Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!uwm.edu!psuvax1!rutgers!cbmvax!daveh
From: daveh@cbmvax.commodore.com (Dave Haynie)
Newsgroups: comp.sys.amiga.hardware
Subject: Re: One more Amiga RAM question. ;*)
Message-ID: <9894@cbmvax.commodore.com>
Date: 28 Feb 90 21:06:15 GMT
References: <928@orange9.qtp.ufl.edu>
Reply-To: daveh@cbmvax.cbm.commodore.com (Dave Haynie)
Distribution: na
Organization: Commodore, West Chester, PA
Lines: 89

In article <928@orange9.qtp.ufl.edu> sutherla@qtp.ufl.edu (scott sutherland) writes:

>	How can the A2091, which uses 16-bit RAM, and the A2630, which
>uses 32-bit RAM, use the same type of chips, with the only difference
>being the SPEED??????????  

The parts are both 256K x 4 bits.  For the 2091, there are 4 chips in a
bank -> 4 x 4 = 16.  For the 2630, there are 8 chips in a bank -> 4 x 8 = 32.

>If this is true, do I use the same number of the faster chips for 2 Megs 
>on the A2630?? 

You need 16 chips for 2 megs on the A2630.  These are in ZIP packages, vs. 
the DIP packages used on the A2091.

>	The specs on the A2630 state that it comes with 2 Megs of 
>100 ns RAM.  According to the Amiga World review (March 1990), they
>should have used 80 ns or 70 ns RAM, like the Hurricane and other
>'030 boards.  

AmigaWorld, like many such magazines, tries but doesn't really understand
the issues.  Putting 80ns DRAM in the A2630 isn't going to help at 25MHz.
The timing isn't a continuous thing, it's quantized on 40ns boundaries.
You'd undoubtedly loose a wait state going to 70ns parts, but 70ns parts
are much more expensive than 80ns or 100ns, currently.

>My question is this.  If I leave the 2 Megs of 100 ns RAM in there and 
>decide to add two more Megs of 32-bit RAM, DO I HAVE TO USE 100 ns RAM, OR 
>CAN I USE 70 OR 80 ns RAM TO IMPROVE PERFORMANCE????   What will be the 
>effect, if any, (especially adverse effects) of having 2 different speeds of 
>RAM on the board?  Would I be better off removing the original 2 Megs and 
>using the same chips for all 4 Megs?

You could drop in 80ns or 70ns parts and the board will work, but it's not
going to run any faster, even if you replace the 100ns parts.  DRAM don't
set the speed of a system, just the potential.  The memory control logic is
what desides how fast the DRAM will be run.  There's nothing more you can 
do with 80ns at 25MHz that I'm not already doing with the 100ns parts, though
I suppose if all parts were changed to 70ns, the system could easily be
re-tuned to run faster.  Easily, meaning via a changed PAL; this isn't 
something the average person could do in a couple of minutes.

>	A couple of extra questions.  I have been told that I MUST use CMOS
>chips and not NMOS chips.  None of the places in the Computer Shopper
>specify whether the chips they list are CMOS or NMOS.  How do I differentiate?

It doesn't matter for the A2630, but in some systems it may.  Over the life of
the 256k x 4 parts, most vendors switched from the NMOS processes used in the
older 256k x 1 or 64k x 4 parts to CMOS.  As a result, most of the CMOS parts
are a bit faster here or there, though there's no way to tell for sure other
than by checking the data sheet.  Based on Commodore's habit of buying any
DRAM they can get in the Far East, plus the time at with the A2630 was
designed (finished it a year ago, right in the midst of the DRAM price
crunch), the A2630 can use any 100ns NMOS part.  You have to realize that
while the speed rating of a DRAM is how you buy them, that's only one of
ten or more very critical timing parameters associated with the particular
chip.

>And finally, Dave Haynie told me that the chips for the A2630 were NOT only 
>256K x 4 CMOS DRAM 100 ns chips, BUT that they must be in the ZIP package.

The ZIP package is the physical parameter.  The board is set up for this,
the DIP parts such as those used on the A2091 simply won't fit.

>What in the world is a ZIP package???

Start with a normal IC, if you know what these look like.  Most DRAMs and TTL
are in DIP packages; the package is relatively flat, and there are rows of 
pins going down both sides.  Imagine you can modify this package.  Take one
row of pins and bend it straight out, away from the body of the part.  Now,
take the other row of pins, bend them out, and then rotate them around to the
same side as your first row.  Move them slightly down, so instead of being
directly in-line with the first set, they're in sort of a zig-zag arrangement.
What you have now is essentially a ZIP package.  They're designed for higher
density applications; you can fit two ZIPs in the space of one DIP, sometimes
even more depending on the width of the DIP.

In any case, it's a package.  There are DIPs, ZIPs, and surface mount packages
for most DRAM.  Some companies take a mess of surface mount packages, mount
them on a small circuit board according to an industry standard, and produce
a module called a SIMM.  That's pretty much the physical packaging story.

>Scott Sutherland


-- 
Dave Haynie Commodore-Amiga (Systems Engineering) "The Crew That Never Rests"
   {uunet|pyramid|rutgers}!cbmvax!daveh      PLINK: hazy     BIX: hazy
                    Too much of everything is just enough