BitWrangler wrote:

I mean look at that kit up there, slapping that together in 1979 would probably have been good for a Master's thesis in E-E, now hobbyists do it.

Your answer:

anthony wrote:

this particular xt board is 2 layer, they are very cheap, it like slighty expanded zx-spectrum board. for 486 chipsets you'll never find datasheets ( i presume). for super s7 chipsets boards there must be at least 4 layer pcbs until certain impedance required. if so, there must be 6 layes pcb.

most of you don't realize how much complicated task it is to produce even 25-year old harware replica

----

Jade Falcon wrote:

XT class hardware is nowhere as complex as a 486 system. If this was not the case we would see 486 class system kits.

Heh... Even with XT class isn't as easy as it looks. Check the link i posted before, and you will see all the troubles people had to assemble the Xi 8088 kit and make it work stable. I still see there some guys which don't get their assemblies to work in a stable way.

Even though I like the 'idea / dream' of designing and manufacturing a new 486 motherboard with lots of complicated features (or what about a super socket 7 one?), this is currently not realistic at all, unless someone of you guys (or investors in a crowdfunding project) has an enormous amount of money to spend on this project and wants to take a massive loss. We should not forget that we are a very tiny niche market. How many people really still use a 486 (or any 15+ years old classic) pc system? Probably at best a couple thousand worldwide, but I'm afraid it's more closely to 2000 users than 5000 users.. And how few of them actually want to buy a new 486 board for a lot of money??

I also agree with the argument that old motherboards are still plenty to be found on online sources, for relatively little money.

To make this an interesting project, the new 486 board should be <100 euro / U.S. dollar assembled (so no DIY kit). With such low demand this is mission impossible.

hyoenmadan wrote:

Your answer: […]
Show full quote

BitWrangler wrote:

I mean look at that kit up there, slapping that together in 1979 would probably have been good for a Master's thesis in E-E, now hobbyists do it.

Your answer:

anthony wrote:

this particular xt board is 2 layer, they are very cheap, it like slighty expanded zx-spectrum board. for 486 chipsets you'll never find datasheets ( i presume). for super s7 chipsets boards there must be at least 4 layer pcbs until certain impedance required. if so, there must be 6 layes pcb.

most of you don't realize how much complicated task it is to produce even 25-year old harware replica

----

Jade Falcon wrote:

XT class hardware is nowhere as complex as a 486 system. If this was not the case we would see 486 class system kits.

Heh... Even with XT class isn't as easy as it looks. Check the link i posted before, and you will see all the troubles people had to assemble the Xi 8088 kit and make it work stable. I still see there some guys which don't get their assemblies to work in a stable way.

Im aware of the problem with that XT kit, there are others that are far better.

This thread reminds me of when I was in 3rd grade and some friends decided we should make a movie.

And we didn't have a camcorder... or a script... or money... or...

Dreams. We did have dreams.

By the way, if you really want to make something, then start with AT style I/O shields. If that turns out to be a success, then think about manufacturing baby AT cases. Those are really scarce in good condition without discoloration and corrosion. I would buy both if prices are comparable to regular I/O shields and ATX cases.

SpectriaForce wrote:

By the way, if you really want to make something, then start with AT style I/O shields. If that turns out to be a success, then think about manufacturing baby AT cases. Those are really scarce in good condition without discoloration and corrosion. I would buy both if prices are comparable to regular I/O shields and ATX cases.

I feel pretty confident that the entire market for new AT cases is maybe 500 people. If someone thought they could make money on them, they'd already be getting made.

You'd have better luck with getting someone to make a retro looking ATX case, that might be a more palatable prospect. But still I'd wager the market is small.

Spent my youth with 8088s and everything after, and really appreciate DOSBox and ScummVM and especially MUNT but wondered what was going on for new production hardware from that era, especially things like 486-ish performance, but Ebay made it clear that vintage will COST you right now, its crazy. Should have kept all that stuff 😀 Wondered how all the non-gamers dealt with this all if emulation didn't work, so I'm trying an experiment now to assemble a new production 486-ish system.

Criteria is that is must be easy to buy new, must play anything Sierra w/ "Quest" in its name. Secondary is that talkies work with the MT-32 and a joystiq for WC2.

There is a wealth of info here on some options like ZFx86 and alibaba motherboards, but looks like you can get a Vortex86 in PC/104 or ISA format, get a passive ISA backplane, and potentially be in business? I'm giving it a try. The cheapest Vortex86 I found was PC/104 for their 800mhz but without video for about $139. Outside of ebaying some israeli PC/104 to ISA cards without any headers pre-soldered I couldn't find a cheap (less than $15) way to get a ISA backplane interfaced so I had to move up.

