Comment by tonysdg | Hacker News Reader

Comment by tonysdg | Hacker News Reader

Parent Full thread

That's the whole idea behind the RISC-V platform [1]. It's well-funded, well-designed, has amazing pedigree (David Patterson literally wrote the book on computer architecture). It's even licensed such that companies can use it in proprietary designs.

And yet, making the actual switch is a tremendously difficult task. On the software side alone, it requires recompiling every application you want to run on your platform. That doesn't even touch the cost of rolling the actual hardware. There are more than a few examples of this: Intel's Itanium, Oracle's SPARC, Berkeley's MIPS, Transmeta's Crusoe, etc. Sure, these all had niches (embedded systems, research hardware, a few high-end servers) -- but breaking out into the mainstream (a.k.a. like x86-64 and ARM/ARM64) is damn near impossible.

[1] https://riscv.org/

This would be a bit different. Itanium, Sparc, Crusoe, ARM... all these CPUs were offering is the promise of better performance in some aspect.

Nobody expects RISC-V to beat i7 or Ryzen in any benchmarks.

IMO its purpose is to be the in hardware what GNU/Linux was in software.

edit: GNU/Linux, not just Linux :)

> This would be a bit different. Itanium, Sparc, Crusoe, ARM... all these CPUs were offering is the promise of better performance in some aspect.

Crusoe and ARM were not about the promise of better performance, but of being much more energy-efficient for the intended purpose (though since a few years ARM tries to get into a (more) high-performance field).

Energy efficiency kind of falls under the "performance" category. Just not raw MIPS performance but "performance per watt".

What I wanted to say is that the major selling point of RISC-V would be its open-ness rather than anything performance (or power) related. In that respect it has already captured the attention of interested parties, regardless of that the actual performance will be.

I heard that so far riscv chips (a few were taped IIRC) were just too slow. I deeply believe that with simple and open silicon (cpu, gpu) you'd get a huge amount of visible benefits for end users that don't require Apple A10 levels of performance.

- Less bugs

- Potentially better applications since the lower layers are stable and don't require you to fight it (as an example, pre vulkan drivers were hell, linux software has to circle around that)

Reply on news.ycombinator.com