According to the infinite posts about how RISC-V will revolutionize computing and how it’s the next big thing, the opposite should be happening.
So, how are we here, with downsizing, rather than the hypergrowth that has been prophesized?
SiFive was one of the loudest RISC-V outfits, but far from the only one. That said, it’s very much not a great look.
RISC-V is still destined to take over the embedded CPU market, just like Linux was destined to take over the server market, just not overnight.
Okay, but why? Like, what are some tangible reasons or metrics why this is likely to happen?
I keep seeing this point reiterated, yet there are no market trends reflecting it.
Market trends go slow in the beginning, but will pick up pace as it grows. If you look at the number of different implementations and how the support for them is growing over time, progress is looking quite good.
As for tangible reasons - the other poster already mentioned that the only competition is ARM, who charge money. There are other problems with ARM too.
First off, it’s one single company. If it goes under for whatever reason, what happens to the ISA? Who gets the rights? What happens during that time of flux where no one has the rights? A company like TI derives quite a lot of its revenue from ARM. If that IP was compromised, TI’s very existence could be threatened.
Another tangentially related problem is what happens if ARM decides it wants more money? Or if it decides it doesn’t want to sell it’s IP anymore? You get the same threats as above.
Building on top of the same pile, what if a competitor buys ARM out? They could cut you out of your market entirely. One would hope that market regulators would catch this, but that’s not a guarantee.
So if it’s such a big issue, why haven’t companies moved on already? Why now? What’s special about RISC-V?
The answer to that is simple - software. RISC-V being an ISA guarantees software compatibility across all implementations. No one wants to develop their own ISA, because no one will write software for it. However, once educational instititutes/large R&D centers have done the heavy lifting of getting things up and running, sw should be a solved problem. It is one thing all microarchitecture designers will agree on, so all the software guys can work once.