Type 3 Complex
Rf90953a.exe Reference disk for Type 3 Complexes
Rd9095a.exe Common Diagnostics for all 859x / 959x Systems
52G9619 1993 BIOS Original Mitsubishi M5M27C201 PLCC32
192-101 IBM PS/2 Enhanced 486/50 Processor Upgrade

Specifications 
    Memory supported, cache, features
    Support for >1GB Disks as IML Drive 
    False ARTIC Diag Error
Type 3 Evergreen cpu upgrade that worked
"M"/ Upgrade 486DX 50 MHz   57F1597
    Daughtercard
      Why DX50 is on a Daughtercard
       Downlevel Complex Incompatible with some Streamers 
      Daughtercard Rework
         Rework on Right
         Rework on Left
    Base Board
    Diagnostic LEDs
  Enhanced Dual Path Memory Parts
U27 Burning 


"M"/ Upgrade 486DX 50 MHz   57F1597 FCC ID ANOIBM486B50 Sticker is 71G6207.

DX50 Base Board

CR1, CR2 Diagnostic LEDs
J1   Solder pads for 25 pin header
J2 Socket for Daughtercard
OS1 40.0000 MHz MCA Bus clock
OS2 22.1184 MHz Osc- Dunno.
U1,22 Voltage Regulators
U14 96F1920 Memory Data Buffer
U15 96F1921 Memory Address Buffer
U23 87X0079
U25 10G7459ESD ???
U33 92F1477ESD ???
U27 41G9251 BIOS ('92)
        52G9619 BIOS ('93)
U41 10G7808 DMA Controller (Type 2)

U27 BIOS. Mitsubishi variant of the 27C201, (256K x 8-bit.)
87X0079 What is this? It's on all later complexes.

J1  Earlier versions had a double pin-row connector at that place ... intended for the (or: one) predecessor of the Systems Management Adapter used on "high availability network servers".

Diagnostic LEDs
   The 50-MHz Type 3 complex in a 90/95 (submodel code 28 and 29), has two LEDs; CR1 and CR2. During POST, CR1 comes on momentarily and CR2 stays off. If the LEDs work any other way, suspect that the processor board is defective. 
   Use the LEDs to help differentiate between a processor board or a system board failure. If you are instructed to replace one of the boards, and the problem still exists, replace the other board (also reinstall the first board). (Ed. Reports coming in make me wonder about any of this. If the board passes diagnostics, and works normally, do you really care about the LEDs?)

Enhanced Dual Path Memory Parts
The Type 3 has the 96F1920 Memory Data Buffer / 96F1921 Memory Address Buffer base board, 96F1992 Data Buffer / 78F1639 Address Buffer daughter card. The reason for the different chip numbers may be the use of "Packet Data Transfers" on the Adapter side I/O buffer.

* Enhanced dual path memory design (Dual Bus Interleave). Although Base 1 allows both the processor and busmasters to access memory concurrently through two paths, the Base 3 and 4 has buffers at both paths to provide better performance. Also the buffer on the adapter side (I/O buffer) uses packet data transfers for writes. This means 16 bytes are collected and this packet is written in one cycle to memory as opposed to writing for every 4 bytes received (as with unbuffered systems).


Daughtercard 71G6190
OS1 50 MHz CPU Clock
SP1 71G6205
U1 87X0079
U2 78F1639 Memory Address Buffer
U3 486DX-50 (5v)
U4,5,7-12 L2 Cache
U6 96F1992 Memory Data Buffer
U13,18 L2 Cache controller Set
U14,15,17 TI SN74BCT 2160-12FM
0

U13, U18 L2 Cache Controller
U4, U5, U7-U12 Toshiba TC55B329J-12

Why DX50 is on Daughtercard
Peter spake thusly:
   The DX50 is on the daughtercard because of the odd Intel cache chipset. The chipset has limits on the distance it can be installed away from the CPU. The only was to use the CPU and cache together was to place them on a daughtercard.

Ed. Also think of the addition of the memory address buffers, which are as big as the Memory Data Buffers. You have to shoehorn in TWO fairly big chips, along with the discrete L2 cache, PLUS the L2 cache controller set.

