5150 Supersoft Memory faulire - Which Chip ?

[KLund1]

Hi again.
I have a 16-64 5150 board with a bank 0 ram fault.
All other banks are unpopulated, and the SW1 is set to only bank zero. I have done the piggy back thing to all nine chips, but I get the same fault from SuperSoft/Landmark Rom test. I am not certain my that all my extra ram chip are good. Most probably are but Iam not sure which. (I with there was a cheap ram tester like there is for LS74xx IC's)
When I run the test I first show location 4050, then on the 2nd try, it show location 4000. That is is I am reading it right. it looks like the characters are doubled or missing.
See picture.
Which chip(s) are bad?
Thanks

 

[Ruud]

First: why the "piggy back thing"? Why not replacing the DRAMs one by one?

Anyway: if the address starts with a 4 then the fault is in the second bank. 4000h = 16 KB. The first 16 KB range from 0000h to 3FFFh. But as this bank is empty. something else is going on. And having seen the picture of your screen, definitely something weird is going on. Check the minuszerodegrees site: this image .

Good luck!

 

[modem7]

I have a 16-64 5150 board ...

Just in case you did not see it, look at the warning at the top of [here].
It means that we need to be careful about what Supersoft/Landmark diagnostics against RAM.

...with a bank 0 ram fault.

Well, I would say that is an assumption at this point. For example, per the above pointed-to warning, on a 16KB-64KB motherboard, the 16K CRITICAL MEMORY REGION can fail if bank 0 is good.

I have done the piggy back thing to all nine chips, but I get the same fault from SuperSoft/Landmark Rom test

Per [here], the piggybacking technique is not a reliable method.

All other banks are unpopulated, and the SW1 is set to only bank zero.

On a 5150 motherboard, the RAM switches do not enable/disable banks of RAM. Plus, the Supersoft/Landmark diagnostics do not read the same switches to determine bank population. The Supersoft/Landmark diagnostics are expecting a 16KB-64KB motherboard to be fully RAM populated, and accordingly always test up to 64 KB

Which chip(s) are bad?

Something to keep in mind is that the diagnostics are not testing RAM chips; they are testing the RAM system. Hopefully in your case, RAM chip/s will turn out to be the problem cause.

Check the minuszerodegrees site: this image .

The image at [here] would have been better (for content perspective, not appearance) - the one for the 16KB-64KB motherboard.
On that motherboard, the two latter memory tests of the Supersoft/Landmark diagnostics only go as high as 64 KB.

it looks like the characters are doubled or missing

Obviously, that confuses the matter.
I have seen a clone video card display something similar (when using the Supersoft/Landmark diagnostics).
Do you have another suitable video card ?

 

[KLund1]

Many thanks for the detailed replies!
I understand that SuperSoft wants all banks to be filled.
What I failed to put in the initial post is that this early 5150 dose not boot at all. From the very excellent MinusZeroDegrees web site, a failure of a ram chip in the first bank will result in a no boot/post situation. So I do not get a ram error code on screen that I could use to find bad ram in the other 3 banks. Did I get that right? That is why I was hoping the supersoft with it's known issues for this early board, might point me to the bad ram chip in bank 0. These are all solder on to the board. I would rather not unsolder them all, and put in sockets. My solder skills are not that good, and the less I have to do on this old board the better. I also hope it is just the ram chips and the the ram system as was pointed out.
I had the Supersoft program not complete as it does the high/low and eight beeps for every missing ram chip, I think.
I only have the one IBM green MGA card.

 

[modem7]

I understand that SuperSoft wants all banks to be filled.
What I failed to put in the initial post is that this early 5150 dose not boot at all. From the very excellent MinusZeroDegrees web site, a failure of a ram chip in the first bank will result in a no boot/post situation.

Well, one of quite a few things that will result in a 'motherboard appears dead when IBM BIOS ROM is fitted' symptom.

But that symptom together with the Supersoft/Landmark diagnostics getting as far as the 16K RAM check, certainly points to either:

* bad RAM chip (or chips) in bank 0; or
* faulty RAM system.

I just now got out a good 16KB-64KB motherboard (fully RAM populated), then disabled RAM chip U43 (bit 5 of bank 0), then ran the Supersoft/Landmark diagnostics.
So, bad was one bit in bank 0, with all other RAM good.

The 16K CRITICAL MEMORY REGION test failed. During that particular test, it first showed bit 5 at address 05040. Then the address changed to 04000. Then the address changed to 0C000. But the bit 5 indication was consistent.

Your 'corrupted' display makes it difficult to work out which bits are indicated at fault.

My solder skills are not that good, and the less I have to do on this old board the better.

I agree.

I only have the one IBM green MGA card.

I used an IBM MDA card, and saw no screen 'corruption' at all. My other IBM MDA card is the same.

You have no CGA or EGA cards, whether they be IBM made or clones ?

 

[KLund1]

Sorry it took so long to bet back to this. Been swapped at work.
I filled all banks with the pulled ram chips. I put them in a random order. But all banks are filled, all facing the correct direction as the solder bank 0 chips.
I found a working EGA card.
I ram SUpersoft again for about 45 minutes. It is stuck on the same test. It can't get past it it. I let it try 70 times.

Do you think I should pull out U45 first and see what happens. Or might this be the ram system?

