No luck getting third RAM bank working in AST Six Pack Plus

[themaritimegirl]

(EDIT: The title should say *sixth* RAM bank, not third.)

My IBM 5155 has 256 KB of memory on the motherboard, and 256 KB on an AST Six Pack Plus card. I recently bought 18 DRAM chips to fill the last two banks on the card to get the full 640 KB. I installed the chips and set the card's DIP switches for all six banks, and when I powered up the computer, it only counted up to 256 KB before stopping and booting (with no error). If I turn banks 5-6 off again, it counts up to 512 KB and boots with no problem.

I removed the chips in banks 3-4, and moved the new chips from banks 5-6 to banks 3-4. The computer sees 512 KB and boots fine. A memory test came out clean. So I know the new chips are working fine.

I installed the old chips back in banks 3-4 and the new chips back in banks 5-6. I turned on just bank 5, and the computer can now see 576 KB. Still no luck getting bank 6 to work.

I've spent the last two hours rotating the chips in bank 6, swapping chips with other banks, cleaning pins, etc etc... I can't get bank 6 on the card to work at all. My best guess is that the last two banks have never had chips in them, and may be slightly oxidized. But a dozen reseats, and cleaned pins on the chips hasn't solved that. I can't help but wonder if there's something simple/stupid that I have forgotten. The starting address on the card is set correctly, to :40000.

Any ideas?

 

[modem7]

Bank 1: 256 to 320
Bank 2: 320 to 384
Bank 3: 384 to 448
Bank 4: 448 to 512
Bank 5: 512 to 576
Bank 6: 576 to 640

I installed the chips and set the card's DIP switches for all six banks, and when I powered up the computer, it only counted up to 256 KB before stopping and booting (with no error).

So, when all six banks on the SixPakPlus are selected, the SixPakPlus is behaving as though no banks are selected (i.e. 5160 only counting the 256 KB on the motherboard).

If this was a problem with bank 6 functionality only, then the 5160 would be counting up to 576 KB.

The symptoms sound like a problem in the circuitry that generates the various bank-select lines.

 

[mikey99]

Try removing all the chips from Bank 6 , leave the dip switches set for all 6 banks, and see if you get a memory error at boot.

Also, maybe the dip switches aren't making good contact, try flipping them back and forth a few times.

There are several versions of the AST SixPack, is this the short or long card ?

 

[modem7]

Try removing all the chips from Bank 6 , leave the dip switches set for all 6 banks, and see if you get a memory error at boot.

An error is not expected in that scenario due to an apparent design flaw in the BIOS of the IBM 5155 and 5160. See [here].

 

[themaritimegirl]

If this was a problem with bank 6 functionality only, then the 5160 would be counting up to 576 KB.

The symptoms sound like a problem in the circuitry that generates the various bank-select lines.

It actually does the same if there is simply a poor connection to a chip. When I first tested the new chips in banks 3-4, it behaved the same, only counting up to 256 KB and not issuing an error. When I reseated the chips, it counted to 512 KB successfully. That's why I wonder if there's a really crusty socket. All the solder joints look good, at least. It will suck if it's an issue with the bank selection circuitry, like you mentioned.

This is the long version of the card.

 

[modem7]

This is the long version of the card.

The variations of the long version are shown at [here].

It actually does the same if there is simply a poor connection to a chip. When I first tested the new chips in banks 3-4, it behaved the same, only counting up to 256 KB and not issuing an error. When I reseated the chips, it counted to 512 KB successfully.

So, at that point in time, replacing chips in banks 3 and 4 (which does not require a change of switch settings), resulted in bank 1 not seen by the 5155, or bad bits in bank 1 being seen by the 5155.

The same symptom would have been seen if the SixPakPlus was not seated correctly. (You would have had to remove the card from the 5155 in order to swap RAM chips in/out.)

That's why I wonder if there's a really crusty socket.

If the only problem is a 'crusty socket', then the expected results of enabling all six banks on the SixPakPlus are:
* 'Crusty socket' is in bank 1: 5155 counts up to 640 KB when socket is 'good', and up to 256 KB when socket is 'bad'.
* 'Crusty socket' is in bank 5: 5155 counts up to 640 KB when socket is 'good', and up to 576 KB when socket is 'bad'.

So, if there is only one 'crusty socket', it is not in bank 5.

But maybe you have multiple 'crusty sockets'.

In any case, you can rule out 'crusty sockets' by repeatability of the symptoms you presented earlier.
Try this:

1. During the following, do not remove the SixPakPlus from the 5155 (i.e. avoiding undue physical disturbance).
2. On SixPakPlus, set switches to enable first 4 banks. Watch 5155 count to 512 KB (expected result, as you observed earlier).
3. On SixPakPlus, set switches to enable all 6 banks. Watch 5155 count to 256 KB (unexpected result, as you observed earlier).
4. On SixPakPlus, set switches to enable first 4 banks. Watch 5155 count to 512 KB (expected result, as you observed earlier).
5. On SixPakPlus, set switches to enable all 6 banks. Watch 5155 count to 256 KB (unexpected result, as you observed earlier).

If you observed that, that kind of repeatability cannot be caused by a 'crusty socket'.

 

[themaritimegirl]

Okay, I pulled the computer out again:

Try removing all the chips from Bank 6 , leave the dip switches set for all 6 banks, and see if you get a memory error at boot.

It counts to 256 KB and boots with no error.

maybe the dip switches aren't making good contact, try flipping them back and forth a few times.

Done; no dice.

The variations of the long version are shown at [here].

My card has none of the part numbers shown there. Mine says 201177-010. It's actually written 02-01177-010.

The same symptom would have been seen if the SixPakPlus was not seated correctly.

