Testing SIMMs for IBM SCSI Caching Adapter

Extending Cache Memory (from Alfred)

SIMMs for SCSI Caching Adapter
8550 as a Source?

Content by Carlyle Smith CSEZ Enterprises Wilmington, DE. Edited by Major Tom.


SIMMs for SCSI Caching Adapter

Here is a way to test your 30-pin 1 MB SIMMs to see if they will work in your IBM MCA caching SCSI adapter. I found some in my junk SIMM bag that are now working fine in my 95 XP machine, The key is to find 1 MB SIMMs that have pins 24 and 26 shorted to ground (pin 9 or 22). These will work in the adapter.

This newsgroup previously reported that the adapter cache could be upgraded from 512 KB to 2 MB by replacing the 256 KB SIMMs with 1 MB matched SIMMs. However, it was said that only IBM SIMMs of the type used in the PS/2 Models 25 and 30 would work; that industry standard 30-pin 1 MB SIMMs would not work. Not wanting to pay $20 or more for a pair of IBM branded modules, I thought I would check out if I had any 30-pin SIMMs that would satisfy the POST cache testing procedure of PS/2 Models 90 and 95.

Table I below shows the comparison of standard pin assignments as compared to the IBM module design. It is clear that if one has an FET VOM (low probe voltage, less than 5 VDC), you can test SIMMs you might have laying around to see if they could be used, instead of paying large dollars for the almost non-existent SIMMs recycled from early PS/2s.

**With care, you don't even need a Field Effect Transistor VOM, nor a digital VOM, only an old-style direct battery VOM, as long as the probe voltage is 5 VDC or less.

Table II identifies the SIMMs that I tested.
**Note that those parts indexed 1, 2, 5, and 7 all work.
**Note that part 4 works but only at 256 KB per module, not the 1 MB module rating, as the 1 MB presence detect requirement is not satisfied
**Note that the part index number 2 below is the original SIMM found on the caching controller

Table III shows the results of testing the SIMMs in the SCSI adapter in a Model 95. It is clear that if you probe pin 24 of a module, and find it connected to ground (pin 22), the module may work in the SCSI adapter. If pins 22 and 24 are not connected, it surely won't. If it is a 1 MB SIMM and pins 24 and 26 are both grounded to pin 22, it will work in the cache slot with no problem.

**At about a buck a module, you're not risking much to test the SIMM. Again, if there is no contact between pin 22 (ground in *all* 256 KB and 1 MB 30-pin SIMMs), that is, if the resistance is infinite, don't go any farther. The SIMM is *not* compatible with the IBM caching SCSI controller.

I also found that I could take one of the Toshiba 1 MB modules and one of the OKI 1 MB modules, and the machine booted up with no problems. This probably would not be a good idea for the long haul, but might be useful to get the adapter working in a pinch.

Table I - SIMM Pin Assignment Differences

Pin Number Industry Standard IBM Design
19 A10 RAS1
24 No conn. Pres. Det. IBM?
26 Data bit 8 Out Pres. Det.
Open = 256K Grounded = 1 MB
28 CAS Parity No Conn.
A10 = Address bit 10
RAS1= Row Address Strobe 1
CAS = Column Address Strobe
In both types:
   Pins 9 and 22 are Grounded
   Pins 1 and 30 are +5 VDC

Table II - Identification of SIMM Modules Tested

Idx Chip
Cnt.
Chip Legend Substrate Legend
1 9 TIH8909L7SN TMS 4256 ?ML TM4256 OU9
122L8909
2 2
1
OKI M514256-80J
OKI M51256-80
MSC2324-12YS3A
221007
3 9 AAA 2801J-07
NMBS
NMBS 18184268
TCC-T-H7N
MM2801J95-07
4 3 QC ST513300J-06
33C160
1MG STB-C
5 9 TC511000AJ-10
Toshiba
THM91010 AS-10
89303
6 9 HYB511000AJ-70
Siemens
7 9 OKI M511000-10
91201B
NSC2314-12YS9A 923009
8 9 AAA 1M 300J-07
NMB5 1116
9 2
1
HYB514400BJ-60 HYB511000BJ-60 HEA59100

Table III - Module Test Results Matrix

Idx Pin 24 Pin 26 Pin 28 Works? Reported Cache Size
1 Ground Open Open Yes 512 KB
2 Ground Open Open Yes 512 KB
3 Open N/T N/T No
4 Ground Open Open Yes 512 KB
5 Ground Ground Open Yes 2 MB
6 Ground 1 Kohm Open No
7 Ground Ground Open Yes 2 MB
8 Open N/T N/T No
9 Open N/T N/T No
N/T - Not Tested

Parts 5 and 7 above have been working fine in both FCC ID ANO6451018 and FCC ID ANOSPRIME FRU 85F0063 caching controllers for some time. Actually, I did the testing back in September, and reconfirmed the operation in mid-January.

Perhaps you could check out some generic SIMMs and find them quite workable/useful.

The ones who will benefit are those who are operating on a limited budget (otherwise known as PS/2 addicts) and who need to spend their bucks in more important areas. (Ed. Like Pizza and Jolt Cola...)


8550 as a Source?

I think I might have stumbled on to another possible source for IBM SIMMs for these cache slots - the Model 50 (8550).

I had to repair one for a guy, and found that it had two 512 KB 30-pin SIMMs in it. The chips are identical to those found on one version of the 256 KB SIMM normally used on the adapter, except both sides of the module are populated, each with nine 256 Kbit chips. Of course, this precludes using them in the cache slots, because the pins on the backside interfere with leaning them back into the retainer clips.

However, looking into the specs for the 8550, the max memory on the planar was 2 MB - that is, two 1 MB SIMMs in place of the 512 KB SIMMs. So check your old Model 50s for some IBM-specific modules. Forget the 50Z - it used one 72-pin module, not two 30-pin modules.

Checking the 256 KB, the 512 KB, and the 1 MB IBM SIMMs, the Presence Detect pin grounding table looks like this:

Size 24 26
256 KB short open
512 KB open short
1 MB short short

Where the continuity between pins 24 and 26 versus pin 22 are measured. This fills the logic out nicely. Did anybody other than IBM make 512 KB SIMMs?

Content created and/or collected by:
Louis F. Ohland, Peter H. Wendt, David L. Beem, William R. Walsh, Tatsuo Sunagawa, Tomáš Slavotínek, Jim Shorney, Tim N. Clarke, Kevin Bowling, and many others.

Ardent Tool of Capitalism is maintained by Tomáš Slavotínek.
Last update: 09 Dec 2024 - Changelog | About | Legal & Contact