IBM PC/AT with EMEA

[lyonadmiral]

I'm trying to decipher this error message when running advanced diagnostics:

REFRESH FAILURE

200002 0002

18:03:14
error -
12928KB MEMORY 203

What module on the EMEA could it be?

 

[modem7]

I'm trying to decipher this error message when running advanced diagnostics:

Which is the case:
* Your IBM 5170 was running fine, then all of a sudden, developed a problem.
* You are in the process of modifying the configuration of your IBM 5170.

If the latter, that increases the range of possibilities. (Incompatibility, conflict, misconfiguration, etc.)

REFRESH FAILURE
200002 0002

200000 = address 2 MB
0002 = bit 1

Presumably, you only have one card that provides RAM above the 1 MB address mark, that being the EMEA, the card described at [here].

The fact that only one bit is shown in error highly suggests a faulty SIMM.

error -
12928KB MEMORY 203

A second error, or somehow tied in with the first !
Again, above the 1 MB address mark.
203 is listed at [here] as a "Memory Address Error".

What module on the EMEA could it be?

I could not see anything in the 'Hardware Maintenance Service' manual for the IBM 5170.

Hopefully, the problem cause is a SIMM, and not the card circuitry.

Worth a try. If you haven't already:
1. Re-seat the EMEA in its ISA slot.
2. On the EMEA, re-seat all of the SIMM's.

The EMEA does not need to be fully SIMM populated, and so, and so if we presume a faulty SIMM, a strategy becomes something like:

1. Have the minimum amount of SIMM's in the EMEA, which according to what I am reading, is two, in the upper most slots. (You should confirm that.)
2. Run the diagnostic.
3. LOOP: If the diagnostic reports no error, add two more SIMM's. If an error is indicated, maybe one of the two SIMM's just added is the problem cause.

 

[lyonadmiral]

They're the modified 1 meg SIMM's so they work in the EMEA; I ran through and tested all SIMM's two at a time. Each set comes up with a different refresh error but not enough of a problem to stop booting.

 

[modem7]

Which is the case:
* Your EMEA equipped IBM 5170 was running fine, then all of a sudden, developed a problem, triggering you to run the diagnostics.
* You have yet to get your EMEA to a working state. You are running the diagnostics as part of diagnosis.
* The IBM 5170 appears to be okay. I just decided to run the diagnostics.

 

[modem7]

Notice how the bit-in-error changes:

0002 = bit 1
0004 = bit 2
0800 = bit 11
1000 = bit 12

I think at least four of your SIMM's each has a single-bit fault.

If we look at 0800 for example. The 08 portion is the HIGH byte, and the 00 is the LOW byte. The photo at [here] informs me that the HIGH byte corresponds to the SIMM socket on the left. And so the SIMM in the left socket has the bad bit (80 = bit 7).

 

[modem7]

Each set comes up with a different refresh error but not enough of a problem to stop booting.

All of the error addresses are above the 1 MB address. You may not (yet) be running anything that uses RAM above the 1 MB address.

 

[lyonadmiral]

Which is the case:
* Your EMEA equipped IBM 5170 was running fine, then all of a sudden, developed a problem, triggering you to run the diagnostics.
* You have yet to get your EMEA to a working state. You are running the diagnostics as part of diagnosis.
* The IBM 5170 appears to be okay. I just decided to run the diagnostics.

I sent some modules to the UK to be modded so they would work in the EMEA; since actual 1 meg modules I haven't been able to find, only the 512k and 256k modules. They came back, I put them in, the system counts up all the way, and nothing triggers a problem that prevents the post, the error only occurs when running the memory test in advanced diagnostics.

 

[modem7]

I sent some modules to the UK to be modded so they would work in the EMEA; since actual 1 meg modules I haven't been able to find, only the 512k and 256k modules.

To turn them into the 'IBM proprietary' versions.

They came back, I put them in, the system counts up all the way, and nothing triggers a problem that prevents the post, the error only occurs when running the memory test in advanced diagnostics.

Background:

In Rudd's Diagnostic ROM (RDR), and the Supersoft/Landmark Diagnostic ROM (SLDR) , is a test known as 'slow refresh'. It appears to be designed to detect RAM chips that pass the memory test in the POST but are perhaps, a little 'weaker-than-normal' in holding charge in one or more cells. Such RAM chips might be problematic. You should read page 34 of the SLDR's user manual (at [here]). So, if I was seeing intermittent RAM problems in one of my computers, replacing any RAM chips that fail the slow refresh test might be the fix.

BTW. After adding a 'slow refresh' test to RDR, I then tested the new version of RDR against a few IBM 5150 and 5160 motherboards. The slow refresh test failed a 4164 RAM chip, and I put the 4164 aside as a means of regression testing the slow refresh test in future version of RDR. But I did not need to do that. The motherboard had given me no problem whatsoever. If had not run a slow refresh test on the motherboard, I would be 'none the wiser'.

IBM Advanced Diagnostics

Could it be that the 'REFRESH FAILURE' error message results from a slow refresh test that the IBM Advanced Diagnostics does ?

 

[modem7]

Further to last:

Regarding the screenshots in post #3.

The KB figure at the bottom matches the 'TESTING' xxxxxKB figure at the top.

'REFRESH FAILURE' is I think, a user-friendly message.

The 'xxxxxKB MEMORY 203' line appears to be a kind of summary: a 203 error occurred when testing xxxxx KB of RAM.
As for the 203:
In the PS/2 error code list at [here], a 203 is described as '203 (PS/2) Memory address error or refresh failure'.
So, in some contexts, 203 has two possible meanings, and in your case, it means 'refresh failure'.

 

[mikeyp]

Hi both

Some info that may help.

I followed the instructions here to modify the modules. http://john.ccac.rwth-aachen.de:8000/misc/ps2cache/

They were fully tested in a retro chip tester pro (RAM and PARITY tests) and POST tested in my AT. My EMEA card is faulty and can only use 4MB RAM currently. I don't have time to figure it out and repair it at the moment.

Lyonadmiral has confirmed that with OEM IBM 256k and 512k modules, he doesn't get the refresh failure. I didn't run IBM advanced diag against my own RAM or lyonadmiral's. It didn't occur that there would be such issues beyond POST tests. I will try this myself and report back. I will also try the above suggestions.

I don't know enough to be able to say what is causing this failure. It's possible the memory we have used isn't of suitable spec. If it's that there's another connection or trace cut required that isn't documented, it's going to be near impossible to track down unless anyone else can weigh in with the required knowledge.

 

[modem7]

They were fully tested in a retro chip tester pro (RAM and PARITY tests) and POST tested in my AT.

We know that the POST does a rudimentary RAM test only. And I have not looked into the Retro Chip Tester Pro to see how extensive its RAM tests are.

I think the big question is, is this worth worrying about in a hobbyist environment if the RAM is usable? If (repeat: if) the subject test is a 'slow refresh' one, then I would say, no. For me, it would different if this was a computer used in a non-hobbyist environment.