VOGONS


Reply 40 of 123, by dionb

User metadata
Rank l33t++
Rank
l33t++
GL1zdA wrote on 2020-03-06, 17:11:

Given that the cheapest NuXT is $280, I don't see how you can make a new cheap 486 motherboard. If you can't get an existing 486 chipset, creating one from scratch in FPGA would probably be prohibitive and I don't think ao486 performance is what people are looking for when they build a 486:
https://www.youtube.com/watch?v=G32lGUVeodQ
Personally, I would buy only such board if it had performance comparable to the late 486 PCI boards. On the hardware side it would need at least several PCI and ISA slots, so micro-ATX is a no-go. 486 Baby-AT boards had often as much as 7 slots, including long VLB ones.

VLB is an interesting proposition. Are there still massive stocks of MCA connectors to use for that?

Not so convinced uATX is out of the question. That gives you four external slots. I/O will already be integrated on motherboard. Because PCI and ISA are mirrored you can share an external slot between them, so you could do 1x PCI and 4x ISA (+VLB if desired), or 2x PCI and 3x ISA.

What would you want to use?
- VGA? The design here integrates that. Not at all sure that's a great idea, but it certainly saves a slot.
- NIC? Yep, 1 ISA or PCI for that.
- Sound? Two ISA should be enough.

So you're only using three slots for a fully featured system, even assuming two sound cards (decadent). That leaves a slot for whatever else you want, like a discrete video adapter. I'd say the form factor is more than viable. Full ATX would limit choice of cases, increase costs & complexity significantly and forces you to go for a big-ass system.

Reply 41 of 123, by Nitroraptor53

User metadata
Rank Member
Rank
Member
GL1zdA wrote on 2020-03-06, 17:11:

Given that the cheapest NuXT is $280, I don't see how you can make a new cheap 486 motherboard. If you can't get an existing 486 chipset, creating one from scratch in FPGA would probably be prohibitive and I don't think ao486 performance is what people are looking for when they build a 486:
https://www.youtube.com/watch?v=G32lGUVeodQ
Personally, I would buy only such board if it had performance comparable to the late 486 PCI boards. On the hardware side it would need at least several PCI and ISA slots, so micro-ATX is a no-go. 486 Baby-AT boards had often as much as 7 slots, including long VLB ones.

No, I'd be using FPGA's and a LOT of integretion. Like I said, I'm going to use EISA and PCI. It WILL fit.

Reply 42 of 123, by SirNickity

User metadata
Rank Oldbie
Rank
Oldbie

I mean... most of the stuff going on in a 486 is 25 or 33MHz. If you're an overclocker with Pentium envy, maybe 50MHz. That's not much to ask of modern FPGAs.

The direction this project is going sounds to me a little more like scratching a personal itch than something that will fill a broader need. Just IMO. Maybe I'm wrong?

Reply 43 of 123, by Nitroraptor53

User metadata
Rank Member
Rank
Member
SirNickity wrote on 2020-03-06, 19:07:

I mean... most of the stuff going on in a 486 is 25 or 33MHz. If you're an overclocker with Pentium envy, maybe 50MHz. That's not much to ask of modern FPGAs.

The direction this project is going sounds to me a little more like scratching a personal itch than something that will fill a broader need. Just IMO. Maybe I'm wrong?

https://www.youtube.com/watch?v=pwzHurzTNTY&t=929s read the comments.

Reply 44 of 123, by SirNickity

User metadata
Rank Oldbie
Rank
Oldbie
Nitroraptor53 wrote on 2020-03-06, 19:43:

Right, so what you're citing there is a lot of anonymous enthusiasm for "a 486 version" of the NuXT. Here's the problem:

For one, people don't actually know what they want. They would LOVE an all-in-one board with built-in video and sound and networking and USB and CF and WiFi and.... except, "I don't want that video chipset. What about Cirrus?" "NO! Definitely S3." "Could you maybe do one with dual SLI Voodoo 3s? I would definitely buy that." "Tseng or I'm out."

But then: "$50?! Are you NUTS? I can go down to my uncle's house and exchange him a 6-pack of Pabst Blue Ribbon for like 20 old 486 motherboards! Ohhh.. you said $500! 🤣 Good luck with that."

Reply 46 of 123, by Nitroraptor53

User metadata
Rank Member
Rank
Member
maxtherabbit wrote on 2020-03-06, 22:48:

trying to include onboard video seems retarded to me, there is absolutely no shortage of reasonably priced PCI cards on the secondary market

I mean, if you have extra space, it seems worth it...

Reply 47 of 123, by hyoenmadan

User metadata
Rank Member
Rank
Member

In ReactOS team they have a mantra... "First show a working and documented proof of concept. Then we can talk about features, wishes, magic dust, unicorns... And funding bucks". Before even you can talk about board designing, you first need a working chipset for your project and CPU target... Or a finished SoC if you choose that route. Designing these isn't easy, specially if you want all sorts of external IO and compatibility with existing hardware... And then there is the proper board design, with all the requeriments specific for an FPGA/ASIC with external PCI IO as you seem to want. There is a reason why you don't see out there a lot of FPGA designs out there, even simple ones (I for example, would like a better IDE-SATA converter chip than the existing JMicron crap, totally OpenSource), and everyone prefers projects with microcontrollers.

Reply 49 of 123, by maxtherabbit

User metadata
Rank l33t
Rank
l33t
Nitroraptor53 wrote on 2020-03-07, 01:17:
maxtherabbit wrote on 2020-03-06, 22:48:

trying to include onboard video seems retarded to me, there is absolutely no shortage of reasonably priced PCI cards on the secondary market

I mean, if you have extra space, it seems worth it...

seems like a great way to burn development resources on something that will not satisfy more than a fraction of users even if you do it perfectly

Reply 50 of 123, by Deunan

User metadata
Rank Oldbie
Rank
Oldbie
Nitroraptor53 wrote on 2020-03-08, 16:19:

https://i.ebayimg.com/images/g/ZtEAAOSwAKxWU3g2/s-l1600.jpg could this board be tweaked with an onboard CPU, remove to EISA slots, and a few other tweaks?

