Hi guys, it's nice to see you're all still in down here! I've been away for a long time, just until yesterday when MCA fever hit me again and I acquired a Bull DPX/20 200 workstation, which seems to be an OEM manufactured IBM 7006 (-41x?). At least the case and mainboard are the same.
Unfortunately the box is pretty stripped: no HDD, no CD-ROM, no drive cages, but a floppy drive, a POWER Gt4xi graphics adapter with the evil 3W3 connector and... no RAM!
The seller gave me 8 pieces of Kingston KTM-9595/32 which he believes to come from "some RS/6000", but the 7006 keeps yelling some

c01 100 200 300 411
c02 100 200 300 412
c03 100 200 300 413
c04 100 200 300 414

codes at me, which I translate to "bad RAM in slots A, B, C, D" using the 7006 operator guide.

Looking that Kingston RAM up shows it is at least used for the 9595 server. No reference to the RS/6000 though. Some refer to it as parity, others as ECC.

Can anyone tell if this is the right RAM for that system?
If not, which one do I need? FRUs would be really helpful.

Also, if anyone is sitting on a stash of bored drive cages, I could help him out...

Eventually I'd like to restore this beautiful beast to an operational condition. It had been asking so sadly...

Thanks, peace and out,
Christian

On Monday, 1 November 2021 at 19:32:41 UTC, Christian Holzapfel wrote:
> Hi guys, it's nice to see you're all still in down here! I've been away for a long time, just until yesterday when MCA fever hit me again and I acquired a Bull DPX/20 200 workstation, which seems to be an OEM manufactured IBM 7006 (-41x?). At least the case and mainboard are the same.
> Unfortunately the box is pretty stripped: no HDD, no CD-ROM, no drive cages, but a floppy drive, a POWER Gt4xi graphics adapter with the evil 3W3 connector and... no RAM!
> The seller gave me 8 pieces of Kingston KTM-9595/32 which he believes to come from "some RS/6000", but the 7006 keeps yelling some
>
> c01 100 200 300 411
> c02 100 200 300 412
> c03 100 200 300 413
> c04 100 200 300 414
>
> codes at me, which I translate to "bad RAM in slots A, B, C, D" using the 7006 operator guide.
>
> Looking that Kingston RAM up shows it is at least used for the 9595 server. No reference to the RS/6000 though. Some refer to it as parity, others as ECC.
>
> Can anyone tell if this is the right RAM for that system?
> If not, which one do I need? FRUs would be really helpful.
>
> Also, if anyone is sitting on a stash of bored drive cages, I could help him out...
>
> Eventually I'd like to restore this beautiful beast to an operational condition. It had been asking so sadly...
>
> Thanks, peace and out,
> Christian

PS/2s which support ECC SIMMs require 39-bit ECC SIMMs. RS/6000s require 40-bit ECC SIMMs. Kingston KTM-9595/xx are 39-bit and only suitable for PS/2s, but useful for >64MB capacities.

Check Appendix C of this:
<http://ps-2.kev009.com/rs6000/manuals/Diagnostic_Information_for_Micro_Channel_Bus_Systems.PDF>

Argh, so it's close, but not right.
Thank you for that! I'll get in touch with the seller and have him check on the remaining memory.

So little remaining memories, so many questions...

I assume, it *could* run my ancient copy of AIX 4.3.3, would happily interact with my PC style Model M, and any SCSI CD-ROM drive would be just fine for the start, right?

And with this video thing... The card is a 11H4913, so it is a "POWER Gt4xi (1-5) 24-bit Graphics Card", according to some awesome know-how site.
Is its companion, the "Processor Card FRU 52G1343 (Placed in Lower Numerical Slot than Graphics Card)" actually required for the operation?
There is an empty MCA slot right next to it, so I assume the processor card missing.
If I recall it right, it's not as simple as dropping in an XGA-2, hmm?

