Ardent Tool of Capitalism

Type 0 Processor Complex

91f8526.zip Reference Disk Type 0 Processor Complex v1.00 (bootable) (from David Beem)
rf90386a.exe Reference Disk Type 0 Processor Complex v1.00 (non-bootable) (zipped image)
Note: No diagnostics disk needed, diagnostic files included on the refdisk.

Model 90 with Type 0 Complex

Type 0 "Minus" No L2 Cache
Type 0 "Plus" 64 KB L2 Cache
System Firmware
   ROM Images
Type 0 "Minus" - IML or not?
Type 0 and 8590 DBA ESDI Support
   ESDI Support for Convenience Partition?
Type 0 Booting Experiences
Type 0 Difficulties with 9595

Note: "Type 0", "Minus", and "Plus" are unofficial designations.

Type 0 "Minus" No L2 Cache, P/N 64F0782, FCC ID ANO386201A |

This complex uses the same core logic as the 8570 Type 2 planar, the Later P70 planar, and the 5550-S/T Stage 2 planar.

Type 0 "Plus" 64 KB L2 Cache, P/N 04G3884 or 84F9541, FCC ID ANO38620C1A |

U2 82385-20 Cache controller - also used on 8580 Type 3 planar and 8570-Axx CPU boards.

This complex is built around the same core logic as the "Type 0 "Minus" complex, but it adds a cache subsystem based on the intel 82385 controller.

Supports four 85 ns and 80 ns 2 MB SIMMs on the risers for 8 MB total and additional 8 MB on an expansion card. Supports Spock from one to three can version. Most likely the short Tribble as well.

Sold as a part of the "special bid" PS/2 Model 90 XP 386 machine — model/type 8590-402. Shipped with a single memory riser, planar with the two DBA-ESDI connectors populated, no intake fan, and no complex air baffle. Known -402 are from WorldCom, Aetna, Royal Bank.

System Firmware (POST & BIOS)

Type 0 "Minus" Complex

Complete firmware stored in EPROM. IML image optional. More information HERE.

Type 0 "Plus" Complex

Simplified firmware stored in EPROM. IML image required for the system to operate.

ROM Images

Type 0 "Minus" Complex

57F2943 - 19 Aug 1989, rev. 0, 27C020-200 (U49)

Type 0 "Plus" Complex

84F9413 - 01 Aug 1990, rev. 0, 27C010-150 (U21)

Type 0 "Minus" - IML or not? (from Major Tom)

Turns out that the humble T0- platform is more interesting than one would think... The firmware actually supports two POST modes! IML and ROM-based POST.

The general IML process looks very similar to what we know from the T1 platform. But there's one major difference - if the IML process fails, it falls back to a purely ROM-based POST method. Interestingly enough, this mode is only available on Model 90 (planar ID FF6Fh). Why? My guess is that the ROM-resident Stage 2 code is only compatible with the Model 90 planar. Model 95 compatibility could be added via either a model-specific IML image or an universal image that has support for both machines.

It makes you wonder why didn't they do the same thing with the T1 - T3 boards... The larger ROM chip wouldn't add much to the BOM cost. The bank switching logic seems to be trivial as well - looks like they are simply toggling the A17 line of the ROM via a specific port (I wonder if this works on the T1 platform as well...).

Neat, ey? Yeah, but unfortunately there's a catch... There appears to be a fatal bug in the "ROM to RAM" routine. When all other operations are done they do a final word-by-word compare of the ROM data with its shadow copy in RAM. Problem is that they used a wrong jump instruction - jz instead of jnz. This means that the check will *fail* if the copy *matches* the source. Instead of jumping to the reset vector F000:FFF0h (to continue with POST Stage 2), the function re-enables the ROM, returns, and the caller then triggers error 0121 - "Problem detected during memory remapping" (256KB ROM Error), and the system is halted.

IBM reused some of the ROM to RAM code for the flash-based T4 platform. Portions of the code are identical between the two, including the verification routine - with one notable difference - the T4 code has the jump instruction corrected of course.

The 099900xx error codes have the same meaning as the corresponding I99900xx IML codes.

From Rich Wolos:
   I had these -402 beasts. Converted them all to 486, no problems now. Just have to reload the IML on hard drive. Usually do a LLF first. The -402 will only accept 4x 2/85 (2 MB 72-pin SIMMs FRU# 92F0104) = 8 MB on motherboard, 8 MB more on an expansion card. Bootable refdisk for that 90386 is only available by backing up. Keep a copy before you switch, may be useful someday <g>.