See that photo? That's what you are trying to make, with these features baked-in. It's probably not even a 4 but a 6 layer PCB, that alone will cost a lot.

I don't want to rain on your parade but when you ask a question "Can this be done?" the answer is actually another question - "How much time and money do you have?". And I'm assuming not a lot money, and you don't want the project to take 20 years to show something. So your only other option is to keep it simple - which means basic mobo with add-on cards for everything. And that won't be a walk in the park either.

My advice: Make the mobo accept 3V3 CPUs only, there's no shortage of these and most people would want a DX2 at the very least I think. Have 16, 25, 33MHz clock options and you can go down to 33MHz and up to 66 (or 100 for DX4).
Pick one RAM standard, stick to it. Either DIMM, or FPM/EDO. Preferably 3V3 DIMMs. Forget VLB, 16-bit ISA slots only, unless you can find some source of PCI host chips. There are also chips that take care of the whole "chipset" thing, like interrupt, DMA, timers - even 386 mobos had those. Either pick one or have an FPGA do all that. The FPGA could also do PCI host but you'd better find someone willing to write that code, or pay for existing IP (and the licenses are usually per device, not per project). With all that and an eager team, you can have a beta board in 6-12 months. That's with some very basic BIOS, mostly lifted from other, existing projects.
With FPGA you could do some simple IDE2SD or something, or connect DRAM through it since many FPGAs can use DDR+ and maybe even give you small but fast L2 cache based on internal RAM blocks.

That's how I'd do it, if I had the time. That's be still some 200 EUR per board I think, but might just attract enough hobbyists to earn some money.

Reply 54 of 123, by Dant

User metadata
Rank Member
Rank
Member

I think the idea of building new boards for 486 or even Socket 7/Super 7 Pentium and K6 is a great idea but, like the NuXT, I think as much readily available hardware should be used as possible. Building an FPGA for large-scale integration is no simple task, and having to handle the complexities and edge-cases of the 486 memory bus, caching schemes, non-standardization of VLB and the lack of documentation behind EISA will add too many hurdles to overcome for your first go at this.

I'd recommend instead just getting an available off-the-shelf chipset (maybe there's a big stock of PCI UMC chipsets on Alibaba?) and building a good, stable, high-quality board around that. It might be even better, TBH, to build something around the Vortex86 platform, even. Going for a straight-up integrated CPU/chipset that can take modern memory but still performs and has compatibility like a 486 system would still be great.

Reply 55 of 123, by BloodyCactus

User metadata
Rank Oldbie
Rank
Oldbie

Im doubtful. JD just copy-pastad Sergey's designs for the NuXT thats stitched together by the isa bus. Throwing in a complex 486 chipset , non isa bus etc, PCI and its chipsets... thats a lot of work. Way more than what went into NuXT. I dont really see the point.

--/\-[ Stu : Bloody Cactus :: [ https://bloodycactus.com :: http://kråketær.com ]-/\--

Reply 56 of 123, by SodaSuccubus

User metadata
Rank Member
Rank
Member

As great as the idea of modern 486 boards and the NuXT are, until they come down in price, i personally don't see much point in them.

Assuming it'd cost as much as a NuXT (hell, probably more tbh), and it was a barebones unit, for that same price you could just wait and grab a decent 486 board off the 'bay. Probably a bundle with processor and ram included. I see those go for auction often enough.

Reply 57 of 123, by SirNickity

User metadata
Rank Oldbie
Rank
Oldbie

It isn't going to come down in price. It can't possibly sell well enough to be a commodity item, so there's no economy of scale for the expensive parts like the R&D, board design, and FPGA development. If it were ever prolific enough to recoup the up-front cost of converting the FPGA design into an ASIC, that would help. But it won't sell enough to subsidize the initial design costs, much less iterative optimization.

This is the unfortunate dynamic of consumer electronics, where people have gotten used to paying $100 for a multi-core tablet with WiFi and a terabyte of flash storage, and $3 for some small adapter / converter / doo-hickey from China. Small-run niche products don't stand a chance in that economy, and consumers aren't willing to cope with the fact they can't have everything they want.

The only way something like this will get made is as a labor of love, where someone(s) is/are willing to eat the design costs, and sell the product at barely a break-even point for the raw materials. Even then, you'll get indignant cries of "OMG! It costs WHUT?! That's such a rip-off! And I know because I'm obviously so well-versed in product logistics that I would know a BOM from a hole in my #$%."

Reply 58 of 123, by LewisRaz

User metadata
Rank Member
Rank
Member
Nitroraptor53 wrote on 2020-03-08, 16:19:

https://i.ebayimg.com/images/g/ZtEAAOSwAKxWU3g2/s-l1600.jpg could this board be tweaked with an onboard CPU, remove to EISA slots, and a few other tweaks?

I have 2 of those exact boards and the CPU comes on a seperate board that is again half the size of that motherboard, which is already huge.

My retro pc youtube channel
Twitter

Reply 59 of 123, by Deunan

User metadata
Rank Oldbie
Rank
Oldbie

I was looking for some FDC controllers for my own projects and by chance I found a chip called FDC37C932APM - that's basically a SuperIO chip with pretty much most of the PC mobo basic stuff in it (keyboard controller, RTC, CMOS RAM, FDC, two serial ports, IDE, the works). It can still be bought in some small quantities so might be useful for a project like that. It's a PQFP-160 package, not the easiest to solder but also not BGA.