http://bitsavers.org/components/amd/_dataBooks/1981_AMD_Am2960_Series_Dynamic_Memory_Support_Handbook.pdf

Page 76-83, physical.

DESIGN ADVANTAGES OF THE Am2965/2966
Compared with Schottky parts such as the Am74S240 or
Am74S244, which are used as RAM drivers today, the
Am2965/66 RAM drivers offer more advantages than just a RAM
driver having no external source resistor,

First, as Figure 8a shows, propagation delays for the Schottky
Am74S240 or Am74S244 are measured at 1.5V. which is not
where the RAM thresholds are. They are at o.aV. 2.4V and 2.7V
as shown in Figure 8b.

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Huh, another SWAG, one Am2966 diddles 8 address lines, the other diddles
CAS [not] and RAS [not]

But the IBM cards seem to lack the address and CAS/RAS drivers, or
mebbee the Kingston card uses a layout closer to a real 72-pin SIMM, and requires poodle-faking?

On 8/29/2021 14:43, Louis Ohland wrote:

http://bitsavers.org/components/amd/_dataBooks/1981_AMD_Am2960_Series_Dynamic_Memory_Support_Handbook.pdf

Page 76-83, physical.

DESIGN ADVANTAGES OF THE Am2965/2966
Compared with Schottky parts such as the Am74S240 or
Am74S244, which are used as RAM drivers today, the
Am2965/66 RAM drivers offer more advantages than just a RAM
driver having no external source resistor,

First, as Figure 8a shows, propagation delays for the Schottky
Am74S240 or Am74S244 are measured at 1.5V. which is not
where the RAM thresholds are. They are at o.aV. 2.4V and 2.7V
as shown in Figure 8b.

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

As that little, white alkaloid poison seeps into my body, I see possumbilities...

What if IBM had cubes that performed similar functions to the Am2966?
The Am2966 did the same as the 'S240 and 'S244

page 70 physical

The Am2965 and Am2966 are pin-compatible with the popular
'S240 and 'S244 with identical 3-state output enable controls. The
Am2965 has inverting drivers and the Am2966 has non-inverting
drivers.

Something that is NOT klar is the Am2966 COULD be used with other 2900
series chips to form an EDC memory system, but in my learned experience,
the 8580, all variants, are parity only, no ECC. So, the benefit of the
Am2966 is only integrating or eliminating external components....

"FUNCTIONAL DESCRIPTION
The Am2965 and Am2966 are designed and specified to drive the
capacitive input characteristics ofthe address and control lines of
M08 dynamic RAMs. The unique design of the lower output
driver includes a collector resistor to control undershoot on the
HIGH-to-LOW transition. The upper output driver pulls up to V cc
- 1 .15V to be compatible with M08 memory and is designed to
have a rise time symmetrical with the lower output's controlled fall
time. This allows optimization of Dynamic RAM performance.

The Am2965 and Am2966 are pin-compatible with the popular
'S240 and 'S244 with identical3-state output enable controls. The
Am2965 has inverting drivers and the Am2966 has non-inverting
drivers.

The inclusion of an internal resistor in the lower output driver
eliminates the requirement for an external series resistor, therefore
reducing package count and the board area required. The
internal resistor controls the output fall and undershoot without
slowing the output rise.

These devices are designed for use with the Am2964 Dynamic
Memory Controller where large dynamic memories with highly
capacitive input lines require additional buffering. Driving eight
address lines or four RA8 and four CA8 lines with drivers on the
same silicon chip also provides a significant performance advantage
by minimizing skew between drivers. Each device has
specified skew between drivers to improve the memory access
worst case timing over the min and max tpo difference of unspecified
devices."

On 8/29/2021 15:02, Louis Ohland wrote:

Huh, another SWAG, one Am2966 diddles 8 address lines, the other diddles
CAS [not] and RAS [not]

But the IBM cards seem to lack the address and CAS/RAS drivers, or
mebbee the Kingston card uses a layout closer to a real 72-pin SIMM, and requires poodle-faking?