On 11/1/21 13:53, Christian Holzapfel wrote:
> So little remaining memories, so many questions...
>
> I assume, it *could* run my ancient copy of AIX 4.3.3, would happily interact with my PC style Model M, and any SCSI CD-ROM drive would be just fine for the start, right?
>
> And with this video thing... The card is a 11H4913, so it is a "POWER Gt4xi (1-5) 24-bit Graphics Card", according to some awesome know-how site.
> Is its companion, the "Processor Card FRU 52G1343 (Placed in Lower Numerical Slot than Graphics Card)" actually required for the operation?
> There is an empty MCA slot right next to it, so I assume the processor card missing.
> If I recall it right, it's not as simple as dropping in an XGA-2, hmm?
>

All of the DPX/20 I've seen are the same as the IBM manufactured
systems. The only special machines are Bull Escala where they had some
oddities such as the PowerPC 620 actually ship (!). So yes your AIX
4.3.3 will work and will be a good fit for this system.

It sounds like you have a GT4i, which will run on its own. And 3w3 is
quite forgiving you just need a Sync-on-green monitor.
http://ps-2.kev009.com/sog/

Regards,
Kevin

Am 01.11.21 um 21:53 schrieb Christian Holzapfel:
> So little remaining memories, so many questions...
>
> I assume, it *could* run my ancient copy of AIX 4.3.3, would happily interact with my PC style Model M, and any SCSI CD-ROM drive would be just fine for the start, right?
>
> And with this video thing... The card is a 11H4913, so it is a "POWER Gt4xi (1-5) 24-bit Graphics Card", according to some awesome know-how site.
> Is its companion, the "Processor Card FRU 52G1343 (Placed in Lower Numerical Slot than Graphics Card)" actually required for the operation?
> There is an empty MCA slot right next to it, so I assume the processor card missing.
> If I recall it right, it's not as simple as dropping in an XGA-2, hmm?
>

Hi Christian,

as for the memory, your machine needs fast page ram with parity (no
ECC!): https://ardent-tool.com/RS6000/7006/7006-41x.html

The POWER Gt4xi consists of two cards. If one is missing it will most
likely not work. If it does work it's a nice graphics card with true
color support for up to 1280*1024 pixels.
https://ardent-tool.com/RS6000/gt4xi_graphics_display.html
If you want to make use of a POWER Gt4xi your monitor must support sync
on green.
What you probably want to have is a GXT500D. This is just an example:
https://www.ebay.de/itm/283171852255
This one has a 13W3 interface which can be adapted to standard VGA.

I don't know any video card which works with both Power PC and x86 PC.

My 7006 runs AIX 4.3.3 from a 3,5" 9GB hard drive, that sits under the
floppy drive. No drive cage is required. To install AIX on your machine
you need the key to turn the lock in the service position, a SCSI
CD-ROM, a SCSI drive, AIX CDROM media and null modem cable to an other
PC with a telnet client.

If you want to get rid of the Kingston ECC ram I gently raise my finger.

Good luck and good night

Wolfgang

Hi everyone.
I made the system pass the RAM error codes by installing 4x 8 MB PS/2 parity RAM. It's enough for a first try, but hopfefully not the final setup ;-)

With key in normal mode, the system now hangs on a 223 code (attempting a normal IPL from SCSI attached devices).
What baffles me is that none of my HDDs likes to spin up when connected to the internal 50-pin SCSI port.
I've disabled the external port (due to lack of a proper terminator), tried two different 50-pin cables and two disks.
They all spin up fine without the SCSI connected, so power is there.
Tried some 9.1 GB SCA drive with an adapter, and an IBM DALS-3540 (540 MB, set to ID 3, spin up, termination on).
Is there a special cable required, is the planar toast or am I missing anything else?

With the key in maintainance mode, it hangs on a 262 code (keyboard not detected), although it flashes the Model M's lights.

Any thoughts?

Wolfgang, I'll check if you can have the leftover ECC RAMs. You're welcome to have it, if the seller doesn't want it back. Im on it.

