Which architecture to start with?

[Schol-R-LEA]

So, what do you think: which would be the best way to go for now?

MIPS is good. Not because I work for the company, but because it's a well-known arch, it's rather regular and if you want a board with it to play, there are many options. There are inexpensive boards capable of running RetroBSD, OpenWRT, Linux, Android.

I agree that MIPS - and it's descendants such as MIPS-X, DLX, Loongson (which is less a descendant than a flat-out knockoff), and OpenRISC - are among the cleanest and simplest high-performance architectures around (most of those which are simpler are either PICs, or older 8-bit designs such as the PDP-8 or the MOS6502), and I love them to death. However, they don't have more than a fraction of the installed hardware base of x86 or ARM, even taken together and including all the various embedded systems based on them. This may change as more Loongson-based laptop and mobile systems get built, but it is unlikely that those will be sold outside of the People's Republic of China in significant numbers.

While there are a handful of SoC and makerboard kits for MIPS, there aren't narly as many as for ARM, and AFAIK no stock form factor motherboards at all (apparently, there used to be an eval board called Jaguar ATX, but the company that made them seems to be out of business). Perhaps what is needed is a Kickstarter campaign to raise funds for the development of an ATX or ITX form-factor MIPS desktop motherboard? It would have to focus on Linux or maybe one of the BSDs as it's primary OS, and it would be a niche market item at best, but there are quite probably enough devs and engineers who would be interested in it to at least do something. Given that Imagination Technologies (who bought out MIPS Technologies in 2013) seem to be looking for something like that, it might even have a real product someday

[alexfru]

While there are a handful of SoC and makerboard kits for MIPS, there aren't narly as many as for ARM, and AFAIK no stock form factor motherboards at all (apparently, there used to be an eval board called Jaguar ATX, but the company that made them seems to be out of business). Perhaps what is needed is a Kickstarter campaign to raise funds for the development of an ATX or ITX form-factor MIPS desktop motherboard? It would have to focus on Linux or maybe one of the BSDs as it's primary OS, and it would be a niche market item at best, but there are quite probably enough devs and engineers who would be interested in it to at least do something. Given that Imagination Technologies (who bought out MIPS Technologies in 2013) seem to be looking for something like that, it might even have a real product someday

I'm not sure you really need a MIPS-based PC that's almost exactly x86-like, except for the CPU. There's Tavolga Terminal AKA Meadowsweet Terminal. But you don't even need that. Creator CI20 runs Linux just fine. Simply attach a keyboard and a display.

[Schol-R-LEA]

I was thinking mainly that it would make it easier to build desktop support if it could use as much stock hardware as possible. However, to be fair, with so many peripherals using USB, eSATA or similar external buses, for most systems the only thing that you really need a backplane for these days is a high-performance video card, and even external video GPUs seem to be coming back into use.

To be honest, I was never a fan of the motherboard/high-speed memory/backplane bus approach that has dominated systems since the mid-1970s, anyway; it scales terribly, and causes many of the problems related to overheating, inability to use heterogeneous multiprocessing, etc. that are still with us today. Since the early 1990s, I have favored a more modular design philosophy, with separate units for CPU and main memory, secondary and tertiary storage (as well as additional shared main memory), video output (preferably with the GPU and video memory integrated into the monitor itself, and perhaps having an additional local CPU for controlling it), and so forth on a multi-master wired bus that doubles as a LAN bus, but unfortunately, that's not the way things went, and no one ever stepped up on that because they were all too busy copying each other to improve things.

I'd say that maybe that is the Kickstarter someone needs to do, but I lack the hardware skills to do it myself, In any case, if memory serves someone tried that a few years ago and didn't get any interest.