design a pci card with a isa bridge chip that could work on motherbords with south bridges that would suport such a thing. a seprate pcb with a isa slot and a header that would connect to the bridge card with ribbons.

Areal Vortex 2 emulation. Or has it been done..?

Joakim wrote on 2024-02-12, 18:05:

Areal Vortex 2 emulation. Or has it been done..?

you mean A3D? there are software libraries emulating it in software.

Here's an ambitious one: have someone start to build CRT monitors again. 😁
Since modern video cards don't have RAMDACs anymore, perhaps have HDMI + DVI + DP on the monitor with RAMDAC inside the monitor.
(Yeah, I know, chances of this are close to null. Making CRT monitors is not exactly cost-efficient and takes a lot of skilled labor... unlike modern-day LCDs, which can be slapped together by machines in a number of seconds.)

I'm pretty sure the assembly lines to make the kinescopes are long sold off and scrapped so getting good newly made kinescopes will be quite difficult and very very expensive...
Electronics side proposes some challenges too, particularly final cathode amps where there's nothing modern that will work there. Most rest can be done without much problems at all. Perhaps some of the old parts can be revived through places like Rochester Electronics who seem to specialize in making legacy parts, although at very uncompetitive prices 🤣.

Has anyone capped RAM on a 386?

I mean, until that point it was feasible to cap a processor (1Mb on a 8086, 16Mb on a 80286) but a 386 could address 4Gb of RAM. Did anyone build a mainboard capable or addressing 4Gb for 386 CPUs?

Zup wrote on 2024-03-08, 16:00:

Has anyone capped RAM on a 386?

I mean, until that point it was feasible to cap a processor (1Mb on a 8086, 16Mb on a 80286) but a 386 could address 4Gb of RAM. Did anyone build a mainboard capable or addressing 4Gb for 386 CPUs?

4GB was addressable by a 386DX, but the SX could only read 16 MB, due to the 16-bit external data bus. I'd be interested in a board that could host 4GB, but as far as I know, there are no 30 pin simms that would make that feasible. I guess that if you created hardware that could read DIMMs on a 386 board, it may be doable, but not sure it would be realistic for the hobby market, even if some collectors are relatively wealthy 😄

Zup, you'll need to build your own chipset to control & refresh that amount of memory, likely possible with a nice fpga.

debs3759, the sx had a 24bit address bus which limited it to 16M of RAM. Not the 16bit data bus.

weedeewee wrote on 2024-03-08, 16:15:

Zup, you'll need to build your own chipset to control & refresh that amount of memory, likely possible with a nice fpga.

debs3759, the sx had a 24bit address bus which limited it to 16M of RAM. Not the 16bit data bus.

Oops, my bad. I blame old age 😀

debs3759 wrote on 2024-03-08, 16:12:

Zup wrote on 2024-03-08, 16:00:

Has anyone capped RAM on a 386?

I mean, until that point it was feasible to cap a processor (1Mb on a 8086, 16Mb on a 80286) but a 386 could address 4Gb of RAM. Did anyone build a mainboard capable or addressing 4Gb for 386 CPUs?

4GB was addressable by a 386DX, but the SX could only read 16 MB, due to the 16-bit external data bus. I'd be interested in a board that could host 4GB, but as far as I know, there are no 30 pin simms that would make that feasible. I guess that if you created hardware that could read DIMMs on a 386 board, it may be doable, but not sure it would be realistic for the hobby market, even if some collectors are relatively wealthy 😄

The only thing I found was that things like the 386 DX compaq servers had both a large amount of onboard and the possibility of additional memory expansion boards that used proprietary (but simm like?) socketed memory that used the highest densities available at the time. I think the amount you could have is somewhere around 256-384MB of memory. But the server line was replaced with the 486 when it came out, and even while being somewhat similar, this precluded 386 motherboard designs from really ever supporting more than that. These were insanely expensive options, and you gotta remember, when 386 was king, that was during the memory chip shortage period. So the 386 never really got used to its full potential here.

The main thing the 386 was useful for was the expanded virtual memory space. It made multitasking much easier. It basically fixed the 286.

An 8088 or NEC V20 replica/substitute with an 21st address pin and 64KB of SRAM would be cool.
That would allow XT users to have HMA on MS-DOS 5/6.

A missing A20 gate isn't that much of a problem to MS-DOS, really.
It's mainly CP/M era applications that need the address wraparound.

Edit:

The main thing the 386 was useful for was the expanded virtual memory space. It made multitasking much easier. It basically fixed the 286.

Hi, I think the 80386 is much more capable than most people know. By using segmentation, it can handle up to 64 TB of virtual memory.

The use of segmentation also allowed to designate segments as containing data and application code.

So there was memory protection based on segmentation, years before NX/XD Bit and Window's DEP were around.

Both 80286 and 80386 had same fathers, also, I vaguely remember.
So without the 80286, the 80386 wouldn't have existed in its final form, maybe.

The main use of the 80286 or iAPX286 was data processing systems, PBX systems and such. It wasn't designed for PCs yet.

The 80386, by contrast was designed to be like a mainframe. This made porting easier, 32-Bit code was previously being used in mainframe environments.

It completely skipped the application as a PC processor so to say, and somewhat overdid. 😉

It was more than what IBM wanted it to be, also, I think.
That's why, for some time, I think IBM had supported the 80286, at best. The 80386 made them afraid.

debs3759 wrote on 2024-03-08, 16:12:

Zup wrote on 2024-03-08, 16:00:

Has anyone capped RAM on a 386?

I mean, until that point it was feasible to cap a processor (1Mb on a 8086, 16Mb on a 80286) but a 386 could address 4Gb of RAM. Did anyone build a mainboard capable or addressing 4Gb for 386 CPUs?

