Subscribe via RSS
18Nov/260

Sharp AX386 HD

This unit appeared on Yahoo Auctions and I was mildly taken aback. I thought I was looking at an X68000, but realised quickly it was chunkier. There's an AT 386 in there, styled like an X68k! Yes please! Prior to instigating buyers-regret, I googled around to see if others had any of these units... just to know what I'd need to look out for, but there wasn't that much information. Here are the original auction photos:

The googlin' continued and I stumbled across a German site with information, but it just mentioned that this machine was Sharp's foray into 100% compatible x86s. The design seems to have won a Good design award, but that's cheating as the design was already well 'designed' via the X68000. Seems someone has already produced a 3D Model of it? Cute! So there's a cult following... There's also a hint of a screen resolution of 1120×768 so that Kanji can be displayed correctly... That could be fun/problematic!

It was initially offered at a high price, but failed to attract a bidder. It was then re-offered at a heavily-reduced starting price, and so I jumped... and after a mild bidding war, it was packed and sent! Buyee did a great job here. I've never seen expanding-foam in plastic bags used for 'filler' packaging, but it worked perfectly.

After unpacking, the dirt on the side of the case was obvious. I went back to the auction pictures and noticed that the seller chose NOT to present this side of the case in any of the photos. Hah. Nice. Cleaning it wasn't easy and a bit of paint was removed... but we were now ready to see it work.

Release the Electrons

The seller had already powered it up and shown the green LEDs on the auction listing, so I went for it. It powered on nicely displaying the Trident BIOS and then the usual warining of a dead CMOS battery and no keyboard. It needed an AT and ... somehow... I'd misplaced all my PS/2 to AT adapters!

Learning something new every day, I found out that AT and PS/2 are identical and it's just a form-factor thing? I quickly rigged up an adapter and got past the warnings. It initially only showed 4mb of RAM and no HDD? I opened up the case and found a dangling MFM cable hanging from a 16-bit MFM Controller Card! Woah. So this AX386 "HD" isn't quite what it proclaims to be. It also seems that when they removed the HDD, they also took whatever bracket/sled held it in. Something for the 3D printer.

I suppose, looking at the auction photo above, you can see that the MFM cables run south and there's another grey ribbon cable heading into the HDD area? Knowing now how big the case is, the HDD would've actually already protruded from the rear of that support.

Further to this, another sign of monkeying-around! In the back of that picture you can see a four-pin header dangling. Actually... there's a closer-up auction shot:

The plug-header has an orange and grey wire coming out of it. Turns out that's the HDD activity light on the front of the case and it's meant to be plugged into the rear header of the MFM controller card. If you then look at the rear-case-shot at the top of the article, you'll also see that ISA Slot 1 is labelled with (HD) after it. So this card is specific to this machine and it's meant to be in Slot 1. I moved it as-such and re-connected the activity light.

Anyway, I digress. Let's get an XT-IDE in there. It booted, but the floppy drives fail badly... The system contains two Epson SMD-400 drives on a straight-through ribbon cable. The internet seems to think these are 720kb and common to the Atari/Amiga realm, but these two are non-standard. They're actually SMD-440L (as per the finer-print) and therefore 1.44mb in capacity and requiring drive-select! They both have DS0/1/2/3 jumpers on the side.

They're also missing Pin 2 on their headers? Can I use a Gotek or standard floppy drive? Look at the cable.. it's hard-keyed:

Anyway, let's try repair them first.

Capacitors

Long-story-short, both drives were extracted and inspected. All electrolytic caps on both drives were toast. I replaced with ceramic 10ufs, but neither drive wanted to work. Drive A just kept spinning at full-tilt and Drive B formatted disk without ever wanting to read/write afterwards.

I relegated them to be glorified eject-button holders... removing them entirely would cause the case to look incomplete.

Using PC Drives

So, if I want to use PC drives instead, I need to make sure the pin-out is correct, and that the drives can be recognised as A and B. With the existing drives, this is done by the drive-select jumpers. Unfortunately, PC floppy drives are locked to drive-select 1 and will always show up as Drive B:. PC floppy drives expect a twist in the data ribbon cable to choose between drives A and B.

Fortunately, this is the internet and someone has almost always done this already. Turns out that some older floppy drives have markings on the PCB showing how to reconfigure their drive-select settings!

I opened one of mine and the DS jumper (aka 0 ohm resistor) was staring me in the face. I quickly shifted it over to DS0 and I had a working, formattable, bootable Drive A! Let's try OS/2...

Hah.. that actually worked... but it was slow as molasses.

Ominous Signs...

