GSETUP regarding expanded memory for IBM 5170

[modem7]

You're going to love this one. I found that the chip in Bank 1 for what had been Bit 12 moved back to Bit 14 displayed the old error of 80000 4000 201. So, the chip was defective. I replaced the chip thinking, did we get them all; not quite looks like there is a new error message 80008 1100 201. Now the 1100 is signaling bit's 12 and 8 as the sources for new error's.

The POST reports the first RAM error encountered, then does no more RAM testing. One fixes that error and then sees if the POST detects more, and so on.

However, it's the 80008 that has me confused. Is the message saying that I have two errors one in the high bank and one in the low bank?

RAM chips can fail in different ways. Some examples:
- Unresponsive ('dead').
- Partially faulty: One or more bits in the chip's cell matrix are faulty.
- Partially faulty: Addressing related problem.

More background: The POST tests using words, not bytes. Because of that, failure address will always be an even number.
E.g. Test word at address 0, test word at address 2, test word at address 4, and so on

Regarding the first example above. The first address of the bank will usually show as faulty. E.g. 80000. However, I have seen the IBM 5170's POST sometimes show the second test address, e.g. 80002

Regarding the second example above. Let's pretend that there is only one faulty cell. If that chip was to be fully tested, it would be discovered that only one cell address fails a read/write test, all the other addresses testing good. The failure address is going to depend on exactly which cell has failed.
So if on one of my 5170's I saw 80010 0002 201, I would say to myself something like, "16 (10 hex) addresses past the start of a bank. That is probably a RAM chip that has a faulty cell somewhere in it." (Not a 'dead' chip.)

The '80008 1100 201' that you are seeing is unusual in that there are two bits (i.e. two chips) showing up at exactly the same 'not the first' address.

We know now that your Cheetah card is providing the 512-640K addresses space from its bank 1. I suggest that you swap out the bits 12 and 8 chips of bank 1, with chips from another bank (which assumes that the chips in the other bank are good). See how that goes.

 

[alejack12001]

The POST reports the first RAM error encountered, then does no more RAM testing. One fixes that error and then sees if the POST detects more, and so on.


RAM chips can fail in different ways. Some examples:
- Unresponsive ('dead').
- Partially faulty: One or more bits in the chip's cell matrix are faulty.
- Partially faulty: Addressing related problem.

More background: The POST tests using words, not bytes. Because of that, failure address will always be an even number.
E.g. Test word at address 0, test word at address 2, test word at address 4, and so on

Regarding the first example above. The first address of the bank will usually show as faulty. E.g. 80000. However, I have seen the IBM 5170's POST sometimes show the second test address, e.g. 80002

Regarding the second example above. Let's pretend that there is only one faulty cell. If that chip was to be fully tested, it would be discovered that only one cell address fails a read/write test, all the other addresses testing good. The failure address is going to depend on exactly which cell has failed.
So if on one of my 5170's I saw 80010 0002 201, I would say to myself something like, "16 (10 hex) addresses past the start of a bank. That is probably a RAM chip that has a faulty cell somewhere in it." (Not a 'dead' chip.)

The '80008 1100 201' that you are seeing is unusual in that there are two bits (i.e. two chips) showing up at exactly the same 'not the first' address.

We know now that your Cheetah card is providing the 512-640K addresses space from its bank 1. I suggest that you swap out the bits 12 and 8 chips of bank 1, with chips from another bank (which assumes that the chips in the other bank are good). See how that goes.

The explanation you have given Is something that you don't learn in school or even on the job.

I suspected that your reply was going to suggest that substitution on other chips in other banks to narrow the errors and hopefully get to the conclusion that would allow 128-KB to be added to the Base memory. I'll keep you informed as I proceed.

 

[alejack12001]

From the chart made below, error is occurring in every chip from Bank 1 on this Cheetah Combo Memory card and has to be replaced as far as the memory. Could there be something else wrong?​

Address Indicator​	Bit Indicator in h​	Binary Bit Indicator​	Bit Row​	Replaced​	Comments​
80010​	0200​	1000000000​	9​	Yes​
80002​	0880​	100010000000​	11 & 7​	11 & 7 Yes​
80080​	0020​	1000000000​	5​	Yes​
80202​	0201​	1000000001​	9 & 0​	9 No & 0 Yes​
8027E​	0400​	10000000000​	10​	Yes​
80202​	1000​	1000000000000​	12​	??​	Already Replaced In earlier Post​

 

[modem7]

From the chart made below, error is occurring in every chip from Bank 1 on this Cheetah Combo Memory card and has to be replaced as far as the memory. Could there be something else wrong?