Versions Incompatible with Network Streamers
   The downlevel processor card is identified by the lack of a module in position SP1. This card may cause data loss or undetected change in the:
   Lanstreamer MC 16 PN 58G8999
   Lanstreamer MC 32 PN 92F8941
   Etherstreamer MC 32 PN 59G9081


Specifications

Memory
Supports 1MB and 2MB 85ns memory, 2MB, 4MB and 8MB 70ns / 80 ns memory and 4MB and 8MB ECC 70ns memory.  The memory SIMMs must be installed in matched pairs (speed, size and type).
   All 64MB of memory are addressable by 32 bit DMA. SIMMs must be installed in pairs of matched speed and size. Mixing Parity and ECC is not allowed, not even if the pairs are matched.

Type 3 Complex Capabilities
Cache: 8kb L1, 256kb L2 cache
* 40 MB per second streaming data transfer support.
   NOTE: The M has the functional equivalence of the SynchroStream controller, but in discrete chips. The N complex is the same, where it is Streaming capable, but does not have the integrated SynchroStream chip.
* Error Checking and Correcting (ECC) memory controller
* 256KB Level 2 memory cache (write-through) is standard.
* 20 MHz DMA; 32 bit DMA;  DMA supports Subsystem Control Block.
* Faster bus arbitration (than Base 1) for busmaster performance.
* Enhanced Dual Path Memory
* Subsystem Control Block enabled
* Vital Product Data support.
* Synchronous Channel Check support
* Data bus parity support
* A logging facility is provided (for ECC or system errors).

M Complex Block Diagram


This does not agree with the actual scans of M complexi, where both the base -AND- daughter card have a pair of Memory Data Buffer /  Memory Address Buffers. YMMV.


Daughtercard Rework
Dr. Jim sez: Documenting patch wire locations is great, but remember that there are often foil cuts associated with them.  And you can't always see all of those.   Ed. One can dream, can't we?

  I have seen one card with the older BIOS 41G9251 ('92) that has NO rework. The three DX50 I have with the new BIOS 52G9619 ('93) all have the rework. Why? Haven't a clue.

Daughtercard Rework on Right

Daughtercard Rework on Left



>1GB Drives as IML
   The M complex supports >1GB drives as the IML source. No complex BIOS upgrade is needed (or available). However, the SCSI controller needs the enhanced SCSI BIOS of 92F2244 / 92F2245 to IML properly.

False ARTIC diag error
   SYMPTOM: When running diagnostics on the ARTIC Portmaster Adapter/A in an 8590 or 8595 with the 486/50MHZ processor card installed, a false 14220, E0DE error may occur.
   PROBLEM ISOLATION AIDS: Check the date of the Portmaster Adapter/A diagnostic file (DEV141.DGS) on the system reference diskette. Files dated 1-31-90 or earlier will exhibit the symptom.
   FIX: Download updated Portmaster Adapter/A option diskette   Dead.


U27 Burning
  Rick Ekblaw plays deep safety:
   The M5M27C201JK is a 2 Megabit part, organized as 256K x 8, with a 12V programming voltage.  Logically, it's equivalent to other 2 Megabit EPROMs in the 27-class, such as the Intel/TI/AMD/Atmel 27C020 or the Fujitsu/NEC/SGS/ST 27C2001 (see HERE, for example).  However, the DIP form factor was still in common use in this time frame, and some of the PLCC varieties did not offer a windowed part (One-Time Programming only).

   So, the good news is that you could use almost any 2Mb PLCC32 EPROM or PROM that you can find to act as a replacement for the M5M27C201JK as the BIOS chip in your M complexi (where it is only being read by the system).  For programming the EPROM/PROM, you have to choose a part for which you have appropriate algorithms in your programmer, because they vary.


> I was all set to install a power stacker 133/586 on my dx50 in the 9595-omt. I for some reason decided to RTFM while having a cuppa. they say the powerstacker will not work with dx50. what cpu can I use to juice up the dx50.?

Jim Shorney
    Pay no attention to the manual.  Set the multiplier to x3, the cache jumper (if present) to write-through, and plug it in.  It will probably work fine.

9595 Main Page