Ardent Tool of Capitalism

Japanese SCSI Adapters

SCSI adapter by Hiroshi Yamashita (old version)

PS/2 SCSI /A ("Tribble")

Short SCSI /A (A/16) (DIP Terminator)
Short SCSI A/16 (SIP Terminators)
2/3 Length SCSI/A (50-pin Header + Card Edge)
Short SCSI A/32 with Cache (DIP Terminator)
Short SCSI A/32 with Cache (SIP Terminators)
2/3 Length SCSI/A with Cache (50-pin Header)
Long SCSI A/32 with Cache and 2.5" HD
SCSI Investigation

Content created by Sandy. Edited and new outlines created by Major Tom.

Short SCSI A/16 (DIP Terminator)

P/N 35G2976 |
P/N 66G1080, FCC ID ANOTPS |

U8 Intel 80C32GB - 10MHz microcontroller as busmaster CPU
RN1 Resistor pack (term.) - Bourns 4120R-003 221/331 - Dual Terminator 220 Ω / 330 Ω
RT1 Polyfuse - Littelfuse 100 HG9K

J3 Pads for 6-pin header; Serial debug/prog. interface. From left to right, the pinout is CHR, ITS, BRN, TXD, RXD, GND

Short SCSI A/16 (SIP Terminators) P/N 07G3169

Short version of uncached SCSI /A with AIC6250EL and old SCSI microcode set.
Long uncached SCSI /A was equipped in PS/55 5570 (analogous to 8580 25 MHz).
I think this short card was an option for 5530-U/V but I'm not sure.

2/3 Length SCSI/A (50-pin Header) P/N 79F1140

2/3 Length uncached SCSI for 5560-W0. This one has 2 internal connectors, one is 50-pin flat ribbon cable connector and the other is 50-pin edge card connector.

Together with the cached P/N 95F4488 these are probably the only MCA SCSI adapters from IBM that have the standard 50-pin internal header. Both used in various 5560 machines plugged into one of the 2/3 length slots.

Short SCSI A/32 with Cache (DIP Terminator) P/N 35G2818 |

This is the Latest SCSI 32/A, probably an option for 5551-Rxx. Basically this adapter is capable to handle drive >1G. Unfortunately most of PS/55s (except 9595 and 8595-J) can't handle IML drive larger than 1G with this adapter due to BIOS issue. 8595-Jxx (1S1P planar with Type 3 50 MHz complex) with this adapter can handle HD over 1 GB as IML drive. SCSI ID for IML drive is not restricted to #6 for this combination.

I once fried the F1 fuse when I attached a Rancho's Single/Differential converter. Really got shocked because that was the only one "> 1 GB capable card" I had back then. I trolled around parts shops in Akihabara and got a similar type of tiny chip named "Poly switch". Electrical specification was a bit different but with the chip my card revived.

Short SCSI A/32 with Cache (SIP Terminators) P/N 07G3063 |

2/3 Length SCSI/A with Cache (50-pin Header) P/N 95F4488 | |

This adapter is an earlier version of SCSI 32/A and was equipped in 5561-W1x/W2x models. SCSI microcode set is 64F4356/57 and it can't handle drive > 1 GB as ID6.

Together with the uncached P/N 79F1140 these are probably the only MCA SCSI adapters from IBM that have the standard 50-pin internal header. Both used in various 5560 machines plugged into one of the 2/3 length slots.

Long SCSI 32/A with Cache and 2.5" HD P/N 35G3785

2.5" SCSI HD is mounted on the adapter. The drive is set to ID #6. The stock HD is WDS280 (80 MB).

All chips on the PCB are same with short SCSI 32/A with cache P/N 35G2818 with Yellow DIP Terminator. Just the parts are reallocated on the PCB.

SCSI Investigation

I'm not sure if I'm correct or totally wrong about the following info. I tested and experienced by myself thus I wrote. I can't explain my experience logically 'cause I can't read BIOS or don't have sufficient documentation about PS/55's SCSI feature.

It is told that PS/55's BIOS lacks, unfortunately, support for over 1 GB IML drive. Even if we use 92F2244/92F2245 SCSI BIOS and 10G4890 microcode, PS/55s such as 5580, 5560 and 5551-N can not accept IML drive larger than 1G capacity. Server models like 9595 (Type 4), 8595 (Type 3) and 9585 are capable to run drives > 1 GB as IML and they are basically same with pure PS/2. In another word, BIOS ROM in Japanese localized processor complexes (and planar) are not sufficient to take drives > 1 GB as it's IML drive even though we use SCSI adapters with enhanced SCSI code. We have such nice cards over here but it's true. It's ridiculous and sad.

Group A Adapters – not capable to handle > 1 GB IML and/or boot drive

Group B Adapters – capable to handle > 1 GB IML and/or boot drive

* Not capable to handle partition > 1 GB as a boot drive. Ed. Huh? But the title says "capable".

Flat package. Don't know correct term for the ROM. It's a flat, SOJ type ROM (write-one-time).

Example:

5580-Y with Type A Upgrade complexes (DX2-66), HD; Seagate ST42400N. Can't make system partition on the Seagate even with an adapter listed under Group B. The Seagate drive IDed as 6 off from 35G2818 can't be recognized as valid drive for IML.

If I use a smaller drive with ID=6 and make system partition on it, then change boot order to Seagate 2G in one partition as C, then system works. Here, if I change SCSI /A to listed under Group A, system stops upon booting with error message "Missing Operating System". If I change older type of SCSI /A to newer one again such as 35G2818 or 07G3063, every thing works fine.

If I change Type A Upgrade to Type 3, the Seagate drive off from 35G2818 and 6451280 (both in Group B) works fine as like as usual PS/2 with Type 3 platform. Also I can give SCSI ID other than 6 to ST42400N.

Conclusion (If I'm correct):

1. Pure PS/55s dose not contain SCSI code in it's BIOS or lacks some important codes relating SCSI. Server models (5580 and 5560) may handle drives larger than 1 GB if BIOS ROM could be replaced. IBM Japan had not ever released such BIOS ROMs nor reference versions. I'm not a BIOS tweaker.
2. SCSI adapters listed in group A can not handle drive larger than 1 GB for IML purpose. Also it can't boot OS even IML drive is smaller than 1 GB and boot order is changed to another drive larger than 1 GB in which OS is installed.
3. SCSI Adapters listed in group B can handle drive larger than 1 GB for IML purpose only when it is used with PS/2 complexes including PS/55 Type 3. In this case SCSI ID for IML drive is not limited to ID 6. With Japanese Type 1/2 and dx2-66 upgrades, SCSI adapters in group B on 5580/5560 can't do it.
4. As a matter of fact, 5551-N can not handle drives larger than 1 GB as an IML drive even though it has 92F2246/2247 microcode set on the planar.

Why Big Blue Japan deleted SCSI code from machine BIOS?