Earlier, in post #81, I wrote, "(which assumes that the chips in the other bank are good)".
Could it be that your Cheetah card has lots of faulty RAM chips, and sometimes, all you are doing is swapping one bad chip for another?
And some of the faulty chips could be intermittent.
Do you have a source of known-good chips that you can swap in?

Maybe you should take a break from the Cheetah card and work on the IBM card that you mentioned in post #74, the 'IBM 512KB Memory Expansion Option', quite a simple card, and one that provides only extended memory.

 

[alejack12001]

The designated chips that were changed were purchased just recently from Jameco. Of course, they got their chips from a manufacturer which they provide free their website. The type are SN41256-10

Yes, it may be time to move on to the IBM card. The only thing that I want to say here is that the chips on that card are doubles meaning that one is soldered on another. So if one is faulty then both are equally faulty.

I also received my 170ns EPROMs to try to put the U17&U37 in to action. I’ll see if I can download the two binaries directly to avoid any corruption and map them directly into the EPROMs and then apply them to the 5170. I’ll keep you posted here as well.

 

[alejack12001]

Typo error not free but from their website.

 

[modem7]

From the chart made below, error is occurring in every chip from Bank 1 on this Cheetah Combo Memory card and has to be replaced as far as the memory. Could there be something else wrong?

Yes.

If you haven't already, I suggest that you clean the edge connector fingers on the Cheetah card.

In determining whether the Cheetah card is causing the symptoms, or whether something else in the 5170 is, you could try out the Cheetah card in your IBM 5162. If you leave the Cheetah card configured as is (providing RAM between the addresses of 512K and 640K), you would need to put a jumper on pins 2 and 3 of motherboard J10 in order to stop the 5162 motherboard from providing that RAM.

 

[alejack12001]

Yes.

If you haven't already, I suggest that you clean the edge connector fingers on the Cheetah card.

In determining whether the Cheetah card is causing the symptoms, or whether something else in the 5170 is, you could try out the Cheetah card in your IBM 5162. If you leave the Cheetah card configured as is (providing RAM between the addresses of 512K and 640K), you would need to put a jumper on pins 2 and 3 of motherboard J10 in order to stop the 5162 motherboard from providing that RAM.

I know that my IBM PC-XT 5162 doesn't house an IBM motherboard 5162, but an AT clone board that has AMI ROM's on it with VLSI chips, AMD 80286 processor with a 10-MHz clock, 80287-10 NPS and 1-MB RAM. The official listing of the motherboard is unidentified 286 motherboard according to Stason.org, which has a similar board with 2-banks of added RAM for SIMS. I know that the board was made in Taiwan. Enough said about the motherboard except I did try the Cheetah card there sometime ago and it's reaction was unusual as the Cheetah cause the motherboard to hang just after POST with no apparent errors as though the motherboard couldn't see the IDE controller. As soon as I removed the Cheetah, everything returned to normal. The actual IBM 5162 motherboard was up for sale in Europe some months ago but 325 euro's with 125 euros for shipping and was too pricey for me.

I'll try cleaning the contacts on the Cheetah as was suggested.

 

[alejack12001]

IBM Expansion (Extended) Memory Card Option 512-KB Test Results​

The block diagram of my card is exactly the same as the 512-KB memory card seen associated with that of the minus zero degrees website as shown in the following link: https://minuszerodegrees.net/5170/cards/5170_memory_board_512K_ram_layout.jpg

The setup was following the instructions given for on memory card of 512-KB starting at address 100000h.The memory chips on the card are dual stacked 128-KB chips Texas Instruments ZA1250NL/AP8510.
I used GSETUP to configure the extended memory and reset the base memory to 512-KB with extended memory to the following configurations:

1. Extended memory set to 512-KB with the Base memory set to 512-KB.

The following error was captured after GSETUP was used to set the extended memory to 512-KB:​

100000 FFFE 201 FFFE, which in binary is 1111 1111 1111 1110 suggests all chips are bad except the last one.​

​

2. Extended memory set to 384-KB with the Base memory remaining the same as step 1.​

​

100000 0001 201 0001, which in binary is 0000 0000 0000 0001 suggest that one chip is bad, being the first chip of either Bank 0 at U51 or Bank 1 at U30.​

​

3 & 4. Extended memory set to 256-KB and 128-KB, which in both cases were the same as 2.​

So, am I setting this card incorrectly or the machines GSETUP incorrectly? In the first try, I get a response that all the chips are defective except 1 while in the lower settings I get the opposite of the first that only one dual stacked chip is bad. The errors seem to point to all except one or the opposite, which make no sense to me. I do believe that there is a problem with this IBM motherboard. I got this motherboard from an eBay seller in Atlanta, Ga. He's in the process of selling the sister board; I am tempted.