Not that I've seen - I've booted with the card ill-seated a couple of times. When it happens, the computer throws an error referencing 40000. I won't rule out the possibility that the card could be ill-seated in such a way that the computer doesn't throw an error, but so far I haven't observed it - the real-time clock still works when the computer boots without acknowledging the extra memory.

1. During the following, do not remove the SixPakPlus from the 5155

Impossible - I have to remove it to access the DIP switches. I will have to trust that the computer will throw a 40000 etc error if the card is ill-seated.

2. On SixPakPlus, set switches to enable first 4 banks. Watch 5155 count to 512 KB (expected result, as you observed earlier).

Counted to 512 KB.

3. On SixPakPlus, set switches to enable all 6 banks. Watch 5155 count to 256 KB (unexpected result, as you observed earlier).

Counted to 256 KB.

4. On SixPakPlus, set switches to enable first 4 banks. Watch 5155 count to 512 KB (expected result, as you observed earlier).

Counted to 512 KB.

5. On SixPakPlus, set switches to enable all 6 banks. Watch 5155 count to 256 KB (unexpected result, as you observed earlier).

Counted to 256 KB.

If you observed that, that kind of repeatability cannot be caused by a 'crusty socket'.

Are you sure? I just tested by enabling banks 1-4, and removing a chip in bank 4. The computer only counted to 256 KB. As such, if the computer works fine with banks 1-5 enabled, but only counts to 256 KB when banks 1-6 are enabled, wouldn't that indicate ill communication with at least one chip in bank 6?

I know what you're getting at - the PC and PC/XT have a dumb method for determining the amount of RAM, basically testing successive addresses until it reaches one with no memory. It seems to me that the SixPackPlus is smarter than that, though - detecting when there's a RAM issue and locking out all RAM banks when it happens. Just a guess, though.

 

[mikey99]

Have you tried cleaning the SixPack ISA card edge connector.......the best way to do this is using a pencil eraser
to clean the contacts. Just rub away until the contacts are nice and shiny. I've 'fixed' a few flaky cards recently
by doing this.

 

[themaritimegirl]

Have you tried cleaning the SixPack ISA card edge connector

Done; no dice. I had hope for that one though - the contacts were rather dirty. I just never thought to clean them since the card otherwise works perfectly.

 

[modem7]

Are you sure? I just tested by enabling banks 1-4, and removing a chip in bank 4. The computer only counted to 256 KB.

I am not seeing what you are.

I pulled out my early 5160. It has the early motherboard and BIOS, matching the 5155. My 5160 has a SixPakPlus, labeled "ASSY 03-01177-010". I think the "010" bit may be a PCB revision.

1. Powered up 'untouched' 5160. 5160 counted to 640 KB as expected.
2. Removed SixPakPlus. 5160 counted to 256 KB as expected. No error.
3. Enabled only first 4 banks on SixPakPlus (by setting SixPakPlus switches 4/5/6 to on/off/off). 5160 counted to 512 KB (256K + 4 banks) as expected. No error.
4. Removed a chip from fourth bank (photo [here]). 5160 counted to 448 KB (256K + 3 banks) as expected. No error.

 

[themaritimegirl]

What the heck. I just replicated what you did, and also got a count to 448 KB. Yet I performed the same test earlier, albeit having removed a different chip, and it stopped at 256 KB.

I am unsure of what's going on here. Still stops at 256 KB with all six banks filled and all six banks enabled. I know it's not a problem with S5, as other configurations where S5 is turned ON work as expected.

Grasping at straws, I removed my XT-IDE controller in case that made a difference. It did not.

 

[modem7]

I additionally ran the DEBUG procedure at [here], using "7000:0" instead of "8000:0" (70000 hex --> 448K).
I read back byte 10 after writing 00, indicating a problem with bit 4.

 

[themaritimegirl]

So CheckIt and DEBUG can be used to test memory that the computer itself doesn't acknowledge? Would it be any use to use one of these tools somehow to test the final bank?

 

[modem7]

So CheckIt and DEBUG can be used to test memory that the computer itself doesn't acknowledge?

Yes.

Would it be any use to use one of these tools somehow to test the final bank?

I don't think so.

If your symptom was that with six banks populated and enabled on the SixPakPlus, the 5155 was only counting as far as the fifth bank, then one hypothesis is bad RAM in the sixth bank. That is where the CheckIt/DEBUG procedures can be used to indicate the faulty bit/s.

But your SixPakPlus is doing something different; disabling all banks when instructed to enable all banks.

 

[themaritimegirl]

Darn, it really looks like my card isn't working correctly, doesn't it.

Well, 576 KB is almost 640 KB...

 

[mikey99]

So based on the info in the link modem7 posted earlier, describing the difference in the way the 5155/5160 test for available memory.......
would testing the card on a 5150 provide any further error indications ? Or would we expect the behaviour to be the same.

 

[modem7]

So based on the info in the link modem7 posted earlier, describing the difference in the way the 5155/5160 test for available memory.......
would testing the card on a 5150 provide any further error indications ? Or would we expect the behaviour to be the same.

Whereas the 5160 motherboard stops at the 256 KB address with no error indication, the 5150 motherboard (64KB-256KB type) would generate a 201 error at the same address (having been informed via SW2 that 640 KB of conventional memory exists).

 

[themaritimegirl]

I posted about this issue on Reddit/r/vintagecomputing, and got a very surprising response - someone else on there has a 5155 with a SixPackPlus, and theirs is having the exact same issue! They can get 576 KB out of it, but not 640 KB - the card just acts like no banks are selected.

A couple other people posted their SPP dip switch configuration, and it matches mine.