Whilst doing all this, the machine would sometimes reboot into monochrome. This was weird, but didn't prevent testing. Other times I'd get a weird gradient display and then other times it'd just not-boot or freeze. I'm thinking it's a dodgy connection in the power switch, but it could be deeper and more sinister. I'll have to dismantle the unit when I next have 2 hours and see what's going on.

On top of weirdness, the BIOS was also always throwing Checksum error on Extended CMOS or Hard disk failed (when trying to use MFM), etc... After seeing how slow OS/2 was, I decided to give the project a rest.

The machine was put on the top shelf and forgotten about...

Reviving it... Properly...

If you hate AI... then stop reading. Bring on 2026 (I first got this machine 2024!) and I'd started spring cleaning. This machine had been staring at me (taunting me?) from the top shelf of the rack and I decided to throw it up on Facebook Marketplace for a song.

A recent buyer, of other things I'd listed on FB, asked about getting a deal on it and that got me thinking: I'd already described all the faults in the listing, but had I really tried hard enough to get ol' betsy going? I've been pretty lazy in the latter half of this year and yeah, I think we already know the answer... Just look at the post-count recently on this blog? Zero-work-done.

So. What's been the cause of this malaise? AI. Let's not beat around the bush. What's the point of working on this old stuff if the 'future' needs 9999TB of RAM... and this thing has 9MB? (I must admit that that's a lot or RAM for a 386). Then again... what's the answer to fixing all of this? AI!? Oh the humanity. Won't they think of the children... etc.

I plugged it all back together, and tried to use all of the BIOS SETUP executables found in the Collection of System Utilities for BIOS Setup on Internet Archive. They all had mild success... but nothing correctly or completely configured all settings required by the BIOS.

My first goal, in this fresh effort to work on the machine, was to disable the second floppy drive. I was only running with a single newer floppy drive on the ribbon and wanted to simply get rid of one BIOS error message successfully: Drive 2 not responding. Gemini kept telling me to use shortcut keys like CTRL-ALT-S or CTRL-ALT-INS, but this BIOS does not have a built-in configuration tool. None of the setup applications above could configure this second drive to be Not Installed, as much as they tried! Instead, I went the reversing-route. In the BIOS Tools archive above is BIOSIMG which seems to be some support tool where you send some company your BIOS image for preservation or help? No idea... either way, it was used to dump the file to SHARP.BIO. I then fed this to Gemini... wanting to know if there were actually any key commands...

sharp.bio is a bios image from my sharp AX386HD computer. I can't get into the BIOS. can you please tell me if there is a magic key combination to press?
There is no magic key combination to enter BIOS Setup—SHARP.BIO does not contain a built-in ROM Setup
utility.
--------------------------------------------------------------------------------
What's inside SHARP.BIO
--------------------------------------------------------------------------------
• BIOS: 80386 ROM BIOS PLUS Version 1.10 03 (Phoenix Technologies Ltd., copyright 1985–1988, BIOS date 01/15/88 at 0xFFF5, with 0077//0066//8899 at 0xFFDC).
• No ROM Setup: There are no Setup menu strings or UI routines anywhere in the 64 KB image (nearly half of the ROM—0x5800–0x9806 and 0xB807–0xDFFF—is empty 0xFF padding, and 0x9807–0xB806 is 0xCC filler). CMOS port 71h writes (OUT 71h) only occur during POST initialization.
• External Setup Required: Like many 1988-era Phoenix 386 BIOSes, configuration is handled by an external DOS/floppy SETUP program, which is why the ROM's own error messages say:
○ Invalid configuration information - please run SETUP program (0x0B9A, 0x4010)
○ Time-of-day not set - please run SETUP program (0x0C32)
--------------------------------------------------------------------------------
Key Combinations That Do Exist in SHARP.BIO
--------------------------------------------------------------------------------
Inspecting the INT 09h keyboard interrupt handler (at 0x29B7, jumped to from 0xE987) and POST code shows only these special keystrokes:
• Ctrl + Alt + Del (0x2AA8): Warm reboot.
• Ctrl + Alt + F1 / Ctrl + Alt + F2 (0x2A16 / 0x2A27): Clears/sets mode flags at BIOS Data Area 0040:00E4 (bit 6) and 0040:00EB (bit 1) (AX/keyboard mode switching).
• Ctrl + Alt + Keypad + / Ctrl + Alt + Keypad - (0x2AA0 → 0x5421): Adjusts CPU speed/wait-state setting via 0040:0015 and chipset ports 22h/23h.
• Space (0x1914): Terminates the POST memory test early.
• F1 (0x1DF8 / 0x1E19): Retries boot or continues past a POST error.
--------------------------------------------------------------------------------
What to do instead
--------------------------------------------------------------------------------
To configure CMOS settings (date/time, floppy/hard drive types, memory), you'll need to boot from DOS (or a Sharp AX386HD system/diagnostics diskette) and run the external SETUP utility—or a compatible generic PC/AT CMOS setup tool from DOS (such as GSETUP or SETUP386).