 

[modem7]

'IBM 512KB Memory Expansion Option' in IBM 5170

100000 FFFE 201 FFFE, which in binary is 1111 1111 1111 1110 suggests all chips are bad except the last one.

That bit 0 showing as good is a red herring; a known behaviour of the 5170 motherboard. We know that it is a behaviour of the motherboard and not the IBM BIOS, because, as shown at [here], the SuperSoft/Landmark Diagnostic ROM (i.e. used in place of the IBM BIOS ROM) reports the same.

The "FFFE" is typical of no RAM being found at the indicated address. (BTW. Sometimes FFFF is shown instead.)

100000 0001 201 0001, which in binary is 0000 0000 0000 0001 suggest that one chip is bad, being the first chip ...

And yet, here, RAM is being found at the same address (1 MB).

And the '0001' highly suggests a bad RAM chip, or one that just needs to be re-seated in its socket.

So, am I setting this card incorrectly or the machines GSETUP incorrectly?

Case #1 is suspect, and I think this could be switch settings on the card.

So,
- 512 KB of base/conventional RAM on the motherboard.
- 'IBM 512KB Memory Expansion Option' set to start its RAM at the 1 MB address mark.
- GSETUP used to configure CMOS SETUP for: base/conventional=512K, expansion/extended=512K

On the 'IBM 512KB Memory Expansion Option', the switch settings would be set per [here]. Step 1 done, then row 1 of step 2. In my 5170's, I have 512 KB of extended memory, provided by one of these cards. A photo of the switch settings follows.

And in case of 'dirty' switches, try adjusting each switch between the on and off positions a few times. And something that we have seen on these forums is a switch that has not been pushed 'fully home'.

 

[alejack12001]

'IBM 512KB Memory Expansion Option' in IBM 5170


That bit 0 showing as good is a red herring; a known behaviour of the 5170 motherboard. We know that it is a behaviour of the motherboard and not the IBM BIOS, because, as shown at [here], the SuperSoft/Landmark Diagnostic ROM (i.e. used in place of the IBM BIOS ROM) reports the same.

The "FFFE" is typical of no RAM being found at the indicated address. (BTW. Sometimes FFFF is shown instead.)


And yet, here, RAM is being found at the same address (1 MB).

And the '0001' highly suggests a bad RAM chip, or one that just needs to be re-seated in its socket.


Case #1 is suspect, and I think this could be switch settings on the card.

So,
- 512 KB of base/conventional RAM on the motherboard.
- 'IBM 512KB Memory Expansion Option' set to start its RAM at the 1 MB address mark.
- GSETUP used to configure CMOS SETUP for: base/conventional=512K, expansion/extended=512K

On the 'IBM 512KB Memory Expansion Option', the switch settings would be set per [here]. Step 1 done, then row 1 of step 2. In my 5170's, I have 512 KB of extended memory, provided by one of these cards. A photo of the switch settings follows.

And in case of 'dirty' switches, try adjusting each switch between the on and off positions a few times. And something that we have seen on these forums is a switch that has not been pushed 'fully home'.

View attachment 1283977

Your switches are slide and mine are rockers; never-the-less they could be dirty, I don't know. I can flip them several times and retest to see if there is a different response. I have the users manual, which showed the switch settings for one 512-KB board placement that matched the table settings you pointed out. If the response is the same then I'll have to presume that the first chip set (since their are a pair) are defective. Is the chip pair U30 from Bank 1 or U51 from Bank 0 the cause of the error, both are labeled 1 in the following link: https://minuszerodegrees.net/5170/cards/5170_memory_board_512K_ram_layout.jpg ? I also used the GSETUP with the 512 at the base or conventional memory and the 512 extended memory. That's how I got the first set of numbers similar to the Landmark picture you showed.

 

[modem7]

Your switches are slide and mine are rockers; never-the-less they could be dirty, I don't know.

