Could it be switch 3 & 4 ?
http://www.minuszerodegrees.net/5160/misc/516 … ch_settings.htm
Switches 3 and 4: Enabled motherboard RAM.

3=ON , 4=ON : Enable only bank 0
3=OFF, 4=ON : Enable only banks 0/1
3=ON , 4=OFF: Enable only banks 0/1/2
3=OFF, 4=OFF: Enable banks 0/1/2/3

also, how the 256kbit chips neeed to be refreshed could an issue.

I already looked at the DIP switches. They looked correct. (Even setting both 3 & 4 to ON gave me just 64kb of memory, which is exactly 1/4 of what the original count should be.)

I've also taken the refreshing into account as well (sort of). I adjusted the A8 delay line to make sure that it matched the original IBM specs in the schematics, but it didn't seem to do anything.

Also, the chips may not have to be refreshed in a certain way. Like I said, the system counts up to its normal count of 256kBYTEs, yet neither type group of memory individually adds up to 256kBYTEs. This tells me that the 256kBIT memory is being accessed in a similar way to the 64kBIT, without the A8 line being activated. Without that A8 line, it might as well all be 64kBIT memory.

I have an extra theory to tack on...

Would the BIOS have something to do with the memory not being supported? (i.e. an early BIOS that only supports 64-256kb?)

Your motherboard is just using the 256k chips as 64k chips.

http://minuszerodegrees.net/5160/motherboard/ … d_revisions.htm

It is possible to upgrade a 256KB 5160 motherboard to support 640KB. An easier alternative is a memory expansion card like the Lo-tech 1MB card.

I had thought about that card, but I am so close to getting this to work. Partly why I want to put 640k on the motherboard is because I have limited space for cards, only two slots, one of which is being taken up by an XT-IDE drive, so I can use the free slot for whatever card I want.

I'm going to go down the A8 line to check if there's a disconnect somewhere.

Have you got photos of your clone board? Plasma's link says IBM had two versions of the motherboard, with one only handling up to 256kB and the other up to 640kB, with a possible mod to convert the 256kB to 640kB by adding a jumper to E2.

This is the clone system I have, a Sanyo MBC-775.

I was able to replace the existing memory with sockets, and instead of the original v2.33 semi-IBM-compatible BIOS, I'm using one of the Yangtech Turbo XT BIOSes from minuszerodegrees. (Although I still have the original BIOS chip)

There are what look like 4 or 5 jumpers, and one of them I've been able to identify as a grounding jumper (5160 Equivalent E1)

Have you traced where the equivalent of U84 is? On the 5160 it's a 74LS158, like your U47 and U46. Pin 4 of that on the 5160 connects to A8 of the 256kb chips. The other chip in the mod is U44, given on the 5160 schematic as a 24510, but on at least one photo is a N82S129 bipolar PROM. Can't see it on yours, but it's probably socketed (what's U45?). On the 5160, pin 1 of that needs to be grounded, which that board does through E2. This assumes the equivalent on your board is programmed the same way as the IBM one.

I already looked at this board as thoroughly as I can without burning my mind out. 😜

When I first got it, this system was actually missing several ICs needed for what I assume was extra memory decoding. I was able to get them through Mouser. (An S280 parity Generator, an extra 158, and a 139)

Also, this system doesn't use a PROM. It uses the equivalent in 244, 245, and 373 logic chips (among others), which is why the board is so heavily populated. It'll be more difficult to put this together.

BTW if you are upgrading the BIOS you may be interested in this Turbo XT BIOS.

I saw that too. 😜 I may just do that.

T-Squared wrote on 2022-01-15, 23:58:

This is the clone system I have, a Sanyo MBC-775. […]
Show full quote

This is the clone system I have, a Sanyo MBC-775.

20220115_175049.jpg

I was able to replace the existing memory with sockets, and instead of the original v2.33 semi-IBM-compatible BIOS, I'm using one of the Yangtech Turbo XT BIOSes from minuszerodegrees. (Although I still have the original BIOS chip)

There are what look like 4 or 5 jumpers, and one of them I've been able to identify as a grounding jumper (5160 Equivalent E1)

Thanks for the picture ! I like to look over old boards. Unfortunately have no helpful information.

All right, good news and bad news.

Good News: I was able to get that Turbo XT BIOS working.
Bad News: The memory problem still isn't fixed, and it's not the fault of the BIOS (i.e. It's not a BIOS mixup, like one that can only count up 256kbytes of memory.)

Okay, looking at everything else, the lines and logic seem connected. (I didn't have to re-map everything.)

From this point forward, everything now seems to involve the 244 and 373 logic equivalent of that 24S10 PROM used in the 5160, because the connections from A16-A19 on the CPU side, and the connections from A8 on the memory side all meet up there. Strangely, the connection to the 24S10-equivalent logic also seems to go off of the CPU/MEM board and onto the video/ISA/Disk controller board. (There are two boards, the CPU/MEM board is at a right angle to the vid/ISA/disk board; apparently they didn't have enough room for the extra logic chips.)

Will an IBM 51X0 system count up only to the amount of RAM that it is able to address?

To clarify, if one address line was severed or not connected, would that drop the amount of memory detectable by the system (like the cheap 16kbit memory, which was actually faulty 32kbit, but still worked) that Sinclair used in the 1980s, or would that break the system entirely? This is assuming that there's some kind of POST failsafe in the BIOS that drops the amount of addressable memory to a power-of-2 or divisible amount and ignores the faulty RAM beyond that point. (i.e. 512kbyte of RAM, but 128k is not properly addressable [maybe 16k does not work], so it drops to 384kbytes.) Of course, I may just be spitting smoke.

I seem to have found an address line (A17) that may not be connected to anything, and may be my problem. (It's not connected to any 37x, 24x, or 138 logic chips.) I'm not sure yet, though. I'll have to continue probing with the multimeter in order to find out.

The BIOS determines the memory size by writing patterns and attempting to read it back. Once the read no longer matches the write (or the 640K limit is reached), it stops counting and that's your memory size.

Yep. I may have a disconnected address line that was never attached at the factory, for some reason. I'll take a look and see what I should do, in the morning.

Ok, looking at it with a logic probe, it seems that A8 is being kept low, but XA16 and XA17 are pulsing. (The latter was the one I thought was disconnected)

So what does this mean?

Dead LS158 somewhere? That 5160 manual has an LS158 the drives A8 with either XA16 or XA17, selected by ADDRSEL. So if both have a signal, but A8 is always low then a problem with 158? Assuming that chip exists on your board, pin 15 is an enable pin that should be low to enable to outputs, maybe it's stuck disabled? Schematic also shows a resistor in series between the output and the RAM, so maybe that's gone open?

Although, A8 does have a pull up on it, so if it's reading low then it must be being driven, which maybe points back to the 158 being disabled.

[edit: added schematic snippet, U84 sits between XA16 or XA17 and A8. Maybe U46 or U47 on your board?]