8580 Common Devices

SHS15F2199 IBM PS/2 Model 80 HMS
SHS64F3995 IBM PS/2 Model 80 HMR
SHS71G4102 IBM PS/2 Models 50, 50Z, 55, 70/80 Upgrades HMM
8580 RETAIN TIPS (*.BOO)

Type 1 Planar (16 MHz, -041 & 071 Models)
Type 2 Planar (20 MHz, -081, 1xx & 3xx Models)
Type 3 Planar (25 MHz, -Axx Models)

Planar Memory (exclusive to Model 80)
8560, 8565 SX and 8580 - Common Devices
Power Supply
8580 Processor Upgrade Options
8580 Diskette Drives (and how to fix them)

8580 Models
Dirty Floppy Drive Controller
Clear CMOS by Shorting Modules to Ground
161/163 Error Code Combination Trivia
Memory Errors
Memory Activation / Deactivation
Model 80 ECAs
Intermittent parity error 110 on 8580 during POST, diags, and apps
>16MB on a 8580 under Linux
64MB Limit on 8570 / 8580 Configuration
8580 Does Not Support SCSI-2 F/W (Corvette)
Base Frame Differences


8580 Models (by Fred Spencer)

Model
Type
CPU Hard-Drive # of MCA Slots
Type Speed Type Size Total 32-bit 16-bit AVE
8580-041 386DX 16 MHz ST-506 44MB 8 3 5 1
8580-071 ESDI 70MB
8580-081 386DX 20 MHz SCSI 80MB 8 3 5 1
8580-111 ESDI 115MB
8580-121 SCSI 120MB
8580-161 SCSI 160MB
8580-311 ESDI 314MB
8580-321 SCSI 320MB
8580-A16 386DX 25 MHz SCSI 160MB 8 4 4 2
8580-A21 SCSI 120MB
8580-A31 SCSI 320MB

For an IBM summary and a list of various IBM Options previously available, jump to Model 80 - Technical specifications.


Dirty Floppy Drive Controller

>There was some dust on the FDD controller below the fan, but I've seen PCBs in even worse condition. This machine seems to be on duty not for long time. I also cleaned the drive (Mitsubishi type), but all didn't help.

Peter said:
   There were a lot dust-related errors on Model 80 - therefore I've explicitly mentioned it on my "Known Errors" page. Office and household dust is often conductive and also causes a lot config lost and date/time errors. The Model 80 PSU fan is over a very sensitive part of the planar.

Time for a Q-Tip and alcohol


Clear CMOS by Shorting Modules to Ground (Record number: H104756)

8580 systems use CMOS to store system configuration. CMOS may become contaminated for various reasons. The current field procedure is to remove the battery and allow CMOS to drain until the configuration is no longer present, which may take several hours.

A faster way to drain CMOS is to short the CMOS modules to ground.

Disconnect The Battery From The 8580 System!

Locate the CMOS modules, identified below and by using a meter lead connected to frame ground, such as the power supply case, carefully drag the meter lead around each of the respective modules, contacting each of the pins. This will immediately drain CMOS. Connect the battery.

CMOS modules are located in the following positions for the following systems:

  • 8580-041,071 (16MHz) in position U186 & U187
  • 8580-081,111,121,161,311,321 (20MHz) in position U146 & U186
  • 8580-A16,A21,A31 (25MHz) in position U132 & U152 (Near Blue Conn.)

If problems are still encountered after draining CMOS, normal problem determination should be performed to determine the cause of defect.


161/163 Error Code Combination Trivia

Peter "remembered" this:
   If the floppy *seems to start* booting (LED lights up, FDD whirs for a second) after the beeps and 161 / 163 display but then "hangs there" press [F1] to force loading. The older Model 80 seem to have a *slightly* different BIOS that not always auto-loads even a valid reference disk. After 161/163 errors the first loading of the reference disk *may* take up to 90 seconds - because the machine BIOS uses the lowest transfer rate available, because it is uncertain about the type of FDD.

