PDA

View Full Version : Kaypro 16 memory problem



phogren
June 27th, 2014, 02:46 PM
i am trying to get a Kaypro 16 going.
I get the following error on boot up:

128kb Good
Mem Par NMI int, 02 Ram Bad 20000
128kb Good

It uses a Phoenix BIOS ver 1.07

It's a real pain to take the motherboard in and out for testing. I tried reseating the chips but no effect. They supposedly came with 256 ram upgradeable to 512. there are no empty ram sockets on motherboard.

suggestions?

Stone
June 27th, 2014, 03:00 PM
It appears that the parity chip in the third bank might be the culprit. Try replacing(switching) it with any chip from the fourth bank and see if the error message changes.

phogren
June 27th, 2014, 04:55 PM
doesn't seem to change anything. I tried switching and replacing (from another MB) and no change at all with the error message.
It has 8 banks of ram, could it be one of the other banks?

Stone
June 27th, 2014, 05:24 PM
Are you sure you changed the third bank and not the fifth bank? Are the banks labeled?

You might try replacing the parity chip in the second bank just to be sure.

modem7
June 27th, 2014, 08:14 PM
There is a Kaypro 16 technical manual at http://maben.homeip.net/static/S100/kaypro/systems/kaypro%2016/Kaypro%2016.pdf

From page 19-6 of that, 8 banks of 64 Kbit chips (8 x 64 = 512)
Therefore:

Bank 0 = 0 to 63 KB
Bank 1 = 64 to 127 KB
Bank 2 = 128 to 191 KB
Bank 3 = 192 to 255 KB
Bank 4 = 256 to 319 KB
Bank 5 = 320 to 383 KB
Bank 6 = 384 to 447 KB
Bank 7 = 448 to 511 KB

The "20000" in the error message will be the hex address of the error. That equates to the 128 KB address.
In my opinion, the "Mem Par NMI int" portion of the error message is indicating that a parity error NMI occured, something that will occur for any bad bit.
The "02" portion of the error message is sure to be showing the bit error pattern. 02 equates to the bit 1 chip.

So, again looking at page 19-6 of the technical manual, I think it is chip U20 (bit 1 within bank 2) that is faulty.

phogren
June 28th, 2014, 11:21 AM
each chip is identified 0-7 and P for the parity chip. I got so frustrated that I pulled banks 5-8 and got the 256kb working and then added 64 at a time. I had to change a couple of rows. Now I get successful memory check at startup to 512kb.
then it will boot off the floppy (hard disk gives a disk error) but I get the following:
Copyright (c) Microsoft Corp. 1981, etc etc
Compatibility Software
Copyright (c) by Phoenix Software Associates Ltd
Copyright (c) 1984 by Kaypro Corporation

Mem Par NMI Int CS:IP = 7f7f:00B5h
Type (S)hut off NMI, (R)eboot, other keys

The boot disk I am using is actually for the 16/2 which is the 2 floppy system without the hard drive.

modem7
June 28th, 2014, 06:42 PM
That reads to me that the power-up RAM test (usually a 'quick and dirty' thing) passed, but later, a RAM parity error occured when address 7F7F:00B5 was read.
7F7F:00B5 hex equates to 7F8A5 hex, which equates to 522405 dec, which equates to about 510 KB.
If you remove the last bank, does the machine operate successfully?

phogren
June 29th, 2014, 08:15 AM
Yes, if I remove the last bank I don't get the error.
Thank you and Stone very much.

Stone
June 29th, 2014, 08:24 AM
You might have one bad chip in that bank or it could be several. You can determine which chips are bad by replacing one chip at a time in bank 2 or 3 or 4 or any bank but the first with one of the chips from bank 8. When you see an error you will know that the replaced chip is bad and you can discard it. Then you'll know how many chips you will need. It might be just one. FWIW, it's always good to have spares as you never know when one of these old chips can fail.

H-A-L-9000
July 2nd, 2014, 08:30 AM
That reads to me that the power-up RAM test (usually a 'quick and dirty' thing) passed, but later, a RAM parity error occured when address 7F7F:00B5 was read.
7F7F:00B5 hex equates to 7F8A5 hex, which equates to 522405 dec, which equates to about 510 KB.
If you remove the last bank, does the machine operate successfully?

I'd read "The NMI occurred while the instruction at 7F7F:00B5 was executed." We don't know if it happened at the instruction fetch or when the instruction read/wrote memory. The NMI lets the CPU push the CS:IP on the stack which is all information on the error the architecture can provide.