~Christian

Will the 7006-42W board (finally found a label on it!) boot from an additional MCA SCSI adapter? Which ones are supported on this machine?
I have a spare Future Domain MCS-600/700 / IBM PS/2 SCSI-2 from a 9576 that I will give a try, but that's probably not what it wants, is it?

Christian Holzapfel schrieb am Mittwoch, 3. November 2021 um 14:43:04 UTC+1:
> Will the 7006-42W board (finally found a label on it!) boot from an additional MCA SCSI adapter? Which ones are supported on this machine?
> I have a spare Future Domain MCS-600/700 / IBM PS/2 SCSI-2 from a 9576 that I will give a try, but that's probably not what it wants, is it?

It's not going further with the FD controller either.

After pulling and checking the NVRAM 3 V battery, the stop code has changed from 223 (attempting a normal mode IPL from SCSI-attached devices specified in NVRAM IPL devices list) to 233 ([..] specified in IPL ROM default device) for obvious reasons, but still no disk spinning up.

I just measured the SCSI control lines /BSY, /ACK, /RST, /MSG, /SEL, /C/D, /REQ, /I/O, with and without a drive attached.
Turns out, they all stay low at all times, so most likely, the host controller has deceased.
The planar PCB looks clean and intact, no burnt fuses or blown capacitors, so probably not much to win there.

What are my options? Will it take Spock, Corvette, Passplay, Cheetah and friends?
If yes, will I need to look out for any special revisions?

Sad thanks,
~Christian

On Wednesday, November 3, 2021 at 3:36:10 PM UTC-4, Christian Holzapfel wrote:
> What are my options? Will it take Spock, Corvette, Passplay, Cheetah and friends?

Assuming that the Bull DPX/20 200 is a direct match for an IBM 7006, then it won't work with any of the PS/2 SCSI adapters, it will want one of the RS/6000 SCSI adapters, and it will probably only boot from a 4-1 (FRU 51G9425). This is similar to a Spock, but it has RS/6000 VPD burned into it, and that makes the critical difference. I purged out my RS/6000 microchannel adapters quite a while ago, sorry.

With regards to memory, you'd likely be looking for P/N 74G1020, FRU 73G3235 (Four 32MB modules yielding 128MB RAM), I probably have some of those hiding in the basement somewhere (might take some digging to find them).

Am 03.11.21 um 20:36 schrieb Christian Holzapfel:
> I just measured the SCSI control lines /BSY, /ACK, /RST, /MSG, /SEL, /C/D, /REQ, /I/O, with and without a drive attached.
> Turns out, they all stay low at all times, so most likely, the host controller has deceased.
> The planar PCB looks clean and intact, no burnt fuses or blown capacitors, so probably not much to win there.
>
> What are my options? Will it take Spock, Corvette, Passplay, Cheetah and friends?
> If yes, will I need to look out for any special revisions?
>
> Sad thanks,
> ~Christian
>

A last try with the internal SCSI port:

* is the external SCSI port disabled? To do this, set the jumpers
exactly to the other position than on this picture:
http://www.supervinx.com/OnlineMuseum/IBM/42T/005-IMG_2763.JPG.html

* unplug all expansion cards including the graphics card

* set the SCSI ID of the hard disk to 0 (a dim memory tells me that
counting is the other way around than on PS/2 machines), passive
termination enabled, no other SCSI device on the cable

* Boot the system in service mode

Please report back with which display code the system stops.

Repeat the procedure with the SCSI cable disconnected and and again tell
us with which display code the system stops.

Wolfgang

Wolfgang Gehl schrieb am Mittwoch, 3. November 2021 um 22:37:15 UTC+1:
> Repeat the procedure with the SCSI cable disconnected and and again tell
> us with which display code the system stops.

Thank you Wolfgang, but the issue remains:

With a known-good SCSI drive at ID 0, external SCSI disabled, termination at the HDD on/off, no adapters at all, the system hangs on
262 (keyboard error) in maintainance mode
233 (boot from IPL) in normal mode
with or without the SCSI cable being plugged.
If the SCSI cable is plugged, the disk won't spin up.