On 8/29/2021 14:43, Louis Ohland wrote:

http://bitsavers.org/components/amd/_dataBooks/1981_AMD_Am2960_Series_Dynamic_Memory_Support_Handbook.pdf


Page 76-83, physical.

DESIGN ADVANTAGES OF THE Am2965/2966
Compared with Schottky parts such as the Am74S240 or
Am74S244, which are used as RAM drivers today, the
Am2965/66 RAM drivers offer more advantages than just a RAM
driver having no external source resistor,

First, as Figure 8a shows, propagation delays for the Schottky
Am74S240 or Am74S244 are measured at 1.5V. which is not
where the RAM thresholds are. They are at o.aV. 2.4V and 2.7V
as shown in Figure 8b.

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Thinking about the Kingston 4MB Model 80 daughtercard. It resembles four complete SIMMs on the card.

Did IBM ever make a double-sided daughtercard? My DIMMing memories are
of single-sided IBM daughtercards.

More wild fantasies... In the early daze, there was a problem with alpha particle that emanated by the memories' own materials. IBM religiously implemented FPM, later real ECC.

Another curiosity... Remember the old 30-pin to 72-pin SIMM Savers? 4
30-pin SIMMs on a PCB that had a 72-pin edge connector.

So... dunno. Mebbee the Am2966 are only there to "fix" issues with the
delays from a double-sided PCB? Consider the Am2966 was a dated design
by the time it was used on the Kingston 4MB card. Late 70s mebbee. '81 ?

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

On 30.8.2021 17:14, Louis Ohland wrote:

Thinking about the Kingston 4MB Model 80 daughtercard. It resembles four complete SIMMs on the card.

Well it's a memory card build from off-the-shelf parts, so of course it
does.

Should we call them "sugar free"?

Did IBM ever make a double-sided daughtercard? My DIMMing memories are
of single-sided IBM daughtercards.

I've never seen one. I'm sure that double sided sugar cube sticks(tm)
wouldn't fit next to each other. It wasn't until later when IBM switched
to the plastic package, and by then the chips had higher capacity, so
they were able to fit them all on one side of the PCB.

More wild fantasies... In the early daze, there was a problem with alpha particle that emanated by the memories' own materials. IBM religiously implemented FPM, later real ECC.

FPM? You mean parity?

Another curiosity... Remember the old 30-pin to 72-pin SIMM Savers? 4
30-pin SIMMs on a PCB that had a 72-pin edge connector.

So... dunno. Mebbee the Am2966 are only there to "fix" issues with the
delays from a double-sided PCB?

Nah, single-sided/double-sided doesn't matter here. Chip count and line
load may however, if that's what you meant.

Consider the Am2966 was a dated design
by the time it was used on the Kingston 4MB card. Late 70s mebbee. '81 ?

It's just a glue. You can compare it to the generic 74xx logic... it's
all ancient, but back then it was still relevant (and it still is to
some degree).

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

So. Think of trying to kludge a multiple SIMM design onto what was
usually a single-side PCB. Mebbee Kingston didn't wandt a card as big as
the IBM 4MB card? There is ample space for the double-sided PCB.

Also, to me, the Kingston card uses an older memory design, while the
IBM 4MB cards use a more "72-pin SIMM" style of a flat SOJ or whatever...

I wonder if the dated chipset on the Kingston card was due to a then
current lack of higher density chips?


On 8/30/2021 10:14, Louis Ohland wrote:

Thinking about the Kingston 4MB Model 80 daughtercard. It resembles four complete SIMMs on the card.

Did IBM ever make a double-sided daughtercard? My DIMMing memories are
of single-sided IBM daughtercards.

More wild fantasies... In the early daze, there was a problem with alpha particle that emanated by the memories' own materials. IBM religiously implemented FPM, later real ECC.

