx8 no

eps@toaster.SFSU.EDU (Eric P. Scott) (02/16/91)

In our last stroll down Memory Lane, our requisition for
Peripheral Outlet memory turned into a Purchase Order to Chip
Merchant, thanks to an "alert" employee who "saved us" $4 per
SIMM.

Chip Merchant's using Samsung chips* on a generic PC board.
(There are no markings other than an inspection stamp.)
They're about 1/8" taller than the 1MB SIMMs.

   ----------  ----------  ----------  ----------  ----------
  |          ||          ||          ||          ||          |
   ----------  ----------  ----------  ----------  ----------
         ----------  ----------  ----------  ----------
        |          ||          ||          ||          |
         ----------  ----------  ----------  ----------
  U U U U U U U U U U U U U U U U U U U U U U U U U U U U U U 

*NeXT currently ships "genuine" Samsung 1MBx8 SIMMs in
 8MB NeXTstations.  Traditional cubes came with Toshiba SIMMs.

The 4MBx9 (parity) SIMMs work fine in the NeXTstation.
The 4MBx8 (non-parity) SIMMs didn't; the machine was convinced
that they were defective parity SIMMs.  Visually, they appear
to be identical to the 4MBx9 SIMMs, but with one chip missing:

   ----------  ----------  ----------  ----------  ....  ....
  |          ||          ||          ||          |    o  o
   ----------  ----------  ----------  ----------  ....  ....
         ----------  ----------  ----------  ----------
        |          ||          ||          ||          |
         ----------  ----------  ----------  ----------
  U U U U U U U U U U U U U U U U U U U U U U U U U U U U U U 

Not knowing the specifics of how the machine determines what
kind of SIMM it has, I don't know if these are trivially
fixable (add a jumper/cut a trace?).  We're probably going to
send these back next week and give them another chance.

					-=EPS=-

anderson@sapir.cog.jhu.edu (Stephen R. Anderson) (02/17/91)

Eric-

I read this article with some interest, since I had 16 Mbytes of 4 Mx8
SIMMS just delivered from Chip Merchant sitting on my desk waiting to
be installed along with my '040 upgrade. Well, rather than give up, I
went ahead and installed them anyway.....and they work just fine. I
put the 8 1x8 SIMMs from the '030 board in, and added another bank of
these Chip Merchant 4x8's, and the machine came up with no problem
(and reports 24 Mbytes physical memory in /usr/adm/messages).

I suspect that you actually DID get defective SIMMs. The ones I put in
look just like your diagram, and I think we must have the same kind,
but as I say, mine seem to work fine.

I should mention that the extra height makes it a tight squeeze to get
the board back in the cube, but it does (just barely) go.

I bought mine from Chip Merchant partly because they were a little
cheaper, but more importantly because Peripheral Outlet (I think it
was them - anyway the place that several people have posted about
recently: my apologies to Peripheral Outlet if I'm wrong) has stopped
taking Purchase Orders. I've gotten good service from Chip Merchant
before (these were ordered on Monday, the PO was faxed to them on
Tuesday, and they arrived on Friday), and no SIMMs I've bought from
them have ever given me any problems (as opposed to some I've bought
elsewhere, for Suns). The only thing is that you have to be pretty
sure yourself what you want, since the people at Chip Merchant can be
guaranteed not to know anything more than the name of what they're
selling. They do know that they have "80 nanosecond low profile 4meg
by 8 SIMMs": YOU have to know that that's what you want. They've never
heard of a NeXT ("Is it IBM compatible?  We sell SIMMs for Macs and
IBM compatibles.")

Steve Anderson

eps@toaster.SFSU.EDU (Eric P. Scott) (02/17/91)

I've been getting mail from people who are apparently confused
about this posting.

I am *not* mixing parity and non-parity SIMMs.

4 4MBx8 SIMMs from the same manufacturing lot, with no other
memory installed, did not work.  I tried both banks.

Replacing *all* of them with 4MBx9 SIMMs *did* work, as does
using only 1MBx8 SIMMs--which was the original configuration.

What I ultimately want is 4 4MBx8 + 4 1MBx8 = 20MB non-parity.

* * * End facts; begin opinions:

There's no question that going from 8MB to 16MB yielded a
dramatic performance improvement.  The 8MB configuration was
heavily paging.  Even in non-GUI "console" mode, it was clear
that 8MB barely cut it.   With 16MB I could load two typical
Applications with no apparent degradation; with a third it began
paging to the 105MB, but no where near as badly as with 8MB RAM.
But it's obvious when you "go virtual."  Another 4MB should let
me run one or two more concurrent apps and still impress tour
groups.  :-)

Part of the "problem" is that so many of our users are used to
working on "primitive" microcomputers where it takes lots of time
to start an application, but it's then fully memory-resident,
and "as fast as the machine will let it."  With 8MB, the
NextStation mostly sucks a thick milkshake through a very thin
straw, with occasional bursts of the "15 screaming MIPS" it's
supposed to show.   Double that, and it pretty much lives up to
the hype.  The Quantum's a dog, though.

BTW, we ordered our NextStation Color machines with 16MB RAM
from the outset--we need them to be usable out of the box when
they finally get here.

					-=EPS=-