The photo in post #91 reveals that you have the switches set opposite of what they should be. Years ago, this was happening often enough that I created the web page at [here].

If the response is the same then I'll have to presume that the first chip set (since their are a pair) are defective. Is the chip pair U30 from Bank 1 or U51 from Bank 0 the cause of the error, both are labeled 1 in the following link: https://minuszerodegrees.net/5170/cards/5170_memory_board_512K_ram_layout.jpg ?

'Chip pairing' is not how I would describe the situation.

0001 indicates bit 0. My recall of this board is that the low half of addresses are supplied from bank 0. (I could look at the circuit diagram, but I'm really confident about that.) So, bit 0 in bank 0 is U50. Hopefully, only a re-seating is required.

 

[alejack12001]

The photo in post #91 reveals that you have the switches set opposite of what they should be. Years ago, this was happening often enough that I created the web page at [here].


'Chip pairing' is not how I would describe the situation.

0001 indicates bit 0. My recall of this board is that the low half of addresses are supplied from bank 0. (I could look at the circuit diagram, but I'm really confident about that.) So, bit 0 in bank 0 is U50. Hopefully, only a re-seating is required.

It could be possible that I have the switches reversed. Let me further investigate. I have ordered some chips in case the re-seating and the switches that could be incorrectly set aren't the root of the problem. Okay, I guess I will have to try re-seating U50.

I have some good news for you. The binaries that you made and sent to me, worked for U17 & U37. I was able to load them into the 256-KB EPROM's successfully this time and mount the chips to the IBM motherboard correctly. Upon re-boot the display of the floppy/hard drive showed at the top with my Transcend 2-GB CF card being recognized. I am still, for now booting from my old MFM hard drive as drive C with the Transcend 2-GB being the D drive. Yay! it works and now the XUB is part of the boot process. I am so pleased and grateful that the binaries were the answer.

 

[modem7]

I have some good news for you. The binaries that you made and sent to me, worked for U17 & U37. I was able to load them into the 256-KB EPROM's successfully this time and mount the chips to the IBM motherboard correctly. Upon re-boot the display of the floppy/hard drive showed at the top with my Transcend 2-GB CF card being recognized.

Using your 'Lo-tech XT-CF-lite rev.2', correct? And presumably with its EEPROM disabled, because with U17+U37 taking over its role as the XUB hosting ROM, it is not required.

Still, the puzzle remains as to why you needed to do down this path. Your 'Lo-tech XT-CF-lite rev.2' (with EEPROM enabled) works elsewhere. Something that can be revisited later.

 

[alejack12001]

Using your 'Lo-tech XT-CF-lite rev.2', correct? And presumably with its EEPROM disabled, because with U17+U37 taking over its role as the XUB hosting ROM, it is not required.

Still, the puzzle remains as to why you needed to do down this path. Your 'Lo-tech XT-CF-lite rev.2' (with EEPROM enabled) works elsewhere. Something that can be revisited later.

Yes, I am using the Lo-Tech XT-CF lite Rev 2. Yes, the ROM the Lo-Tech is disband.

Answer to the second sentence, I guess can be summed up in the way this IBM motherboard has behaved before the implementation of the U17 & U37 and now with the EEPROM’s in place the XT-CF card is recognized.

 

[alejack12001]

Yes, I am using the Lo-Tech XT-CF lite Rev 2. Yes, the ROM the Lo-Tech is disband.

Answer to the second sentence, I guess can be summed up in the way this IBM motherboard has behaved before the implementation of the U17 & U37 and now with the EEPROM’s in place the XT-CF card is recognized.

My answer above is probably not making any sense. What I meant to say was that the 5170 motherboard rejected any attempt to place the Lo Tech XT-CF into the startup process of the machine before triggering MSDOS. I can't say why the built in PROM of the Lo Tech card failed to trigger the XUB; but now with your help and the implementation of the U17 & U37 ROM chips the Lo Tech is working perfectly.

 

[modem7]

What I meant to say was that the 5170 motherboard rejected any attempt to place the Lo Tech XT-CF into the startup process of the machine before triggering MSDOS. I can't say why the built in PROM of the Lo Tech card failed to trigger the XUB; but now with your help and the implementation of the U17 & U37 ROM chips the Lo Tech is working perfectly.

