bfranchuk wrote: The real problem is building an authentic Lilith with today's hardware. For example, I can't yet track down a supplier of a 2901 - AMD don't make them, Signetics don't, Cypress semiconductor don't seem to. Does anyone still make them? So, then you're left with 3 choices: * Build a Lilith out of TTL. * Build a Lilith using an FPGA. * Build it using a Microcontroller. The TTL solution is crazy - even ETH didn't have to do that. The FPGA solution isn't bad, you could probably do it with a smallish FPGA to handle the micromachine and some Flash chips for the Microprogram (2x8-bitx128K would have sufficient bandwidth). But it won't feel very authentic looking at a single chip with 200+ pins. So you might as well do a firmware emulation using a humble MCU. An Arm-based one would be fast enough and could drive a display in real-time. Yes. The Ceres-1, 2 and 3 workstations were 32-bit. Personally I find 16-bit more interesting, more of a challenge. And it's surprising how hard it would be to create a 16-bit self-hosted computer today. 16-bit compilers seem to be MIA or too big to be run on a 16-bit CPU. Even Bill-Buzbee's Magic-1 doesn't self-host. That's one of the other interesting features of Lilith, its Modula-2 compiler is powerful enough to develop entire systems and it *can* self-host. In theory you could expand a Lilith to cope with, say, 16Mb of code, just by extending its base registers (F, G, L, H etc) to 256b boundaries (that wouldn't help much with data access though, since Lilith shares a single 64KW address space amongst all processes and programs). -cheers from Julz @P