Type 0 and 8590 Planar DBA ESDI Support

Update from Daniel Hamilton: Confessions of a Model 90 386 (update)

I took down my Model 90s today and booted them up. I started up the old 402 and got another itch to try to do something with those ESDI ports. I experimented back in early 2006 with this (I went and found my old posts here before trying again, to make sure I wasn't repeating mistakes of the past). :) However, this time I found out something that partially disproves something I had revealed here back then. My original post read:

8590-402 reveals more secrets...

Q: Is ESDI bootable on a Type 0 Processor Complex?
A: No: IML Error I9990080 (IML not supported on primary disk drive)

The answer is not incorrect, but it is *incomplete*. The correct answer is Yes AND No actually. ESDI will boot with the T0 I found today. The catch is that it just cannot be the IML drive. What I did before was remove the Spock adapter and therefore removed the original 160MB SCSI disk that came with the system (which was the IML disk as well). However if the SCSI adapter/disk is NOT removed, the ESDI drive works just fine, and actually will preempt the SCSI disk in the boot sequence by default!

I can see how this would have helped upgrades from Model 70s (which I've read postulated elsewhere is a major reason for the Model 90s). A administrator could simply remove the hard disk from a production Model 70 and just stick it in a new Model 90 and turn it on... presto, the system boots up with the ESDI disk as C:, and the fast SCSI disk sets waiting as D: drive. Neat.

So a correction 4 years in the making is...

Q: Is ESDI bootable on a Type 0 Processor Complex?
A: Yes: ESDI disk will boot as long as another IML-capable disk is installed on the system. IML Error I9990080 (IML not supported on primary disk drive) will occur if only the ESDI disk is installed.

ESDI Support for Convenience Partition?

Tim Clarke opines:
   So, in theory, you could IML from the RefDisk and then boot to an ESDI-only system via the "Start Operating System" option from it's menu, I'd guess. Also, have you considered that the Type 0 complex might have "Convenience Partition" support in its ROM(s), if IBM were experimenting with the IML support for non-SCSI drives at that time? If you ensure that there are no more than 3 partition entries in the master partition table and a good 4MB of space available on the ESDI drive, you could then try the "Create System Partition" option on an ESDI-only setup? The Convenience Partition's type byte is 0x3F, IIRC.

Tom says:
   The T0 and early T1 POST code definitely have support for IML from a DBA drive. But I'm not sure if the system programs are capable of installing a reference partition there.

Type 0 Booting Experiences

I got it to work in another 8590-0H9. The weird thing (and telling) was this...

Two machines:
   8590-402: Stock T0 complex
   8590-0H9: Stock T2 complex

I have a second T0 complex (I had before getting the 402), here-after referred to as T0-2. T0-2 would not work initially if swapped with the original T0 in the 402. No video. I then flipped the password-reset jumper on the planar. Put T0-2 back in, and it worked. Great, now I know the complex works.

I just got the 0H9. It came with a T2 complex. I checked the planar, same FRU as the one in the 402. So I swapped the memory for some 80 ns parity memory (2x 2 MB). Went ahead and flipped the password-reset jumper. Powered up, no video, nothing. I was stumped. So then I took the T0 in the 402 and put it in the 0H9. Viola, it worked. Okay, what's going on? I put T0-2 back in AND IT ALSO WORKED, although it wouldn't just minutes earlier. So I had two functional T0 machines functioning. I seemed to finally have the recipe for success.

The recipe is: One 8590 with Spock SCSI. Put at least 1 PAIR of 80 ns Parity SIMMs in (Model 70 386 Memory). Flip the password-reset jumper. Cross your fingers and push the power switch.

Type 0 Difficulties in 9595

Daniel Hamilton's story with 9595s:
   On the Model 95 XP (0MG), the T0 seems to not be able to access IML via the floppy controller. You'll always get I99900x1 or I99900x2 errors. I took the 160 MB SCSI drive from the 402 and put it in the 9595. It then came back with a I999003x error, Disk IML record incompatible with system board. Someone mentioned the planar ADF is missing and could be causing the problem. I need to investigate this further. I finally ended up with I9990033 errors.