Another curiosity... Remember the old 30-pin to 72-pin SIMM Savers? 4
30-pin SIMMs on a PCB that had a 72-pin edge connector.

So... dunno. Mebbee the Am2966 are only there to "fix" issues with the
delays from a double-sided PCB? Consider the Am2966 was a dated design
by the time it was used on the Kingston 4MB card. Late 70s mebbee. '81 ?

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

On 30.8.2021 15:01, Louis Ohland wrote:

What if IBM had cubes that performed similar functions to the Am2966?
The Am2966 did the same as the 'S240 and 'S244

Usually the drivers and all the other support logic is directly on the
planar, and the memory modules contain only the actual memory chips.

The 8580 planars may be different in this regard. Or it could be
something simple - like the high amount of chips on these cards.

See my "Topology Thoughts" post - some of the IBM cards possibly have
dedicated drivers (or something similar) as well.

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

My curiosity is the Am2966, or some IBM flavor, or an S24x chip was
required. The design of the -Axx supported the functionality of the
Am2966 / equivalent. My suspicion is the Am2966 functionality would be supported on all models of the Model 80.

On 8/30/2021 10:27, Tomas Slavotinek wrote:

On 30.8.2021 15:01, Louis Ohland wrote:

What if IBM had cubes that performed similar functions to the Am2966?
The Am2966 did the same as the 'S240 and 'S244

Usually the drivers and all the other support logic is directly on the planar, and the memory modules contain only the actual memory chips.

The 8580 planars may be different in this regard. Or it could be
something simple - like the high amount of chips on these cards.

See my "Topology Thoughts" post - some of the IBM cards possibly have dedicated drivers (or something similar) as well.

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

So why include the Am2966 at all?

On 8/30/2021 10:41, Tomas Slavotinek wrote:

Consider the Am2966 was a dated design
by the time it was used on the Kingston 4MB card. Late 70s mebbee. '81 ?

It's just a glue. You can compare it to the generic 74xx logic... it's
all ancient, but back then it was still relevant (and it still is to
some degree).

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Like I've mentioned, possibly to avoid loading the line too much. You
are limited by the fan-out capability of whatever is on the other end of
the bus lines. Or maybe different requirements of the off-the-shelf
DRAMs (rise time or who knows what).

On 30.8.2021 18:16, Louis Ohland wrote:

So why include the Am2966 at all?

On 8/30/2021 10:41, Tomas Slavotinek wrote:

Consider the Am2966 was a dated design
by the time it was used on the Kingston 4MB card. Late 70s mebbee. '81 ?

It's just a glue. You can compare it to the generic 74xx logic... it's
all ancient, but back then it was still relevant (and it still is to
some degree).

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

On 30.8.2021 17:41, Louis Ohland wrote:

So. Think of trying to kludge a multiple SIMM design onto what was
usually a single-side PCB. Mebbee Kingston didn't wandt a card as big as
the IBM 4MB card? There is ample space for the double-sided PCB.

Also, to me, the Kingston card uses an older memory design, while the
IBM 4MB cards use a more "72-pin SIMM" style of a flat SOJ or whatever...

What Kingston used there was the standard package of the time.

I wonder if the dated chipset on the Kingston card was due to a then
current lack of higher density chips?

Possibly.

But when it comes to chip density you also have to consider limitations
of the Model 80 memory interface. It provides 4 CAS and 4 RAS lines, but
only 9 address lines. So, that will limit your chip selection... you can
play with the word size and with the chip count but you can't do much
about the address bus limitation. Many later DRAMs were 10x10, 11x11,
12x10 etc. Model 80 is limited to 9x9 max.

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Tom, if I understand, the Am2966 can control 8 address lines. Damnifiknow...

On 8/30/2021 11:28, Tomas Slavotinek wrote:

It provides 4 CAS and 4 RAS lines, but
only 9 address lines

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

And you are correct! But there are two 2966s on the board. :)

One of the chips may be partially unused, or used to drive some of the
other signals (like the strobes for example).