The goal was to get rid of the error in extended CMOS, so I asked it what was stored there:

can you work out what the bios reads from the "extended" area? what settings are saved there?

It actually returned a table:

CMOS/Reg What it stores Where in SHARP.ASM
0x40 Bit 3 (0x08) = Extended CMOS checksum error flag. Set to 0x08 when the 0x40–0x7D checksum fails; checked during POST to print Checksum error on extended CMOS. 0x07B4, 0x0F96
0x4C / 0x4D Floppy mode / AX FDD control. 0x4D (masked with 0x7F) is written to hardware port 0x7D (with 0x01 written to port 0x7C), and selects how 0x68/0x69 are mapped into standard CMOS 0x10 (floppy drive types). 0x1B13–0x1BB0, 0x3F52
0x4E Keyboard hotkey disable flags (checked in INT 09h):
• Bit 0 (0x01): Disable Pause
• Bit 1 (0x02): Disable Print Screen
• Bit 2 (0x04): Disable Ctrl+Alt+Del
0x2AAC, 0x2CE2, 0x2DD5
0x4F Chipset config (bits 0–1). Defaulted to 0x01 on checksum error; bits 0–1 are written to chipset index 0x06 via ports 0x22/0x23. 0x0F88, 0x0FA0
0x68 / 0x69 Floppy drive type pairs (high nibble = Drive A, low nibble = Drive B). 0x68 accepts 3 or 4 (defaults to 0x44 on checksum error); 0x69 accepts 0–4. Used to populate standard CMOS 0x10. 0x0F8F, 0x1B45–0x1B79
0x7E / 0x7F Extended CMOS checksum. 16-bit big-endian sum of 0x40–0x7D (0x7E = high byte, 0x7F = low byte). 0x0F3F–0x0F61, 0x1C03–0x1C3E

All other bytes in 0x40–0x7D are simply zeroed on checksum failure and included in the 0x7E–0x7F sum, without being individually read by SHARP.BIO.

Ok, this is exciting! Somehow Disk 0 and 1 are actually configured in extended BIOS? Not the regular area? No wonder the generic setup tools don't work. I therefore asked it to build a tool which'll let me set those params in the extended area. It would then also have to calculate and store the extended area checksum.

can you write a short c program which can set these values and calculate the checksum for the extended area?

Would you believe it worked? It produced this C source... which I then compiled with Open Watcom C Compiler (open-watcom-2_0-c-win-x64.exe). It told me to run EXTCMOS 4C=04 4D=04 69=40 which would set the configuration and stored the checksum. No more extended BIOS area checksum errors!

IMG 20261118 185800926 HDR

But, a new error: Disk 2? What? I only have Disk 0 and 1... wait a minute...

Oh... look... (apologies for the re-use of this imagery) the motherboard actually has a header for floppy drives 2 and 3! This thing supports 4 floppy drives. I told Gemini this...

ohhhh... the floppy configuration in the ext bios area is for floppy drives C and D (to allow 2x external floppies?
Yes—you're spot on! Looking at lines 7425–7480 (0x3EFB–0x3F80) and loc_1B2F (0x1B2F–0x1BB0):
• Hardware FDD Multiplexer (0x7C / 0x7D): The Sharp AX386HD has a floppy multiplexer controlled by ports 0x7C/0x7D. During floppy POST, it writes 0x00 to 0x7D to test the internal drives (0 and 1), then tests external drive 2; if present, it writes 0x04 to 0x7D to switch to the external bank and tests drive 3.
• 0x68 = Internal drives (0 & 1): High nibble = internal 0, low nibble = internal 1 (only allows 3 = 720 KB or 4 = 1.44 MB).
• 0x69 = External drives (2 & 3): High nibble = external 2, low nibble = external 3 (0–4, where 0 = None).
• 0x4C / 0x4D = Bank / swap mode (written to port 0x7D):
○ 0x00: Internal drives (0x68) mapped as A:/B:
○ 0x04: External drives (0x69) mapped as A:/B:
○ 0x01: External 2 as A:, Internal 1 as B:
○ 0x05: Internal 0 as A:, External 2 as B:
So for one internal 1.44 MB drive and no external drives, 4C=00 4D=00 68=44 69=00 is the normal Sharp AX setting (and if you have an external 1.44 MB drive on the external port as A: with no B:, 4C=04 4D=04 69=40 maps external drive 2 as A:).
can we modify extcmos to control all 4 floppy flags. I want to turn off floppy drives 1,2,3 and just use floppy 0
Updated EXTCMOS.C to control all 4 floppy drives (FDD0–FDD3), update both standard (0x10, 0x14) and extended (0x68, 0x69, 0x4C, 0x4D) CMOS bytes, and recalculate both CMOS checksums (0x2E:0x2F and 0x7E:0x7F).
Rebuild with Open Watcom: wcl -bt=dos extcmos.c
Usage in DOS
• Enable only Floppy 0 (1.44 MB) and disable Floppies 1, 2, 3:
EXTCMOS single
• Or set all 4 drives explicitly (0=None, 1=360K, 2=1.2M, 3=720K, 4=1.44M):
EXTCMOS fdd=4,0,0,0

