rasz_pl wrote on 2026-06-11, 10:35:

Weird chipset choice to map this ram to a window at 2000:0 (128KB) instead somewhere at the top. Makes EMS emulation very slow (copying data around).

Yeah, it's essentially a limulator except using this oddball memory instead of XMS or disk.
Incidentally, XMS could be another way to expose the 384K.

Babasha, I've heard of these Juko 128K-512K bank switching memory implementations before. I've thought about whether you have to keep interrupts off while having the opposite memory bank switched in. If the main program is aware of the bank switching, I suppose not, unless there are already drivers or TSRs loaded there.

jakethompson1 wrote on 2026-06-12, 19:27:

Yeah, it's essentially a limulator except using this oddball memory instead of XMS or disk. Incidentally, XMS could be another w […]
Show full quote

rasz_pl wrote on 2026-06-11, 10:35:

Weird chipset choice to map this ram to a window at 2000:0 (128KB) instead somewhere at the top. Makes EMS emulation very slow (copying data around).

Yeah, it's essentially a limulator except using this oddball memory instead of XMS or disk.
Incidentally, XMS could be another way to expose the 384K.

Babasha, I've heard of these Juko 128K-512K bank switching memory implementations before. I've thought about whether you have to keep interrupts off while having the opposite memory bank switched in. If the main program is aware of the bank switching, I suppose not, unless there are already drivers or TSRs loaded there.

Its better to look at RAM-drive diver disassembled and commented by Indrekis (just translate it)
https://indrekis.github.io/retrocomputing/ibm … %8F%D1%82%D1%8C

rasz_pl wrote on 2026-06-11, 10:35:

Weird chipset choice to map this ram to a window at 2000:0 (128KB) instead somewhere at the top. Makes EMS emulation very slow (copying data around).

Love indrekis blog, had to resort to google translate to read but it was worth it https://indrekis-github-io.translate.goog/ret … &_x_tr_pto=wapp
all blog posts https://indrekis-github-io.translate.goog/pos … &_x_tr_pto=wapp

Indrekis found one BIG bug in EMS driver - in some circumstates stack can be lost during memory switch (in DOS its not so frequent while Windows 3.0 freezes more frequent). I hope in next version 1.3(?) get more stability and compatibility)))

Babasha wrote on 2026-06-12, 19:22:

Next step. Case for my rig.

Little baby or Big brother what you choose?

An XT shouldn't exist without a 5.25" 360KB floppy!

About your SCSI Controller. You know this thread?
Re: Making Future Domain TMC-850MEX SCSI controller bootable

eisapc wrote on 2026-06-15, 10:48:

About your SCSI Controller. You know this thread?
Re: Making Future Domain TMC-850MEX SCSI controller bootable

Tnx!

I know))) Currently i dont ready to solder it and it was so interesting use BIOS EXTENSION socket on motherboard

Some news about shadow RAM usage (part 1)
After bug-hinting by Indrekis and help of GlaBIOS author the EMS driver is way more stable with JUKO ST-12 and SCSI card/HDD
Now it boots and work in Windows 3.0 + Word 1.0

Some news about shadow RAM usage (part 1)
Use shadow RAM as a RAM-disk + benchmarks.

1. SCSI HDD speed

2. Original JUKO RAM-disk driver (direct mode)

3. Universal EMS RAM-disk driver