If the F1 fails and a second F1 brings you into IBM ROM BASIC suspect:

  • the floppy *not* being a valid reference disk (extraction fault?)
  • accidentally snatched a defective floppy for the reference disk creation?
  • having a dirty / misaligned / dysfunctional FDD


Memory Errors

00011000 indicates a system-board parity-check error. If two system-board memory-expansion kits are installed, remove the kit in the connector labeled either J15 or J8. If the problem disappears, replace the kit. If the problem remains, replace the kit in the connector labeled either J16 or J9. If that does not solve the problem, replace the system board.

00011100 indicates a memory-adapter parity-check error. Remove the memory adapters from the expansion slots, one at a time, until the error goes away. Remove the memory-expansion kits from the last adapter removed. Install the memory-expansion kits, one at a time until the error returns. Replace the last memory-expansion kit installed. If that does not solve the problem, replace the adapter.

00021500 indicates a failing system-board memory-expansion kit in the connector labeled either J15 or J8. Replace the kit. If that does not solve the problem, replace the system board.

00021600 or 00022100 indicates a failing system-board memory-expansion kit in the connector labeled either J16 or J9. Replace the kit. If that does not solve the problem, replace the system board.


Memory Activation / Deactivation

The Model 80 activates and deactivates memory in 1MB blocks, except for the first 1MB of system-board memory. For this block, the following occurs.

If the POST detects an error in the first 512KB of system-board memory, the first 1MB block of system-board memory is deactivated and the following occurs:

  • If an additional 1MB block of system-board memory is installed, the addresses assigned to the deactivated block are reassigned to the second block of system-board memory. This is only true for the first two 1MB blocks of system-board memory. After the first two blocks, an error code is displayed.
  • If additional system-board memory is not installed, no address reassignment occurs and an error code is displayed.
  • The first 1MB of memory address space cannot be assigned to adapter memory.

After the first 1MB of memory has been tested, it is assigned addresses. If the POST detects a memory error in any memory after the first 1MB, the 1MB block of memory is not deactivated and an error code is displayed. In this event, the customer-level diagnostics program can be run to deactivate the 1MB block and reassign the addresses. This program is on the Reference Diskette supplied with the system.

After the addresses are reassigned, the defective block is ignored by POST during subsequent power-ons.

If errors occur one at a time, the system deactivates 1MB blocks of memory. However, if two errors occur at the same time on the same 80386 Memory Expansion Kit, the programs on the Reference Diskette cannot be loaded and an error message is displayed.


Model 80 ECAs

Peter said:
   If you don't try to run the Model 80 with more than 16MB RAM (which is a story of its own) and if the busmaster is designed within the limits given by IBM (which is the case for the SCSI's and the XGAs) you won't get into many troubles anyway.
   But there have been busmastering cards known as troublesome - because they tried something "off standard" - like the Buslogic BT-64x cards, which were explicitly named in the original ECA text along with other cards.
   The ECAs had been listed in the earlier versions of the EPRM - but for the last ones they had been thrown out. I'll try to dig out an older EPRM version and see if I can get the original text.

Ah, found it:

ECA008 is the "speaker noise" ECA (valid for 8580-111 S/N below 6019000 and -311 S/N below 6502022) with Planar FRU 90X7390 *and* a 387 FPU installed.

ECA011 is the 386/387 conflict ECA, valid for -111 / -311 with board FRU 33F8415 with EC-level a79694 and / or REA number 18600149 and lower (instead of the EC number).

ECA031 is the "busmaster ECA" but a different text that I recall. This refers also on the "initial system board" FRU 90X7390, which has an EC level below EC C00835 *in either case* and the reference is made for the "Wizard" adapter but the IBM SCSI adapters are also mentioned. So there is a general problem with the busmastering on these early level boards at all. I think they won't work properly at all. The most common failure mentioned is a 110 Parity Error with a busmaster.