And so I compiled, copied over and set those to "0", along with the actual disk 1. WIN! But yeah.. the HDD was still failing. I had a hunch that the controller BIOS wasn't working, so I tried to dump it. Sorry. I didn't actually try... I asked Gemini to write an app to scan and dump the option ROMs. It did it...

Created DUMPROM.C to scan for and dump any ISA/video option ROMs in C000h–EF80h.
Compile with Open Watcom: wcl -bt=dos dumprom.c
What It Does
• Scans C000:0000 through EF80:0000 in 2 KB (0080h paragraph) steps for the 55 AA option ROM signature.
• Reads the ROM size from byte 2 (blocks * 512 bytes; defaults to 8 KB if 00).
• Verifies the 8-bit checksum (00 = OK).
• Saves each detected ROM to ROM_.BIN (for example, ROM_C000.BIN for video BIOS or ROM_C800.BIN for an MFM/storage card).

But yeah, it only found the Trident VGA card... which I suppose was to be expected. It then told me that the card should just-work(tm) and that the EPROM on the card is most-probably microcode for the on-card controller and not an actual option ROM. I believed it... and asked what else could be wrong, telling it that the error message was Hard disk failure...

IMG 20261118 185942568 HDR

Actually, before this I'd set random jumpers and got Hard disk controller failure, so I really should have already known that the controller was being init'd... just not the disk. Anyway, I asked it to extend EXTCMOS.C to describe the current BIOS settings (not trusting the previous SETUP programs) and it returned with:

IMG 20261118 185636514 HDR

I double-checked the BIOS HD0 type which was set to a Type 1 (10mb) for my 10mb drive .. and it was. I then used SpeedStor Storage Manager to init the disk. Despite finding a few bad sectors, the drive was recognised and ready to go. DOS 6.22 was then installed!

MOUSE.COM was then used from Microsoft Mouse 7.x (because this thing uses a BusMouse) and ... that also just-worked(tm).

I Want Twenty MegaBytes?

Another funny moment... I always say this is the internet... someone has already done this... and, I wonder if it's just Google tailoring my search results, but the first result for ANYDRIVE.EXE is my blog from 2017. I actually had to fix the links as they'd all died. I continued to check other links on other posts in this blog and NONE are doing WELL!

Anyway, I've installed a Rodime RO352 and somehow.. since the BIOS only supports 17 Sectors, thought that I had a 20mb drive as it wants 32 sectors via the manual. Turns out... who cares? It's still a 10mb drive. But Anydrive is good Yo! If you have a large drive. I was totally confused by a separate ST-225 drive that I'd also been testing with.

In the end, I think I need to re-add the XTIDE IDE Drive as a secondary device to have room to move. But that would throw me back into a put it back on the top rack stage.

Railroad Tycoon

This just-worked(tm).

IMG 20261118 192200531 HDR

I even installed a sound card for extra fun! Use UNISOUND instead of caring about sound drivers.

IMG 20261118 192019055 HDR

IMG 20261118 192056876 HDR IMG 20261118 192109753 HDR IMG 20261118 192117492 HDR

Of course... the mouse stopped working once the sound was enabled.

IMG 20261118 195322353 HDR

The cause? As partially mentioned above, they both wanted the same IRQ! Can we flip the loading order?

IMG 20261118 195445264 HDR

No.. well... yes... I mean, no errors? But no mouse! The answer....?

IMG 20261118 195807918 HDR

Use SET BLASTER to force your PNP card away from IRQ5 and everything will work fine! Of course, you'll need a Vibra16 sound card... or some other PNP thingy-ma-jig to use UNISOUND.

A Final Note On Video

This machine came from Japan with a standard Trident VGA card installed. It probably would've come from the factory with a JEGA or AX-VGA card installed, to allow the display of Japanese characters... but it seems that card is long-gone. Instead, you can install DOS/V for double-width characters if required.

Fortunately, the BIOS and all text is in English, so maybe that was also changed out?! We'll never know.

Filed under: Retro No Comments