Yes, I know. You have interpreted my "Still, the puzzle remains as to why you needed to do down this path." statement, not as a statement, but as a question to you. Later, if you want, we can retry to get the card working as intended, that is, XUB in the card's ROM. But that is up to you. You may be fine with always using the workaround.

And 'on the back burner' is the Cheetah card, presently configured to provide RAM from addresses 512-640 KB. I have an idea about that, but I need to prove something first before presenting it here.

Right now, it would be good to sort out the 'IBM 512KB Memory Expansion Option'. That issue is looking presently like bad switch settings plus a faulty RAM chip.

 

[alejack12001]

Yes, I know. You have interpreted my "Still, the puzzle remains as to why you needed to do down this path." statement, not as a statement, but as a question to you. Later, if you want, we can retry to get the card working as intended, that is, XUB in the card's ROM. But that is up to you. You may be fine with always using the workaround.

And 'on the back burner' is the Cheetah card, presently configured to provide RAM from addresses 512-640 KB. I have an idea about that, but I need to prove something first before presenting it here.

Right now, it would be good to sort out the 'IBM 512KB Memory Expansion Option'. That issue is looking presently like bad switch settings plus a faulty RAM chip.

In your first reply, I guess I am old school as if it works don't fit it. However, curiosity may take over and I may want to see what's on that Lo Tech ROM to see if XUB code is there.

As you know, I tried different variations with the Cheetah card with the 512-640 and came up empty handed. For some reason this card is not being liked by the system. When I get the opportunity, I plan to step back a bit and see if I can get another card that I have that is really made for an XT machine (that is, 8-bit port) to see if it can fill that 128-KB gap. I have a second AST 6 pack plus that I used on my 5150 that was the twin of the card used in the 5150 there. I would keep the clock off since the 286 has it's own real time clock and setup the card with just 128-KB of RAM and taking off the 256-KB of RAM that are on the card. Setup the card for the purpose of filling that 512-640 gap and use it's serial/parallel port to eliminate the need for a card for that purpose. The type 3 motherboard does have 2-8-bit slots on the board. I'll have to let you know if this theory works or not.

I did receive two sets of the dual stacked RAM chips for the IBM 512-KB Memory Expansion Option, which could be tested later if the AST fails of course. One way or another I plan to get memory added to this machine. Although it may take a while to do so.

 

[alejack12001]

An AST six-pack plus become the way to raise the base or conventional memory from 512 to 640-KB. I stripped the card down to 128-KB and turned off the serial/parallel ports as well as the Real time clock. To my surprise, the IBM 5170 recognized the added memory and of course gave an error that I needed to adjust the conventional memory to 640. After I made the correction then the 5170 rebooted to the 640-kB.

Modem7, you provided a link to the minuszerodegrees picture of the switches for the IBM Extended Memory Option card, which housed the added 512-KB memory. This picture showed the same rocker switches used on the card and the one sentence that was the game changer for that card was the switches were all off even though the red markers were all in the on position. So with that in mind, I asked myself, "Do I have these switches backwards and the answers was Yes, I do." When I flipped the switches in the opposite direction and placed the card in the machine, I got another memory error where the machine recognized the added 512-KB of memory. After I corrected the extended memory, the machine rebooted with 1152-KB of RAM. The picture of the rocker switches were the key element in this whole effort with the IBM Extended Memory Option card.

I don't know if the switch settings are reversed on the Cheetah card. I will have to address this at a later date.

 

[modem7]

An AST six-pack plus become the way to raise the base or conventional memory from 512 to 640-KB. I stripped the card down to 128-KB and turned off the serial/parallel ports as well as the Real time clock. To my surprise, the IBM 5170 recognized the added memory and of course gave an error that I needed to adjust the conventional memory to 640. After I made the correction then the 5170 rebooted to the 640-kB.

Good as an experiment. From the perspective of an AT-class motherboard, the '8-bit data path' card is 'slow'.

This picture showed the same rocker switches used on the card and the one sentence that was the game changer for that card was the switches were all off even though the red markers were all in the on position.

Those red bars/markers on switches have misled people. In the photos at [here], originally, the switch that has red bars didn't have red bars on it. Many years ago, a member got confused by the red bars on his/her switches, and to help other people, added red bars to the photo, and sent to me the revised photo.

The picture of the rocker switches were the key element in this whole effort with the IBM Extended Memory Option card.

It's good to see progress being made. And some relief for you, because it is becoming less likely that you have a faulty 5170 motherboard.