On 30.8.2021 18:35, Louis Ohland wrote:

Tom, if I understand, the Am2966 can control 8 address lines.
Damnifiknow...

On 8/30/2021 11:28, Tomas Slavotinek wrote:

It provides 4 CAS and 4 RAS lines, but
only 9 address lines

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Huh. Imagine the equivalent of four 2MB 80nS on a daughtercard. Just a
few small details. Like how to get past the system set-up at POST.

Huh, something like bank-switching? Have one equivalent of a 2MB 80nS
SIMM on the daughtercard, able to access on-card memory after POST.
Think of it akin to an initializer?

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Two 1MB IBM & Two Hitachi 1MB Memory Cards – Vintage Components

https://www.ebay.com/itm/234092265188

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Non-inverting Am2966 replaces 74S244;

On 8/29/2021 15:02, Louis Ohland wrote:

Huh, another SWAG, one Am2966 diddles 8 address lines, the other diddles
CAS [not] and RAS [not]

But the IBM cards seem to lack the address and CAS/RAS drivers, or
mebbee the Kingston card uses a layout closer to a real 72-pin SIMM, and requires poodle-faking?

On 8/29/2021 14:43, Louis Ohland wrote:

http://bitsavers.org/components/amd/_dataBooks/1981_AMD_Am2960_Series_Dynamic_Memory_Support_Handbook.pdf


Page 76-83, physical.

DESIGN ADVANTAGES OF THE Am2965/2966
Compared with Schottky parts such as the Am74S240 or
Am74S244, which are used as RAM drivers today, the
Am2965/66 RAM drivers offer more advantages than just a RAM
driver having no external source resistor,

First, as Figure 8a shows, propagation delays for the Schottky
Am74S240 or Am74S244 are measured at 1.5V. which is not
where the RAM thresholds are. They are at o.aV. 2.4V and 2.7V
as shown in Figure 8b.

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Looking at photos real quick, I'm pretty sure that the PCB is a 6-layer
jobbie. The two inner most layers are power and ground planes, and the
rest are routing layers.

You can see most of the signal traces - even these that are on the inner layers.

If you look at the bottom side of the board, where the planar connector
pins are visible (with pin A01 marked by the unique square pad) you will
see that the first 8 address lines (MA0-MA7, connector pins A03-A10) go
to U1. The last address line (MA8, pin A11) goes to U2, together with
the next two pins (A12 & 13) - and these are the row strobes (-RAS0 &
1). So yeah, they used the rest for the strobes and what not...

On 30.8.2021 18:59, Tomas Slavotinek wrote:

And you are correct! But there are two 2966s on the board. :)

One of the chips may be partially unused, or used to drive some of the
other signals (like the strobes for example).

On 30.8.2021 18:35, Louis Ohland wrote:

Tom, if I understand, the Am2966 can control 8 address lines.
Damnifiknow...

On 8/30/2021 11:28, Tomas Slavotinek wrote:

It provides 4 CAS and 4 RAS lines, but
only 9 address lines

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

For the double-sided question, remember that IBM wrapped a clear shield around the solder-side of their sugar-cube Model 80 boards to prevent electrical contact with the PSU or other memory board. Don't forget the "mini-modules" of the same pinout on the
Model 80 memory adapter either. I can grab them up for what appears to be an upcoming photo session of extended length.

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

On 30.8.2021 21:31, IBMMuseum wrote:

For the double-sided question, remember that IBM wrapped a clear shield around the solder-side of their sugar-cube Model 80 boards to prevent electrical contact with the PSU or other memory board. Don't forget the "mini-modules" of the same pinout on

the Model 80 memory adapter either. I can grab them up for what appears to be an upcoming photo session of extended length.

Yeah the insulating sheet is pretty much mandatory with the tall sugar
cubes...

More photos?! Yeeees!

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)