Ask HN: What do you think about RISC-V development?
My friend told me he's betting a lot on further development of RISC-V architecture. With China and maybe soon Russia have USA trade ban, and ARM and x86 are licensed to USA companies, it will further develop technological disequality in the world, or make middle-man countries earn more money..
So the question is, do you think RISC-V as an open source architecture emerge as real competitor to ARM? As I understand it, RISC-V is a great hope, but just one part of a
bigger, harder problem. I will certainly be playing with it, and using a RISC-V desktop as
soon as it's viable. Broadly though, what the tech world needs is a democratisation of
hardware. We need to move toward reusable, flexible, stable,
transparent and more standardised, chips (though not all of these
desires are necessarily compatible). We need simple trustable, auditable microprocessors free from
"management engines", user hostile enclaves and pay-per-core
shitfuckery. This isn't only about end user choice and privacy et
cetera, it's about long-term ecology, sustainability and resilience. It's no big stretch to imagine Intel announcing an AAE (Advanced
Advertising Engine) that steals cycles to insert ads directly into
your applications - then millions of discarded products ending up in
landfills as people revolt against technology hijacked to be tools of
manipulation and domination. Problem is, VLSI fabrication is a dangerous, expensive and messy
business. RISC-V goes some of the way to offering a "peoples'
microprocessor", but nobody is going to set up production in their
garage any time soon. Investment in SME boutique fabs would be much nicer than pouring money
into Intel. If we are going to bring manufacturing home, I'd like to
see services not unlike the old vinyl record business, where a small
company or group of hobbyists can buy small-run batches of 100 chips
designed by a web interface - drag and drop four RISC-V cores, some
cache, a neural processor, some audio DSP... and have it drop into a
standard package socket, boot, and be cryptographically verifiable as
the intended design with a simple JTAG scan. I guess we're a long long way from this "silicon Lego", but that's my
dream. Sounds really like long way.
Yes it'd be great to be able to have some "home brew" manufacturing process that could be useful.
Then making CPU could switch to distributed model. But I just read today of silicon wafer shortages, so story goes deeper... I'm curious will things "just work" or will it be like Apple's recent M1 chips where some programs were not compiling/working on it.