I have several HT12 boards, is there a reason you picked this chipset?
Did you desolder one from a board or are they available from some source?

This project could solve availability issues.

Do you have a kicad/boardview file for this, that won't be a problem to upload? Could help repair existing motherboards.

Wow, I love this 😀

I bought a motherboard with a damaged battery. I repaired it. Unfortunately it didn't last long. That's why I decided to build a replica. The HT12 has been installed on many late 286 boards. It shouldn't be difficult to find defective boards.

I bought a tray of ICs from a chipbroaker.

Nelson68k wrote on 2024-12-16, 13:26:

After some failures the 286 board finally seems to run stable. The ram test runs without errors. I have already been able to tes […]
Show full quote

The attachment monitor.jpg is no longer available

The attachment mainboard.jpg is no longer available

After some failures the 286 board finally seems to run stable. The ram test runs without errors. I have already been able to test some extensions. Only the coprocessor test is still open.

The attachment blanck.jpg is no longer available

The attachment populatet.jpg is no longer available

The board runs with 20Mhz and 0 wait state very well so far.

The implementation was done with Kicad. I will put the files into the git repository at a later date. Just like with the 386 mainboard.

I really like this motherboard, I don't know if you replicated an existing board, or if you made a new one, however in any case, I congratulate you.

I hope you can solve all the problems, and that it works well, now you can't find cards like that anymore, and if you find one they cost a lot too much.

Update us punctually on the progress, and if I understood correctly, have you already done a 386?

PC@LIVE wrote on 2024-12-16, 15:00:

I really like this motherboard, I don't know if you replicated an existing board, or if you made a new one, however in any case, I congratulate you.

Thanks. The board is a replica. I have improved a few things. 4-layer instead of only 2. SIM modules can be used. Previously only SIP. I will improve a few small things. https://github.com/Marco-Both/A-one-286

PC@LIVE wrote on 2024-12-16, 15:00:

and if I understood correctly, have you already done a 386?

Yes, I have already made a 386 board. Re: PC CHIPS/FOXCON M396F ver. 2.6 https://github.com/Marco-Both/M396F-Replika And a graphics card. 8/16Bit VGA graphics card AVGA1 / GD5401 clone

Have you tried the MR Bios v1.65 for headland chips yet?

1Part Number   Code      CPU  Chipset                        I/O chip  Turbo
2V039B200      HT12200   286  HT12/HT12+                     None      
3V039B201      HT12201   286  HT12/HT12+                     None      Turbo, active state low
4V039B202      HT12202   286  HT12/HT12+                     None      Turbo, active state high
5V039B210      HT12210   286  HT12/HT12+ with built-in EMS   None      
6V039B211      HT12211   286  HT12/HT12+ with built-in EMS   None      Turbo, active state low
7V039B212      HT12211   286  HT12/HT12+ with built-in EMS   None      Turbo, active state high

Where can I find this MR Bios file?

douglar wrote on 2024-12-17, 14:11:

Have you tried the MR Bios v1.65 for headland chips yet?

I had been looking for it. But unfortunately found nothing. Do you have the files?

EDIT: I found something and am testing it.

Nexxen wrote on 2024-12-17, 14:23:

Where can I find this MR Bios file?

http://vogonsdrivers.com/index.php?catid=77

"MrBios Rom Bundle by jheronimus" has all of the BIOS files that have been found.

"MR BIOS archive sheet" lists all _documented_ BIOS files, but only the ones with a version number listed are in the jheronimus bundle. Unfortunately less than 1/4 of the documented rom images have been found so far.

That's an exciting looking board. I have a pair of 45ns 30 pin simms that I've been saving for a board like that.

Makes me wonder what else could be done to make an uber 286 board.

A PS/2 mouse header? A PLL for selecting clock speeds? An OPTi 82C281 build that has a cache ?

Nelson68k wrote on 2024-12-16, 21:58:

Thanks. The board is a replica. I have improved a few things. 4-layer instead of only 2. SIM modules can be used. Previously onl […]
Show full quote

PC@LIVE wrote on 2024-12-16, 15:00:

I really like this motherboard, I don't know if you replicated an existing board, or if you made a new one, however in any case, I congratulate you.

Thanks. The board is a replica. I have improved a few things. 4-layer instead of only 2. SIM modules can be used. Previously only SIP. I will improve a few small things. https://github.com/Marco-Both/A-one-286

PC@LIVE wrote on 2024-12-16, 15:00:

and if I understood correctly, have you already done a 386?

Yes, I have already made a 386 board. Re: PC CHIPS/FOXCON M396F ver. 2.6 https://github.com/Marco-Both/M396F-Replika And a graphics card. 8/16Bit VGA graphics card AVGA1 / GD5401 clone

Thank you very much ☺️

Your links are very interesting 🧐.

I also like the other boards a lot, and in fact maybe, in cases where the PCB is very ruined, from the battery loss, the best thing is probably to do as you did, transfer everything to a new PCB, but in reality I don't think it's that simple, maybe for PCs like 286 or 386 it might not be too difficult, to make a new PCB, even if among my motherboards to repair I have a MB 486 DX the size of the 386SX, here maybe that would be a nice future project, if you're interested I'll tell you the brand and the Model.

