86Box PC1982 GLaBIOS with 192K shows only 176K

[Torinde]

I don't know if that's 86Box problem, GLaBIOS problem or user problem:

Happens also with V20:

[640-KB]

What kind of memory expansion card do you have configured in 86Box and what are the settings on it?

[Torinde]

None, only the base memory shown.

[640-KB]

If you change to the IBM 10/27/82 ROM, what do you see?

[Torinde]

Interestingly, with 128K and 64K - there is no discrepancy between dropdown, POST, CHKDSK in all three BIOSes. Where per the BIOS bug description there should be discrepancy as on 192K.

[640-KB]

Yeah, that's what I got too when using the IBM ROM - the RAM error, except when I tried booting with DOS 3.3 it would freeze. I suspect that in this state it really only emulates 176K, but is emulating the memory size DIP switches to be 192K, which is what IBM is saying there. GLaBIOS reads the 5150 memory size switches as the maximum RAM to allow, but won't error if the actual installed RAM is less - will just stop there.

I was able to reproduce this in 86Box by:

1. Edit the machine and type 176K in the Memory box, hit OK.
2. Edit machine type again and memory now says 128K (176K is not a possible size to be installed on the motherboard).
3. Hit the up arrow on the Memory which changes the box to say 192KB. Click OK.
4. Now the functional memory is actually 176K but it says 192KB in the box. Boot with either BIOS and you only get 176K of working RAM, even though the DIP switches reflect 192K.

So yeah, I'd say this is technically an 86Box bug (albeit one very obscure) based on invalid input since memory can only be installed in minimum sizes of 32KB blocks, and motherboard RAM in 64KB blocks (*only the first 64K can be 16KB increments on the original 16-64K board), so 176K would never be possible on hardware nor was the 5150 BIOS designed for it. See diagram.

[Torinde]

OK, so I made a ticket there.

What do you think the result should be with IBM BIOS 5150_27OCT82 - per what I read at this bug report:

• at 192K it should show 48K (as 86Box does now),
• at 128K it should show 32K (but 86Box shows 128K),
• at 64K it should show 16K (but 86Box shows 64K).

Do you think that's another bug in 86Box (even more obscure)?