While probing the SCSI bus, I didn't even see anything on the control lines - not even a glitch on power-up.

ekb...@vnet.ibm.com schrieb am Mittwoch, 3. November 2021 um 22:18:32 UTC+1:
> Assuming that the Bull DPX/20 200 is a direct match for an IBM 7006, then it won't work with any of the PS/2 SCSI adapters, it will want one of the RS/6000 SCSI adapters, and it will probably only boot from a 4-1 (FRU 51G9425).

Well, at least the Ardent Tool states that the 7006 should happily accept the Corvette in no perticular revision, also for boot:

https://www.ardent-tool.com/IBM_SCSI/SCSI-FW.html#RS6000_Boot_Support

(at least, if you read it in an optimistic way...)

I took out the planar, and everything looks pretty clean and in-order to me.
All self-resetting fuses F1-3 are ok, no leaked electrolytics or blown tantal caps.
No burnt smell, no smoked traces.
Unfortunately, everything between the SCSI controller NCR 53C720 and the connectors seems to be routed on the inner layers.

Well, there's one thing: One capacitor on the buttom side, C256, seems to have fled off the board.
It looks like it had been there once, but had simply fallen off at some point:

http://www.holzapfel.biz/7006-42W/7006_missing_C256.jpg

However, this is at the opposite corner, far away from the SCSI controller itself and its connectors:

http://www.holzapfel.biz/7006-42W/7006_planar_scsi.jpg

Being connected between a 74ABT240 buffer IC's GND and VCC pins, its job is simply stabilizing the IC's power, not being part of any important oscillator circuit or so.
I simply dropped a standard 100 nF back in there.

While having it open, I replaced one of the electrolytic Cs that I found bad in some other device from the same era (an LPKF 91s PCB mill).

No spinning disks at all.

Wow, this is getting desperate...

Probing traces shows that all I/O pins of the SCSI connector are directly connected to pins of the NCR 53C720 controller IC on the board. No broken lines, and no (safety) logic in between.
So if anyone might have put in a 50-pin cable the wrong way around, it might have roasted the SCSI controller itself.

Before I dig in deeper into the 53C720's host interface, I'll try to come up with some additional SCSI adapter first.
The 7006-42W feature matrix
https://www.ibm.com/common/ssi/ShowDoc.wss?docURL=/common/ssi/rep_sm/w/877/ENUS7006-42W/index.html
lists some "IBM SCSI-2 Fast/Wide Adapter/A" which is a regular "Corvette", isn't it?
Getting a card today is difficult enough, but getting the proper cable with that naughty mini-Centronics plug, well......

Am 04.11.21 um 12:54 schrieb Christian Holzapfel:
> Wolfgang Gehl schrieb am Mittwoch, 3. November 2021 um 22:37:15 UTC+1:
>> Repeat the procedure with the SCSI cable disconnected and and again tell
>> us with which display code the system stops.
>
> Thank you Wolfgang, but the issue remains:
>
> With a known-good SCSI drive at ID 0, external SCSI disabled, termination at the HDD on/off, no adapters at all, the system hangs on
> 262 (keyboard error) in maintainance mode
> 233 (boot from IPL) in normal mode
> with or without the SCSI cable being plugged.
> If the SCSI cable is plugged, the disk won't spin up.
>
> While probing the SCSI bus, I didn't even see anything on the control lines - not even a glitch on power-up.
>
>
> ekb...@vnet.ibm.com schrieb am Mittwoch, 3. November 2021 um 22:18:32 UTC+1:
>> Assuming that the Bull DPX/20 200 is a direct match for an IBM 7006, then it won't work with any of the PS/2 SCSI adapters, it will want one of the RS/6000 SCSI adapters, and it will probably only boot from a 4-1 (FRU 51G9425).
>
> Well, at least the Ardent Tool states that the 7006 should happily accept the Corvette in no perticular revision, also for boot:
>
> https://www.ardent-tool.com/IBM_SCSI/SCSI-FW.html#RS6000_Boot_Support
>
> (at least, if you read it in an optimistic way...)
>

Hi Christian,

it seems you're right and the onboard SCSI is not working. My point is
to get past 262 in service mode, just to make sure, the system continues
POST even without a working SCSI subsystem.