Nelson68k wrote on 2024-12-16, 21:58:

I have improved a few things. 4-layer instead of only 2. SIM modules can be used. Previously only SIP. I will improve a few small things. https://github.com/Marco-

4 layers instead of 2 is hardly an improvement. To me it is a serious drawback but it simplifies your routing task.

douglar wrote on 2024-12-17, 21:10:

That's an exciting looking board. I have a pair of 45ns 30 pin simms that I've been saving for a board like that.

Makes me wonder what else could be done to make an uber 286 board.

A PS/2 mouse header? A PLL for selecting clock speeds? An OPTi 82C281 build that has a cache ?

A ps/2 interposer would easily fit this board, but yes of course having it built-in would be much preferred 😀

This is fantastic c one of the best 286 chipsets! What was the per piece cost for the HT12 chipsets? What is the approximate cost for components and PCB to assemble a motherboard?

Have you tried to run the Harris at 25 MHz? There was a later Headland chip, the HT18C, which may improve overclocking room.

douglar wrote on 2024-12-17, 21:10:

Makes me wonder what else could be done to make an uber 286 board.

A PS/2 mouse header? A PLL for selecting clock speeds? An OPTi 82C281 build that has a cache ?

Integrated PS/2 mouse header would be a nice bonus.
I think leaving the DIP-14 socket for the crystal oscillator is sufficient. We have an adjustable PLL with MHz increment/decrement buttons that fit into DIP-14 (just waiting on a new PCB layout): download/file.php?id=204708&mode=view

feipoa wrote on 2024-12-18, 02:14:

Have you tried to run the Harris at 25 MHz? There was a later Headland chip, the HT18C, which may improve overclocking room.

While the HT18C does support 16MB of RAM, the CPU performance is somewhere between the middle of a 0WS and 1WS HT12 mobo with the same clock speed, with nothing in the BIOS to change the wait state (strings do exist in the BIOS ROM) - even worse, I've had to mark my Win 3.1 286 benchmarks that I've been working on with that HT18C mobo as obsolete, since Windsock 3.30 was reporting either 80 or 90 for the CPU speed when clocked at 24Mhz (as-is from the factory, there is an unpopulated DIP-14 socket for changing the ISA clock speed, otherwise it's 1/4th of the CPU oscillator - I decided to opt for not using the additional oscillator) with no apparent consistency though DOS benchmarks were unaffected, and it was clearly affecting the Windsock detailed GDI benchmarks scores on certain tests that have no bearing on what video card is used, regardless if they're accelerated or not. The HT12, running at 25Mhz with 0WS has consistent CPU scores of 93 or 94 in Windsock and those same specific tests have significantly higher scores - we're talking a ~20-30% improvement.

georgel wrote on 2024-12-17, 21:58:

4 layers instead of 2 is hardly an improvement.

it is for signal integrity

georgel wrote on 2024-12-17, 21:58:

To me it is a serious drawback but it simplifies your routing task.

additional $20 more for peace of mind and not needing to wonder why something crashes, where to apply corrections and keep respinning PCB in hopes its finally fixed.

DEAT wrote on 2024-12-18, 03:35:

feipoa wrote on 2024-12-18, 02:14:

Have you tried to run the Harris at 25 MHz? There was a later Headland chip, the HT18C, which may improve overclocking room.

While the HT18C does support 16MB of RAM, the CPU performance is somewhere between the middle of a 0WS and 1WS HT12 mobo with the same clock speed, with nothing in the BIOS to change the wait state (strings do exist in the BIOS ROM) - even worse, I've had to mark my Win 3.1 286 benchmarks that I've been working on with that HT18C mobo as obsolete, since Windsock 3.30 was reporting either 80 or 90 for the CPU speed when clocked at 24Mhz (as-is from the factory, there is an unpopulated DIP-14 socket for changing the ISA clock speed, otherwise it's 1/4th of the CPU oscillator - I decided to opt for not using the additional oscillator) with no apparent consistency though DOS benchmarks were unaffected, and it was clearly affecting the Windsock detailed GDI benchmarks scores on certain tests that have no bearing on what video card is used, regardless if they're accelerated or not. The HT12, running at 25Mhz with 0WS has consistent CPU scores of 93 or 94 in Windsock and those same specific tests have significantly higher scores - we're talking a ~20-30% improvement.

Reports in this thread mention full stability at 31 MHz with the HT18/C: Re: 3 (+3 more) retro battle stations
The ability to use 16 MB is a nice bonus if you want to run Windows 3.11. On my current 286 system, I need to use an ISA RAM extender card to get up to 7 MB DRAM, but running DRAM on the ISA bus offers slower overall performance compared to the motherboard's onboard DRAM.

Do you have some clock-for-clock comparisons at around 25-30 MHz, between HT12-16A and HT18/C?

douglar wrote on 2024-12-17, 18:57:

"MR BIOS archive sheet" lists all _documented_ BIOS files, but only the ones with a version number listed are in the jheronimus bundle.

I have found the matching Romimages. I split one of them and tested it. It works.

