From what I have gleaned of the tests:
16K CRITICAL MEMORY REGION
For each address within the first 16K, write a byte to the address, wait for I/O delays, then read the byte straight back.
By reading the byte straight back, this test does not rely on the dynamic RAM refresh circuitry working.
MEMORY REFRESH test
A test of the ability of the dynamic RAM subsystem to hold its contents 'long term' (10 seconds), when the refresh circuitry is operated at the normal refresh spec of about
15 microseconds per row (what the IBM BIOS operates the refresh circuitry at).
This test uses the first 16K or RAM, and so if the earlier 16K CRITICAL MEMORY REGION failed, I see that invalidating the results of this MEMORY REFRESH test.
SLOW REFRESH test
This is the MEMORY REFRESH test, excepting that the refresh circuitry is operated at well beyond its refresh spec. In most cases, the RAM chips used in 5150/5160 motherboards will tolerate that. If this was the only (repeat: only) test failing on one of my motherboards, then I would say to myself,
"All of my RAM chips are within spec (which is what the IBM BIOS runs them at), but one or more cannot be pushed significantly beyond spec. Maybe I will, or maybe I will not, do anything about it."
If the earlier MEMORY REFRESH test is also failing (which runs the refresh circuitry within the RAM's refresh spec), it is expected that the SLOW REFRESH test will also fail. [If I cannot throw a ball 5 meters, why would I be able to throw it 10 meters?] I am puzzled as to why the authors of the Supersoft diagnostics run the SLOW REFRESH test if the MEMORY REFRESH ('normal' refresh) test fails.
Initially, I only populated bank 0 and received a "16K Critical Memory Region" again (only at address 05080) but, this time it successfully passed the "System Memory to 40000" test but, failed the "Slow Refresh to 40000" test.
Is that consistent - identical symptoms whenever only bank 0 is populated ?
Or do you get different results each time ?
I then populated Bank 1 as well and this time it passed the "16K Critical Memory Region" but, failed the "Memory Refresh" test and, then again the "Slow Refresh to 80000" test (since I've now populated both Bank 0 & Bank 1).
Is that consistent - identical symptoms whenever only banks 0 and 1 are both populated ?
Or do you get different results each time ?
According to the manual, a "Memory Refresh" failure could indicate a problem with the 8253 Timer channel one or the 8237 DMA controller channel 0.
Since both of these passed the initial test (before reaching the "16K Critical Memory Region" test), which one of the two is the likely cause or, should I be looking elsewhere for the fault?
Ignore the SLOW REFRESH test for now (because it pushes the system out of spec).
You have put known good RAM chips into the system. If you were
always seeing the 16K CRITICAL MEMORY REGION test passing, and
always seeing the MEMORY REFRESH test failing, then I would see that as an indicator that the refresh circuitry is faulty.
You are (sometimes) seeing failure of the 16K CRITICAL MEMORY REGION test. It suggests to me that there is something else going on. If the symptoms are inconsistent, some things that come to mind are:
* Bad RAM chips (which you have disproved)
* Bad PSU (which you have disproved)
* RAM chip sockets that have oxidised contacts
* Bad RAM support circuitry: filter cap/s next to the RAM chip sockets
* Bad RAM support circuitry: other