Most bang for the buck new Amiga open Hardware motherboard

[8bitbubsy]

Originally Posted by Bruce Abbott

Most people are uninteresting.

You are uninteresting.

Originally Posted by Bruce Abbott

Almost as uninteresting as Windows 10.

My point is that a computer needs modern software to be usable to the vast majority of computer users. If it doesn't have that, it's just another niche (for the most part). Windows 10 is another subject on its own, it surely has its flaws and a scary movement towards an OS being a service instead of a product.

Originally Posted by Bruce Abbott

I thought the same. Now my A1200 sometimes gets more use than my PC!

Good for you. However, to over 99% of people, an A1200 is a useless paperweight these days.

[coldacid]

Originally Posted by jbilander

Of course if you are designing a new Amiga it makes most economical sense to put all the custom chip behavior in one large FPGA together with some external RAM and the CPU (also in verilog/vhdl). This is basically what the V4 SA or the Minimig/MiSTer is. It might not be as fun though without Paula, Agnus, Denise etc doing the "chip-dance" on the actual dance-floor and no floppy drive eating disk in the background.

The thing is, we need to cover both bases -- individual chips and all-in-one chips. For a new, from-scratch Amiga design, sure put the custom chipset in a single FPGA. For everything else, individual chips with matching pins are what's needed. A good logic description in VHDL or Verilog should be able to do this.


The other issue is that if you want to properly recreate a chip, warts and all, you need to build up from the transistor level. Unless you know the chip is entirely stateless (which apart from simple logic ICs they never are) just treating it like a black box and seeing what happens when you do something will never be 100% accurate. It might not even be 90% accurate, because you won't pick up the edge cases that'll probably bite someone in the ass down the road. If you can describe the logic from the individual transistors upwards, though, you can guarantee that your implementation will be a perfect match for the existing silicon.

[coldacid]

Switching away from all this back-and-forth about the chips themselves, what would I like in a modern 68k Amiga board?

Certainly USB support, SATA, onboard Ethernet, and maybe even still maintain a floppy adapter. Zorro III bus support. I'm on the fence about PC buses (classic PCI and AGP might be useful given the tech level but I don't think anyone uses those anymore for anything new, and PCIe is probably too OP for a 68k system).

The ability to add co-processors for certain things that a 68k doesn't do (SIMD, codec support, the "FPU" instructions that the '040 and '060 throw back to regular code, etc.) although I'd rather just have a "modern" 68k that has at least some of this functionality on board (really, just the FPU stuff; we should be able to fit the logic of a full 68060 _including_ the pushed-to-code instructions on a modern FPGA). OTOH custom co-processors raises the same kind of fragmentation issues that the different PPC kernels did back in the day.

Ability to use bog-standard PC PSUs without any muss.

[eXeler0]

Originally Posted by coldacid

Switching away from all this back-and-forth about the chips themselves, what would I like in a modern 68k Amiga board?

Certainly USB support, SATA, onboard Ethernet, and maybe even still maintain a floppy adapter. Zorro III bus support. I'm on the fence about PC buses (classic PCI and AGP might be useful given the tech level but I don't think anyone uses those anymore for anything new, and PCIe is probably too OP for a 68k system).

The ability to add co-processors for certain things that a 68k doesn't do (SIMD, codec support, the "FPU" instructions that the '040 and '060 throw back to regular code, etc.) although I'd rather just have a "modern" 68k that has at least some of this functionality on board (really, just the FPU stuff; we should be able to fit the logic of a full 68060 _including_ the pushed-to-code instructions on a modern FPGA). OTOH custom co-processors raises the same kind of fragmentation issues that the different PPC kernels did back in the day.

Ability to use bog-standard PC PSUs without any muss.

Thanks for the feedback.
Yes, as interesting the discussion about FPGA replacement for the chipset chip is, its beyond the scope of this "thought experiment".
I didn't include Ethernet or Wifi in my "spec" so yea that's probably a good thing to include.
For some reason I'd love to see a 040 being used on a mobo, but if Buffee becomes a thing and it performs well, then a cost effective solution is probably to just provide an empty 68000 socket on the Mobo then people could choose between
1. original 68000
2. Various Motorola accelerators (here's the problem though, there are no really affordable 040 accelerators)
3. Vampire
4. Buffee

[eXeler0]

What would be the wisest choice in terms of providing some kind of network connection straight on the mobo (or at least a simple add-on board) ? One of those cheap Wifi modules? Ethernet?
What existing projects could be adapted to minimize the "inventing the wheel" moments.
Plipbox utilized the Parallell port.. Any clever way to speed up the parallell port without breaking to much legacy stuff?
A new mobo likely wouldn't have PCMIA..
(Just thinking out loud here..)

[torsti76]

Originally Posted by eXeler0

What would be the wisest choice in terms of providing some kind of network connection straight on the mobo (or at least a simple add-on board) ? One of those cheap Wifi modules? Ethernet?
What existing projects could be adapted to minimize the "inventing the wheel" moments.
Plipbox utilized the Parallell port.. Any clever way to speed up the parallell port without breaking to much legacy stuff?
A new mobo likely wouldn't have PCMIA..
(Just thinking out loud here..)

Matze's Zorro-LAN-IDE-CP has a nice ethernet part with juicy speeds. I'd prefer that over Wifi solutions because there's virtually no encryption overhead you'd have to offload somewhere...