For about $237 you can get a ISA card mounted version of the Vortex86 including VGA (Advantech PCA-6743VE-Q0A2E), which given some power is enough to PoC if this is workable. Added on a passive 8 slot ISA backplane (PCA-6108-0B2E) for another $50, but there are smaller options there or I guess no option if you don't want better video or sound.

My hope is that a Vortex SBC and ISA backplane is a viable setup for new production retro computing, add some sound or better video if you want, maybe a HardMPU, and have a supported new production computer. Given the EBay prices right now I kind of can't believe someone isn't already doing this at scale for old industrial controls, but who knows if it will even work. Excited to find out.

Might as well develop dream 440BX board.

We should not forget that we are a very tiny niche market.

And 486 are not really that practical to use, even when compared to AT Socket 7 boards.

kanecvr wrote:

Nobody said it would be easy... but it is possible, and it could make money if crowdfunded trough something like kickstarter whi […]
Show full quote

Nobody said it would be easy... but it is possible, and it could make money if crowdfunded trough something like kickstarter while being sold in parallel on it's own website.

To keep cost down, unassembled kits could be sold. Just provide the PCB (with BGA parts pre-soldered) + the rest of the parts as a kit, just like this: http://www.mtmscientific.com/pc-retro.html

pckit2.jpg

I know I'd personally want the unassembled kit, and it would be great fun to put it together myself. Like I mentioned before, BGA parts would need to come pre-soldered, since they require specialized equipment to solder correctly. PGA parts can be soldered by the buyer by means of a SMD rework station witch is not an expensive tool.

I'm going to order one of those boards with 512k of ram. Total cost is 200$, and you get a DIY IBM XT clone - awesome.

Of course there should be an option for pre-assembled boards, for those who lack the skill and equipment, or are simply not interested in that aspect.

Hey, they're selling those local in my state. What a coincidence.

At some point, the design may no longer be considered a "486" and have a less broad appeal. I think the best option would be to clone an accepted favourite 486 PCI or 486 VLB motherboard. Either source new, old stock chipsets or reverse engineer them, then implement them in ASIC or w/FPGA. We would need to source a motherboard design engineer who has the passion and to work for free. It would be nice to have a 20-66 Mhz FSB range. Another nice twist might be to add a custom 2x PLL multiplier option so we could run IBM 5x86c chips at 2x2x33 if desired.

Anyway, for reasons that others have already pointed out, it seems unlikely that this dream will ever be realised. There are a lot of design aspects like inner-trace capacitance and impedence matching which are dependent on many factors, like trace spacing, substrate material, operating frequency, trace length, etc, which would require significant research and trial-and-error. For example, just look up the equivalent circuit diagram for a wire - inductance, capacitance, resistance, shunt and short - add many wires in parallel... Ideally, you would want to find someone who has experience in this and has the free will. Otherwise, start reading, and start saving. I hope you don't have little kids!

I’m entertaining similar idea for some time and even completed relatively simple project of this kind - http://www.vcfed.org/forum/showthread.php?464 … 6939#post426939

At the moment I have a half completed design of the board that can work with 8088/80286 CPUs, so with a button push you can switch between PC/XT/AT286. Biggest issue (at least, for me) - available time. I have not much of free time, and have many other things to do...

If we are talking about 486 class system - in my opinion, it never happens. Only possibility - someone gets full set of factory docs, so you don’t need to design/troubleshoot anything, just put in production. If they need to design ASIC/FPGA - will not happen, will take too much work...

I seem to remember seeing in a video that System 76 has an open source disk I/O controller. Clearly not enough for this endevour, but at least its something.

As of 5 years ago, I was still able to order UMC 8881/8886 chipsets NOS, perhaps you still can? Someone could pick their favourite UMC 8881-based PCI motherboard and spend a year tracing out the components and writing up the Gerber, but it would help if you had experience in motherboard design as you will inevitably miss some connections. Biostar MB-8433UUD is an OK board to copy, but my biggest complaint is that it doesn't have a 74F244 buffer between the TAG and the Cache addresses. This seems to have the impact of requiring reduced cache timings when using more than 256K cache and a faster FSB. Of course, there are several boards which do have the buffer and would be straight-forward to copy. It might be easier to copy the Gigabyte 486 AM/S, but it doesn't have the PS/2 mouse implementation that the 8433 does. One could combine the best of both boards. However, I suspect that using the buffer will results in 66 MHz not working for a FSB. Some modelling might help with this.

