Erik Engheim
2 min readJan 6, 2022

--

It will on desktops which are increasingly using specialized silicon. Just look at the M1 with neural engine, video encoding and decoding hardware, encryption/decryption hardware, vector processing units etc.

E.g. vector processing is very parallel. That means you want small cores because it allows you to have lots of them. A vector processor doesn’t need to have a big ISA. Thus RISC-V is perfect. It allows you to design really small cores which is what you want in this case.

This also is a win for server based chips such as Amazon Web Services. Each web request is generally not very performance demanding, but you need to handle a lot of them.

In a data center each rack server can typically maximum go to 250W. That works poorly for Intel processors which have massive power consumption.

Within that limit you can put in a RISC-V chip with way more cores. If you got 256 cores running on 250 W that will make you a lot more money than 64 Intel cores on 250W. Why? Because you can serve 4x as many users. That translates into lower service prices.

That would make a market for desktop RISC-V. You don’t really need high performance single thread desktop performance anymore that Intel can deliver. Even an iPad was fast enough for almost any task 6 years ago.

What you need is specialized hardware to speed up specialized tasks like AI, image processing, video encoding etc. For servers you want lots of cores.

So very high performance single thread general purpose computing isn’t a selling point anymore.

--

--

Erik Engheim
Erik Engheim

Written by Erik Engheim

Geek dad, living in Oslo, Norway with passion for UX, Julia programming, science, teaching, reading and writing.

Responses (1)