RISC-V's Increasing Influence | Hacker News Reader

RISC-V's Increasing Influence

semiengineering.com

24 pointsvoxadam8 comments

Whilst this article is Europe based, my understanding is that China is fully behind RISC-V so has strong backing to succeed.

riscv has no future. the big platforms want to control everything, and the small players, well, theyre small. it peaked in 2022, and has long since run out of steam.

regarding china, etc: the current political climate will end very soon. in the US, quasi-nationalist trump will be replaced with a pax-americana democrat. xi jinping will be replaced by a pro-west faction who will want to do business with the US. this will mean scuttling all their nationalist projects in favour of buying things from the US ("we're friends with the US now, no need for this second rate domestic stuff"): homegrown riscv replaced with arm cores, homegrown gpu stuff replaced with nvidia, etc. same thing goes for russia, and i dont think i even need to mention the middle east at this point in time.

10 years from now riscv will join the ranks of openrisc, but in the meantime i suppose there's no harm in 'distributing' academic grants and investor money.

Thank you Cassandra.. But you could.be right about all of these geopolitics and it still wouldn't change the fact that NVIDIA announced a billion RISCV chips in their GPUs so far or that China is not going to throw away its insurance policy if the US looks friendly for 4 years. I don't know what you are hoping people keep paying license fees on, but everyone from TSMC to the final consumer is going to be all the happier dividing that money up without a license whenever possible.

This is not what I am seeing at all.

The ISA war is really a two front issue: compute, and control. Control implementation vastly outnumbers compute. Heck, for any given compute motherboard with 4 or 32 cores, there is 10-50 control processors, each with 1-4 cores. A car has hundreds. Many lightbulbs have one or two cores. Basically everything electronic has an MCU these days, because a microcontroller is actually cheaper than a nand chip. Even advanced transistors as used in automotive switching often have a microcontroller, albeit a very rudimentary one.

In this world, riscV is eating ARM and everything else alive slowly but surely.

RiscV will take longer to mature into a good application processor, but it inevitably will. It will take over Arm for the same reason Arm is eating away at x86. My prediction is that in 20 years, it will be the predominant architecture for most sequential computing applications.

Self-reply here since the edit window timed out:

I think the driver here is cost. You don’t see 50mhz ARM cores for $0.07 for a reason, and that reason is that the licensing of the ARM ISA does not make that possible.

So at the low end, riscV is 1/2 to 1/3 the cost for comparable performance based on licensing costs alone. In a market where discrete components sell for thousandths of a dollar, a few cents is huge and a dollar is catastrophic. The difference between ARM and riscV can easily be a 40 percent reduction in BOM cost, or significantly improved features and performance at the same cost.

The CH32V003 is a nice RISC-V chip at a great price, and deservedly popular, but there is in fact an Arm chip in the same space, the Cortex-M0+ Puya PY32F002A, which has the same MHz but a little more RAM (3k) and flash (20k).

Being Chinese there has been some suggestion that they haven't actually paid for the CM0 core, but it's supported in Keil and I think listed somewhere on the main Arm site too, which tends to rule that out.

(WCH has already introduced successors to the CH32V003 with more RAM and flash at the same $0.10 price -- and even the CH570D with 96 MHz, 12k RAM, 240k flash, USB, and a 2.4 GHz packet radio -- but they are not as widely available yet)

Sorry.

    God save the ARM
    It's just an instruction stream
    There is no future
    In England's dreaming

    Don't be told what you want to want to
    And don't be told what you want to need
    There's no future, no future
    No future for you

If you'd said RISC-V has no future in 2021 you might have had some argument.

But then the money and experienced talent started pouring in. Tenstorrent "Ascalon" 8-wide core comparable to Apple M1 or AMD Zen3 is taping out right about now, they want to see it widely used, including in laptops (which I'd expect to see next year).

That's five years or so behind Apple, but pretty close behind Qualcomm.

They say they'll catch Apple / Intel / AMD performance in 2027. That'll be cores. SoCs and stuff you can buy in a shop will be a year or two later. But this decade.

There are a number of others hard on their heels, including open-source XiangShan from China.

Will they be running Windows or MacOS? No. But they'll be running Android, Fedora/RHEL, Ubuntu/Debian and they'll be in data centres and phone and tablets.

Heck ... there are new phones introduced in 2025 with only Arm A53 cores from 2012 e.g. using the Mediatek Helio G50 SoC with 8x A53 such as the Infinix Smart 9 HD, Tecno Spark Go 1S, Tecno POP 9 4G (Indian model).

The majority of RISC-V SBCs and laptops for sale at the moment are similar specs to those phones, but the better ones are 2x as fast with another factor of 2x or 4x coming in the next 12 months.

> [RISC-V] peaked in 2022, and has long since run out of steam.

Not a chance.

RISC-V has become very entrenched in the embedded space. We're probably going to see an ARM/RISC-V duopoly in embedded, long-term. I can see it becoming somewhat popular in the server/AI space also. But, really I don't see a future for RISC-V on desktop/mobile. Nobody can really compete with Apple/Samsung/Google, and smaller players can't really afford to be even slightly incompatible. Sure, we get more powerful RISC-V SBCs every year, but SBCs have always been for people who want to do embedded with style, or hobbyists. And sure, we get things like the Framework/Deepcomputing laptop mainboard, but again, that's for hobbyists.

Reply on news.ycombinator.com