Since there is no monitor and no video card you might as well just pull
the keyboard off.

Give your system a boot device and connect the floppy drive. Disconnect
the SCSI cable and tell us what happens.

Wolfgang

PS: I have a spare corvette. If you want it, contact me offline.

Wolfgang Gehl schrieb am Donnerstag, 4. November 2021 um 19:14:18 UTC+1:
> it seems you're right and the onboard SCSI is not working. My point is
> to get past 262 in service mode, just to make sure, the system continues
> POST even without a working SCSI subsystem.
>
> Since there is no monitor and no video card you might as well just pull
> the keyboard off.
>
> Give your system a boot device and connect the floppy drive. Disconnect
> the SCSI cable and tell us what happens.

With floppy connected, and a random or without a diskette in the drive, with or without keyboard, the system finally hangs at 262 when set to maintainance mode.

This is the full error code list. I had to use my phone's slow motion video recording function to capture the quick ones.
Some look pretty fatal to me, indicating some serious planar failures, if they're not exaggerating:

211 IPL ROM CRC comparison error (irrecoverable)
215 A low voltage condition is present (irrecoverable)
218 RAM POST is looking for good memory
21c L2 cache is not detected (That's ok, there is none on this specific model)
219 RAM POST bit map is being generated
220 IPL control block is being initialized
216 IPL ROM code being uncompressed into memory
290 IOCC POST error (irrecoverable)
288 Adapter card slots being queried
291 Standard I/O POST running
283 Serial port POST executing
272 Tablet Port POST
271 Mouse and Mouse port POST
219 RAM POST bit map is being generated
291 Standard I/O POST running
262 Keyboard not detected as being connected to the system's keyboard port

Hi Christian,

a CRC error indicates that the inital program load data stored in ROM is
corrupted.

Am 04.11.21 um 21:11 schrieb Christian Holzapfel:
> 211 IPL ROM CRC comparison error (irrecoverable)

Whether and how such a problem can be solved, I cannot say.
Unfortunately, I don't have a 7006 replacement planar either. That means
I can't help you any further.

Wolfgang

Gurp. Get a copy of the correct ROM? Like that is even likely?

On 11/5/2021 12:23, Wolfgang Gehl wrote:
> Hi Christian,
>
> a CRC error indicates that the inital program load data stored in ROM is
> corrupted.
>
> Am 04.11.21 um 21:11 schrieb Christian Holzapfel:
>> 211    IPL ROM CRC comparison error (irrecoverable)
>
> Whether and how such a problem can be solved, I cannot say.
> Unfortunately, I don't have a 7006 replacement planar either. That means
> I can't help you any further.
>
> Wolfgang
>

Louis Ohland schrieb am Freitag, 5. November 2021 um 19:58:34 UTC+1:
> Gurp. Get a copy of the correct ROM? Like that is even likely?

It could be worth a try.
Looking at
https://www.ardent-tool.com/RS6000/7006/7006-41x.html
are you referring to U5? This seems to be the EPROM to me.

The label states

P/N 11H5838
E/C D 73041
(c)IBM 5628

Maybe someone has something similar, and a reader, and waaay to much time..?

Louis, if I recall correctly, you bought my 7006-42T some years ago. Are you able to read out a dump?

I'll try to read out mine and identify some checksums or seemingly corrupted areas.

Thank you all for your friendly offers and helpful ideas.

Christian Holzapfel schrieb am Samstag, 6. November 2021 um 02:16:12 UTC+1:
> I'll try to read out mine and identify some checksums or seemingly corrupted areas.

The chip is an ST Microelectronics 27C4001, 512 KB, UV-erasable EPROM, chip ID 20 41.

This ist the contents of my chip:

http://www.holzapfel.biz/7006-42W/11H5838_crc_error.bin

The readable copyright notice from address 0x100 onwards indicates that it's probably the correct chip type.