Got one of the Vortex86DX half-length ISA boards booting last night, just with a quick DOS 7.1 install and some benchmarks, just trying the bare ISA card for now but also go the passive 8-slot backplane which seems to have AT / ATX / and screw terminals for power along with ATX power button support.

Bare board (compact flash slot on the back, lots of headers onboard, they provide USB, PS/2 splitter, and a few other. This is the Advantech PCA-6743VE-Q0A2E:

Vortex86DX supports 8 different speed dividers, the processor docs list them as 1x-8x divide but the BIOS here shows different steps (32x divide!), not sure who is telling the truth. I emailed DMP Electronics and they said a L1/L2 cache disable is their provided AMI BIOS for the Vortex, but Advantech is using their own AWARD BIOS here and isn't showing the option:

Quick benchmark run with the divider set to 1x then 32x, so the max and min:

Thats all I had time for last night, hoping to get some better benchmarks run and some games setup to see how it works, and get the ISA backplane going with a SB16 and eventually a HardMPU too. The onboard video is off the internal PCI bus that isn't exposed on this board, so I'm hoping it works well because it would be great to have a high-speed VGA setup like that.

the dx version is a beast. 1.5gh 486. The SX version is kinda meh. no fpu. Is yours a dx or sx version? (they were afaik all called Vortex86DX, even the SX models). I cant see the chip in the pic at work...

BloodyCactus wrote:

the dx version is a beast. 1.5gh 486. The SX version is kinda meh. no fpu. Is yours a dx or sx version? (they were afaik all called Vortex86DX, even the SX models). I cant see the chip in the pic at work...

DX version at 800MHz max. I've got a 1GHz in PC/104 form factor too but haven't tried it yet.

http://www.advantech.com/products/1-2jkn7b/pc … 18-d5cca41d4ff5

CkRtech wrote:

This thread reminds me of when I was in 3rd grade and some friends decided we should make a movie.

And we didn't have a camcorder... or a script... or money... or...

Dreams. We did have dreams.

Were you in the 3rd grade in '78-'79? Because I and a friend had the same dream! (We wanted to make our own movie in the Star Wars universe back then)

Back on topic, though-- I like the idea of using an FPGA on a PGA132 base to make a sort of "486DXL2-80", with a ribbon to connect to the 387 socket to perform the FERR signal, à la the RapidCAD. It'd be kindof the PC version of the Vampire 600.

I suppose going through all that effort just to pump up a 386 motherboard would be silly, though...

How much did this computer card cost? I couldn't locate the SATA port on yours.

feipoa wrote:

How much did this computer card cost? I couldn't locate the SATA port on yours.

Mouser stocks them in the US, it was $237 for the 800Mhz ISA board, but the PC/104 version at 1Ghz is much cheaper at $139! I'm going to try that with a PC/104 to ISA adapter next. The SATA may need an adapter cable, I'll take a look tonight.

I used to work for a company who did number of designs from 386 to SS7. It's not undoable, but you'd need a lot of dedication. One design took about a year, with two HW engineers working on schematics and then one or two layout guys routing the board. You need a good PCB software with multiple layer support, constraints etc.
I personally did a simpler project, which was NatSem Geode GX1 based processor module (pretty much whole computer in PC104-ish size). I did HW design, bus design and BIOS customization, and that one was easier than full fledged ATX board.
It might be difficult to get all the necessary documentation. The common Intel "yellow" pages wouldn't be sufficient, as the HW design stuff, timing constraints, errata etc. went to "orange" or even "red" pages. I assume those documents are no longer printed and available anyhow.

Kubik wrote:

I used to work for a company who did number of designs from 386 to SS7. It's not undoable, but you'd need a lot of dedication. One design took about a year, with two HW engineers working on schematics and then one or two layout guys routing the board. You need a good PCB software with multiple layer support, constraints etc.
I personally did a simpler project, which was NatSem Geode GX1 based processor module (pretty much whole computer in PC104-ish size). I did HW design, bus design and BIOS customization, and that one was easier than full fledged ATX board.
It might be difficult to get all the necessary documentation. The common Intel "yellow" pages wouldn't be sufficient, as the HW design stuff, timing constraints, errata etc. went to "orange" or even "red" pages. I assume those documents are no longer printed and available anyhow.

Very interesting - so without this documentation it is a futile effort...