[anujdeshpande]

The 8266 was based off an xtensa core instead of an Arm core which is the norm in the industry. Espressif created a replacement for this wildly popular chip with the C3 and based it off RISC-V to keep up with the times and the community. No more toolchain complaints from the community, plus they get to use open standards like RISC instead of xtensa from Tensilica. The never-ARM approach from Espressif has worked out well for them over multiple other alternatives that they chose over the years.

If someone writes a book on the downfall of Arm in a few years, this is going to be a very interesting chapter IMO.

disclaimer - worked at Espressif for a bit (2017-2019)

[PinguTS]

The downfall of ARM?

You know that ARM is used in a number of microcontrollers? There are over a 100 of them in every modern car, also in Chinese ones. So there are a billion a year alone in that area. Then all the peripherals which use either 8051 or ARM based microcontrollers, like simple things like the keyboard controller.

While I think Risc-V has its benefits. I don't know how it compares today to the scalability of the Cortex-M series starting with the M0 and then following the Cortex-A series. But I see that it is still a long way to go par.

[regularfry]

Something can have massive market penetration and still be doomed. Nobody talks about the biggest fax machine manufacturer, after all.

Mind you, whether that applies to ARM is a bigger question, it's like asking whether sand is going obsolete. I have my doubts unless something radical happens in the RISC-V space commercially; I don't think Espressif going all-in would do it on its own.

[ta988]

The RP2040 shows that Arm is far from dead in the space of tiny microcontrollers. And more are coming.

[catiopatio]

The downfall of ARM? What? Are you kidding?

[anujdeshpande]

- As one of the replies said, I was talking mostly about the IoT space, but I think it's true for the non-IoT laptop/server space as well.

- Sure - looking at what Apple's silicon team is doing with ARM would make you think that ARM has nothing to worry about for a while. However, notice that Apple is the only one who is getting a lot done with ARM at the high end. Qualcomm, etc. aren't getting the same amount of performance. I believe that's because of the non-ARM stuff that Apple puts in their silicon efforts. Imagine in a few years, when the small companies designing and licensing RISC-V cores are not that small anymore - folks like Amazon (uses ARM cores in their servers) and Apple might have a serious alternative.

- I think open source models win in the long run, similar to Windows vs Linux, or closed version control software vs Git. The tooling around open source grows over a few years and then suddenly it feels like it's miles better than the closed options. I think similar stuff will happen with silicon.

- ARM themselves worry about this btw - not that long ago they created a website smearing RISC-V which was mostly full of FUD. Their employees revolted and made them get rid of that site: https://www.theregister.com/2018/07/10/arm_riscv_website/

[jkestner]

ARM should be worried. In anticipation of their new IPO, they changed their licensing model to be based on the value of the final device, not the chip itself. Manufacturers are profoundly unhappy with that, and hey, there happens to be ann alternative to invest in now.

Talk to someone who works at a RISC-V startup. There’s a lot of energy there. It’s going to take a while, but startups are working on every class of chip.

ARM changed the fee structure because growth has been flat, and this is a way to get money short term, not grow the addressable market. It will settle into Intel-like stagnation as a disruptor takes its markets. What's ARM's moat?

[adrianpike]

I believe OP is speaking specifically to the IoT space, which is still a bit of a stretch - but slightly more accurate than the massive ARM dominance we're seeing elsewhere.

[m00x]

RISC-V is a much better ISA for IoT since it's so much better aligned. You just need to use the parts that you want instead of having to put in the whole package. Plus, it's free!

[michaelt]

I think the ESP8266 in a way shows the opposite.

ST, Atmel and Microchip maintain massive product catalogues so users can choose between 4MHz, 10MHz, 16MHz and 20MHz for their 8-bit, 8-pin microcontroller, and choose whether they want 1750 bytes of program memory or 3500 - but for the same price customers can get a 160 MHz, 32-bit microcontroller with built in wifi?

Why would expressif pay the engineering costs and supply chain complexity costs to make worse products, when their flagship product is already a great price?

[TickleSteve]

ARM is most dominant in the microcontroller space with the Cortex-M devices. Literally billions of them are made every year.

https://www.arm.com/company/news/2021/02/arm-ecosystem-ships...

[bzzzt]

It's not about the number of ARM cores shipping today, but tomorrow. If RISC-V tools are about as good as ARM and the cores are cheaper it will become the preferred instruction set somewhere in the future.

Seems in this age of open source tools the instruction set is mostly commodity anyway.