The IPL ROM image can be disassembled and decompiled with Ghidra without any troubles - at least Ghidra doesn't complain about any corruptions during analysis.
Turns out IBM uses some weird implementation of the regular CRC-32 checksums. Can't tell if the implementation is normal, off-standard or just weird-looking through the decompiler.
Guess I could identify the memory regions that are included in the checksum check, and what the final value is compared to...but what for? To find out my ROM is corrupt?
Digging into PowerPC assembler is fun, but it won't help me recover the intact image :-/

Maybe someone in continental Europe has a 7006 and could lend me his EPROM for a readout? I'd return it immediately, I promise. Shipping costs and surprise gifts are on me!

On 07-11-2021 18:06, Christian Holzapfel wrote:
> The IPL ROM image can be disassembled and decompiled with Ghidra without any troubles - at least Ghidra doesn't complain about any corruptions during analysis.
> Turns out IBM uses some weird implementation of the regular CRC-32 checksums. Can't tell if the implementation is normal, off-standard or just weird-looking through the decompiler.
> Guess I could identify the memory regions that are included in the checksum check, and what the final value is compared to...but what for? To find out my ROM is corrupt?
> Digging into PowerPC assembler is fun, but it won't help me recover the intact image :-/
>
> Maybe someone in continental Europe has a 7006 and could lend me his EPROM for a readout? I'd return it immediately, I promise. Shipping costs and surprise gifts are on me!
>
Hi Christian,

ISTR some IPL-ROMs and the like had a kind of safety built in, that is: a copy of the image in a secure, non-erasable portion of its memory. In case of terminal corruption of the working copy, one could copy the safety-copy back into the erasable portion and be back in business.
Maybe??

No 7006 here I'm afraid, sorry.

--
Jelte,
Admirer of the letter of IBM with blue Ishiki

JWR schrieb am Montag, 8. November 2021 um 09:43:29 UTC+1:
> ISTR some IPL-ROMs and the like had a kind of safety built in, that is: a copy of the image in a secure, non-erasable portion of its memory. In case of terminal corruption of the working copy, one could copy the safety-copy back into the erasable portion and be back in business.
> Maybe??

Hi Jelte. I didn't see any duplicate copy of the BIOS inside the EPROMs address space. This is where you expected it, right?

I managed to come up with a proper RS/6000 style Model M keyboard, the one with the speaker populated on the bottom.
With this one, it goes beyond error 262 (keyboard) in Maintainance mode and now halts at
260 - "Information is being displayed on the display console"

Looking at the serial port 1 with a fully populated null modem cable at 9600, 8N1 shows:

TESTING COMPLETED

A PROBLEM WAS DETECTED. Please write down the following
Service Request Numbers (SRN) and report them to your
service representative.

SRN LOCATION DEVICE
868-099 00-00-0S SCSI port

Press "ENTER" to Continue.

SRN 868-099 refers to "Description: SCSI adapter test failed." Well, thanks for that...

When pressing "ENTER" on the serial terminal (not the keyboard!), the system tries to IPL from the floppy drive.
At least, this is *some* progress.

The IBM document "RS/6000 Diagnostic Information for Micro Channel Bus Systems, Version 4.3.3" mentions a set of diagnostic diskettes.
Are these available as downloadable images anywhere?

On 11-11-2021 17:28, Christian Holzapfel wrote:
> JWR schrieb am Montag, 8. November 2021 um 09:43:29 UTC+1:
>> ISTR some IPL-ROMs and the like had a kind of safety built in, that is: a copy of the image in a secure, non-erasable portion of its memory. In case of terminal corruption of the working copy, one could copy the safety-copy back into the erasable portion and be back in business.
>> Maybe??
>
> Hi Jelte. I didn't see any duplicate copy of the BIOS inside the EPROMs address space. This is where you expected it, right?
>

Correct. Well, it was a long shot ;)
Shame it didn't work out.

--
Jelte,
Admirer of the letter of IBM with blue Ishiki