ECA035 is "8514/A in 8580" but this references downlevel 8514 cards rather than the Model 80 at all.

ECA040 is the "314MB drive not coming ready" ECA - valid for -311s equipped with a second 314MB ESDI drive - caused by a failure in the HDD logic board.

ECA048 is the second "busmaster ECA", which mentions both 90X7390 *and* 33F8415 boards as potentially affected. They list some cards: Known troublesome are Bicc Isolan, ComTech Channel/2 and /2+, IBM Wizard / Portmaster / XGA (!!), Metacomp Psconnect, Northern Telecom LanStar/MC, Proteon P1840 Pronet-4, Racore 4x16 Network card, Yarc Micro785+. The Lan Technologies Microprint Network is listed as "under review", but that is *still* not the full ECA I have in mind. Maybe it is one of the earlier drafts of the final ECA.

ECA051 is an "OS/2 trap / NMI" but for 25MHz boards.

ECA069 is "162 + 601 POST error" ECA - caused by downlevel FDDs on all early PS/2 (50, 60, 70, 80).

That's what I found so far.


Intermittent parity error 110 on 8580 during POST, diags, and apps (Record number: H006554)

When advanced diagnostics fail to detest a bad system board memory card, swap the cards in J15 and J16 re-run diagnostics.

Advanced diagnostics loads into the first 256K of memory and will test all memory above 256k with multiple bit patterns that stress test the memory cards.Stress testing the first 256K is not possible because the test would write over the diagnostic program. You can test all the memory by swapping the two memory cards and re-running the memory test.


>16MB on a 8580 under Linux

Ed Avis says:
   I upgraded my 8580 from 12MB to 24MB RAM and Linux's ibmmca.c SCSI driver stopped working. (I have a couple of Spock Primes in there.)

To cut a long story short, this is caused by a DMA limitation on the 8580 (and possibly other early PS/2s) that restricts DMA to the lower 16Mbyte of RAM. Michael Lang, the maintainer of the driver, was very helpful and after trying various bug-fixes, suggested changing the flag 'unchecked_isa_dma' to 1. I'm not sure what this does, but it fixed the problem.

So if you have >16Mbyte RAM and you find that Linux crashes on boot with errors from IBM MCA SCSI, edit the file ibmmca.h in the kernel sources and change a couple of lines to read:

    unchecked_isa_dma: 1,                 /*32-Bit Busmaster */

instead of 'unchecked_isa_dma: 0'. Hopefully this will become an option in later versions of the kernel - or maybe with the new 2.4 kernel and 4.0 ibmmca driver it is not needed.


64MB Limit on 8570 / 8580 Configuration

The 8580 SC.EXE only supports 64 MB. If you have somehow stuffed in two MC-64 cards, the system will barf.


8580 Does Not Support SCSI-2 F/W (Corvette) (Record number: H125078)

Compatibility:
   The IBM SCSI-2 Fast/Wide SCSI Adapter/A is supported in the 8590/95 & 9576/77/85/90/95. BIOS on earlier 32-bit systems (8570 & 8580) do not support this Adapter.

Note: Further research by WBST also points to a change in ADF syntax between the older SCSI w/cache and Fast/Wide SCSI adapters.


Base Frame Differences (Record number: H021673)

The two Base Frame Assemblies used for the 8580 are not interchangeable. The Base Frame Assemblies contain the thumb screws to attach adapter cards.

FRU P/N64F0157 is used for Models 8580-041, -071, -081, -111, -121, -161, -311, -321.
It contains a metal back plane required for EMC purposes and should not be removed.

FRU P/N64F0158 is used for Models 8580-A16, -A21, -A31.
It DOES NOT contain a back plane assembly because there are components on the back of the planar.

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: 02 Nov 2024 - Changelog | About | Legal & Contact