 

[KLund1]

"The 16K CRITICAL MEMORY REGION test failed. During that particular test, it first showed bit 5 at address 05040. Then the address changed to 04000. Then the address changed to 0C000. But the bit 5 indication was consistent."

On the first pass it shows 05040, then stays at 04000.

 

[modem7]

I found a working EGA card.

Good to see an uncorrupted display.

I ram SUpersoft again for about 45 minutes. It is stuck on the same test. It can't get past it it. I let it try 70 times.

Basically, the diagnostics want/need you to fix the RAM problem before they will go on to do the other tests. That may be because some of the other tests use RAM.


I have just now done some experiments on my motherboard and observed something odd. If I disable a chip by grounding pin 13 (data out), the diagnostics report the bit as faulty. But if I remove a chip, the diagnostics do not report the bit as faulty.

Time to head to bed.

 

[modem7]

Note: The following applies to the 16KB-64KB version of the 5150 motherboard. It has yet to be ascertained if it is applicable elsewhere.

Note: RAM chips have different modes of failure. By grounding pin 14 (data out), I am simulating one of those modes. Below I am referring to that as SFVGDO (simulated failure via grounded data out).


With one SFVGDO chip in each of the banks (0/1/2/3), it was clear that the diagnostic's 'FAILING BITS' represents failures across all 4 banks. And the display of those SFVGDO bits was quicker than I could discern. It was certainly not a case of: There is SFVGDO bit in bank 0 appearing, and then, oh, there is the one in bank 1 getting added to the list.

Per my earlier post, simulating bit failure by chip removal rather than by SFVGDO, had the unexpected result of not showing failure of that bit.
By removing all chips in banks 1/2/3, the diagnostic then only showed SFVGDO bits in bank 0:

1. On the first pass (only the first pass) of the 16K CRITICAL MEMORY REGION test, the address against MEMORY ERROR AT ADDRESS was 05040, with the FAILING BITS showing the SFVGDO bits in bank 0.

2. The SYSTEM MEMORY AT ADDRESS 04000 test then ran. The address against MEMORY ERROR AT ADDRESS changed from 05040 to 04000. The FAILING BITS remained as is, showing the SFVGDO bits in bank 0.

3. The diagnostic looped back to start the tests again, however in that iteration, and all subsequent iterations, the MEMORY ERROR AT ADDRESS remained at 04000.

Photo at [here].


With this motherboard, and suspected bad chips in bank 0, it appears valid to simply remove the chips in banks 1/2/3 in order to exclude banks 1/2/3.

Back at post #1, you had banks 1/2/3 unpopulated. You also see the 05040 to 04000 transition that I see. But at that time, you had the 'corrupted' display. What bits are reported in error now if you remove the chips in banks 1/2/3 ?

 

[KLund1]

Very interesting read. Thanks again for all the help!
I pulled banks 1,2,3 and ran the test again.
I get the exact same results on the first pass as posted above in #7. 05040 then 04000, rinse and repeat. Also with the same failing bits.

 

[modem7]

05040 then 04000

Same as what I see, so we ignore that.

Also with the same failing bits.

But what are those bits ? All I have to refer to is the photo in post #1 where the display is corrupted with things doubled up.

 

[KLund1]

See pic in post #6
bits 7 to 1 are X'd

 

[modem7]

See pic in post #6
bits 7 to 1 are X'd

From my perspective, that was when you populated banks 1/2/3, thus possibly introducing extra faulty bits (due to bad chips).
But I will take it from your answer, that with the RAM in banks banks 1/2/3 removed, the diagnostics show faulty bits of 7 to 1.

That is a lot of bits, and interesting that bit 0 always shows as good.

Originally, I used to think that the 16K CRITICAL MEMORY REGION test did not rely on operational RAM refreshing (that the later RAM test did), but by experiment, I discovered that it did. At post #39 at [here] are some results of mine from simulating failure of RAM refresh, and seeing what the impact was on the 16K CRITICAL MEMORY REGION test. Then, I saw only one or two bits reported in error, and they varied.

However, maybe the count of bits depends on the particular motherboard.

 

[KLund1]

Also very interesting. But is starting to sound more like the ram system, and not only the ram chips.
But I am still willing pull 1 or more ram chips off the board. I'd like to have some idea of which one(s) to pull and see what happens with other ram chips
What do you think?

 

[modem7]

Also very interesting. But is starting to sound more like the ram system, and not only the ram chips.
But I am still willing pull 1 or more ram chips off the board. I'd like to have some idea of which one(s) to pull and see what happens with other ram chips
What do you think?

I guess try replacing any one between 7 and 1. If that bit disappears from the diagnostic's list, you then assume that the other six RAM chips are bad as well. If not, more diagnosis required.

 

[modem7]

I do not know if you have had a chance yet to replace one of the indicated failing bits, but something interesting turned up.

That is a lot of bits, and interesting that bit 0 always shows as good.

A recent post about the Supersoft ROM running on an IBM 5170 that has no RAM at all fitted, showed similar: bit 0 passing, and bits 1 though 7 failing.
Photo at [here].

On my IBM 5160, I repeated the no-RAM exercise (using the 5150/5160 version of the Supersoft ROM), and got the same result: bit 0 passing, and bits 1 though 7 failing.
Photo at [here].

But is starting to sound more like the ram system, and not only the ram chips.

Yes it does look that way.