PC@LIVE wrote on 2024-12-17, 21:35:

It might not be too difficult, to make a new PCB, even if among my motherboards to repair I have a MB 486 DX the size of the 386SX, here maybe that would be a nice future project, if you're interested I'll tell you the brand and the Model.

Maybe. I worked together on the two boards 286 and 386 for about 2 years. In several revisions.
The chipset for the 386 board can still be purchased from brokers.

georgel wrote on 2024-12-17, 21:58:

4 layers instead of 2 is hardly an improvement. To me it is a serious drawback but it simplifies your routing task.

4 layers (with separate ground plane) have a much more stable power supply and less noise on the board. Therefore almost all boards from 286 upwards use this. Why is this a serious drawback?

maxtherabbit wrote on 2024-12-17, 22:40:

A ps/2 interposer would easily fit this board, but yes of course having it built-in would be much preferred 😀

Maybe one day.

feipoa wrote on 2024-12-18, 02:14:

What was the per piece cost for the HT12 chipsets? What is the approximate cost for components and PCB to assemble a motherboard?

If I remember correctly, I paid €25-30 per chipset. I'll have to check.

feipoa wrote on 2024-12-18, 02:14:

Have you tried to run the Harris at 25 MHz?

The board will not start properly. Switching from low to high at a later time does not work either.

feipoa wrote on 2024-12-18, 02:14:

What is the approximate cost for components and PCB to assemble a motherboard?

The board does not have many components. Chipset, CPU, ISA socket and PCB are the most suitable costs. I estimate the costs to be around 80-100€. Depending on where you buy the parts.

feipoa wrote on 2024-12-18, 02:14:

We have an adjustable PLL with MHz increment/decrement buttons that fit into DIP-14 (just waiting on a new PCB layout): download/file.php?id=204708&mode=view

That looks good. I would be happy to test it.

feipoa wrote on 2024-12-18, 04:54:

Do you have some clock-for-clock comparisons at around 25-30 MHz, between HT12-16A and HT18/C?

I have the exact same board in that link. I was using 24Mhz mostly because of the ISA bus synchronising to the CPU clock if you don't solder a DIP-14 socket at the top-right corner of that board and set JP9 to 1-2, and with the PT-606E multi I/O card I was getting data corruption if I went to 12.5Mhz ISA bus speed - 12Mhz is perfectly stable. While it's nice to have a secondary IDE slot, I've never used it so I decided to switch back to the PT-606G, and now I seem to be getting bottlenecked at 13.5Mhz ISA bus speed.

My RAM for the HT12 (it is simply labelled HT12, not a variant) is not binned well enough to get above 25Mhz 0WS - running at 27Mhz corrupts the processing of CONFIG.SYS, but even at 25Mhz I get several stability issues. I can't even use a ET4000 without it immediately crashing in Wolf3D unless I drop down to 16Mhz 0WS, I only have 32/40/48/50/54/66Mhz oscillators available to use. It was doing fine with Windows benchmarks at 1024x768x256, but it crashed very early during the full Winbench 3.11 suite (as opposed to the useless Graphics WINMARK test). S3 911/924/801 cards will not load Windows at 256 colours, but are perfectly fine at 16 colours. The GD5429 is one of the most reliable cards I have, only showing a minor glitch with one specific test in the full Winbench 3.11 suite at 1024x768x256 and having an off-by-one issue with loading the Wolf3D title screen a single time directly after running Doom8088, which had glitched the rest of the image. No other issues that I've seen, even with running a full set of benchmarks at 800x600x65K in Windows.

With all that said, I freshly compiled some new benchmarks using the GD5429, the PT-606G for multi I/O and a 3com 3c509B that I use for XT-IDE. In addition to three runs of wolf_286, I also used three runs of WolfensteinCGA and a single run of the most recent release of Doom8088 using the doom2my and doom213h EXEs for benchmarking:

HT12 25Mhz 0WS, ISA bus at factory default of 8Mhz
doom2my - 18.119
doom213h - 15.976
wolf_286 - 17.2
wolf3dc - 27.4

HT18C 25Mhz, ISA bus set to 8Mhz via additional 32Mhz oscillator
doom2my - 14.857
doom213h - 12.839
wolf_286 14.6
wolf3dc 21.4

HT18C 25Mhz, ISA bus set to 12.5Mhz via CPU oscillator (factory default)
doom2my - 15.534
doom213h - 13.570
wolf_286 - 16.0
wolf3dc - 22.7

The CPU performance is fairly obvious with the Doom8088 benchmarks, as there's significantly less breathing room to be gained from increasing the ISA bus speed compared to wolf_286 or wolf3dc. The HT18C is good if you really need to multi-task a lot with Windows 3.1, but outside of one specific use case with a Doom map editor that works on Standard Mode that was claiming to use 6MB of RAM, I can't think of any genuine use case where I would prefer it over the HT12, especially when I get properly binned RAM which hopefully should be shortly after Christmas.

Thank you for checking this. DOOM's performance difference between the HT18C and HT12 is surprisingly high. If the delta is so large, then it would be better to use the ISA RAM extender with the HT12.