4GB was addressable by a 386DX, but the SX could only read 16 MB, due to the 16-bit external data bus. I'd be interested in a board that could host 4GB, but as far as I know, there are no 30 pin simms that would make that feasible. I guess that if you created hardware that could read DIMMs on a 386 board, it may be doable, but not sure it would be realistic for the hobby market, even if some collectors are relatively wealthy 😄

I was actually pondering "the point" of something I thought of doing along similar lines. IDK if anyone remembers me noting that Lava PCI i/o cards actually have a Xilinx FPGA on them, I had been thinking of how one might reprogram one to do something interesting and then I started wondering the other week about making them an RDRAM interface, then you'd have a chunk of RAM at PCI speeds. Probably not real useful when you can maybe get same speed from a SATA card and cheap SSD and have it non-volatile, but maybe a bit interesting on say a Pentium classic or MMX to have big RAM cache, RAM disk or something. Was also pondering if one could add an ARM devboard to it and emulate a voodoo .

Wireless NIC on ISA slot supporting 802.11b/g/n + drivers for "dead" operating systems like DOS, win3.xx/9x.

You can technically replace 30-pin DRAM sticks with custom SRAM sticks. I have a design mocked up.

Speeds are ~10s for all accesses (page, random) and allows you to max out your BIOS settings... If your BIOS is already maxed, then not sure if there's benefit.

What you want to do is have 16-64MB of L2 cache, writeback, no chipset intervention...

no idea if possible, but an 8086 mainboard with 16 bit isa slots. Does that exist, if the 16 bit isa slots are possible to use on an 8086 ?

analog_programmer wrote on 2024-03-09, 05:07:

Wireless NIC on ISA slot supporting 802.11b/g/n + drivers for "dead" operating systems like DOS, win3.xx/9x.

This is in progress for the PicoMEM. You need a Pico W to do it

MikeSG wrote on 2024-03-09, 08:54:

You can technically replace 30-pin DRAM sticks with custom SRAM sticks. I have a design mocked up.

Speeds are ~10s for all accesses (page, random) and allows you to max out your BIOS settings... If your BIOS is already maxed, then not sure if there's benefit.

What you want to do is have 16-64MB of L2 cache, writeback, no chipset intervention...

I would love to have 512mb of this in my tyan s1564D

weedeewee wrote on 2024-03-09, 10:28:

no idea if possible, but an 8086 mainboard with 16 bit isa slots. Does that exist, if the 16 bit isa slots are possible to use on an 8086 ?

Maybe trying to fit a 8086 in a later motherboard is the way to achieve this?

weedeewee wrote on 2024-03-09, 10:28:

no idea if possible, but an 8086 mainboard with 16 bit isa slots. Does that exist, if the 16 bit isa slots are possible to use on an 8086 ?

Hi, I guess it works with some ISA cards.
Say, an 16-Bit UMB or EMS card, 16-Bit NIC, 16-Bit IDE card or a 16-Bit VGA card (no linear frame buffer at 15-16 MB).

The 8086 has fewer data/signal lines than an 80286, address space is 1 MB rather than 16 MB.

A few things are missing, also.
The additional IRQ and DMA lines of AT platform are absent on XT motherboard designs, normally.

But otherwise, the basic signals of ISA could be made available, I think.

The Olivetti M24 had some 16-Bit extension, which was looking a bit like the VLB extension often seen on 486 motherboards.

I've found some pictures here: https://mastodonpc.tripod.com/personal/m24cards.html

Jo22 wrote on 2024-03-10, 08:31:

Hi, I guess it works with some ISA cards. Say, an 16-Bit UMB or EMS card, 16-Bit NIC, 16-Bit IDE card or a 16-Bit VGA card (no l […]
Show full quote

weedeewee wrote on 2024-03-09, 10:28:

no idea if possible, but an 8086 mainboard with 16 bit isa slots. Does that exist, if the 16 bit isa slots are possible to use on an 8086 ?

Hi, I guess it works with some ISA cards.
Say, an 16-Bit UMB or EMS card, 16-Bit NIC, 16-Bit IDE card or a 16-Bit VGA card (no linear frame buffer at 15-16 MB).

The 8086 has fewer data/signal lines than an 80286, address space is 1 MB rather than 16 MB.

A few things are missing, also.
The additional IRQ and DMA lines of AT platform are absent on XT motherboard designs, normally.

But otherwise, the basic signals of ISA could be made available, I think.

The Olivetti M24 had some 16-Bit extension, which was looking a bit like the VLB extension often seen on 486 motherboards.

I've found some pictures here: https://mastodonpc.tripod.com/personal/m24cards.html

Hate to tell you that you've told me nothing new. Maybe someone else has some use out of the info 😀

weedeewee wrote on 2024-03-10, 09:03:

Hate to tell you that you've told me nothing new. Maybe someone else has some use out of the info 😀

No no, that's fine. I didn't mean to tell you anything new. It was intended more like a confirmation to your idea.

Now that we talk about it, there seems to be a sort of an implementation of your idea.
Some modern XT backplanes seem to feature the full ISA connector, probably for future expansion.

An 8086 or 80186 CPU card could make good use of it.
Maybe for an experimental memory expansion card.

If the backplane isn't using the 16-Bit connector part for itself, it could be used for about anything.
It would solely be up to the CPU card to define the pinout (pins of all slots would still be wired in parallel, though).

Here's an ISA retro backplane meant to be used with the Xi 8088 processor board.
It has a POST card built-in, it seems.

https://www.retrobrewcomputers.org/doku.php?i … backplane:start

If an 80186 CPU card existed, it could make use of the 16-Bit part of the edge connector.