[emmet]

> RISC-V is 25x slower than a top of the line Apple M-series chip

I don't think anyone has put the kind of money into a RISC-V processor that Apple has in order to develop the 3nm M4.

I was going to say it isn't an apples to apples comparison but I will restrain myself.

[sgerenser]

True, but who is going to put in the money? It's not a foregone conclusion that RISC-V will get enough investment to ever be competitive with state-of-the-art Arm or x86 chips.

[AnthonyMouse]

This question is sort of like, how is Linux ever going to be competitive with state-of-the-art proprietary Unix?

Suppose Facebook are tired of paying a premium to Cisco et al and decide to commission their own network equipment. That stuff doesn't have to be competitive with x86 on single thread performance, it just has to be reasonably power efficient. So they take some existing free RISC-V core and make a few improvements to it and use that. But they publish the improvements, because they're not actually trying to be a hardware OEM and if someone else takes their design and does the same thing, they know they get those improvements for their next generation.

So then that happens. Google want the same thing and make more improvements. Netgear use it in a consumer router, and they're not big enough to improve the chip, but they ship it in a product that sells a million units, so widespread use causes the community to optimize software for it and fix bugs. At this point Samsung or Qualcomm realize they only have to improve the SIMD support a little and they can stop paying ARM for their low and mid range phone SoCs. But if half of Android devices are now RISC-V and Qualcomm are already designing the high end cores themselves, why pay ARM for that either? So now it's in the high end phones, and someone starts putting the same chip into laptops.

All it really takes is for enough people to not want to pay ARM to create an ecosystem that allows everybody else to do the same thing. The free designs eat the low end of the market and then the high end uses the same architecture because why wouldn't it?

[bee_rider]

I’d imagine anyone who doesn’t have Apple super-duper special ARM license from the 90’s (or whenever) will be better served by RISC-V in the long run, right? Why deal with license issues?

I don’t know if it will happen, but it would be extremely funny if Intel cut off Arm and went with RISC-V. (False reports of the death of x86 have been around for decades, but it is bound to happen eventually, right?)

[emmet]

Absolutely fair point. I'm only pointing out the fact it hasn't been proved to be a limitation of the architecture yet.

Can't write off the first car only able to go 15km/h because your horse is able to do 40km/h.

[gjsman-1000]

Well, the good news for RISC-V (I say this with half honesty, half sarcasm), is that most of the RISC-V investment is happening in America and China. Their access to venture capital, talented engineers, and a decent economy makes the UK (where ARM's fighting from) look like Mississippi backwaters. ARM is disadvantaged against RISC-V geographically, economically, and politically; and judging by their interest in scare tactics a few years ago, I think they know it. Perfect conditions for a possible quick erosion of their technological lead.

[alephnerd]

> makes the UK (where ARM's fighting from) look like Mississippi backwaters

Most of ARM's design work is done in the US (Austin), India (Bangalore, Noida), and China (Beijing), though ARM China should basically be treated as a separate company at this point due to corporate shenanigans.

That said, in the chip design space (which tends to be concentrated in the US, Israel, India, and China), RISC-V has become much more popular for commodity embedded usecases because of the less restrictive licensing meaning better profit margins, which is allowing fabless chip startups to potentially leap ahead of ARM

[talldayo]

There's also the fact that ARM has a total of like 2 architecture licensees, and everyone else has to use piss-slow Cortex designs. If there was competition happening between ARM cores it would be a more interesting story, but right now ISA has taken a backseat while OEMs fight over TSMC access.

[wmf]

There's more like a dozen architectural licenses but they're mostly used for server chips that were canceled. The Cortex X925 is getting close to Apple/Nuvia BTW.

[gjsman-1000]

> has a total of like 2 architecture licensees

And that probably only happened because Apple co-founded ARM.

[brucehoult]

That was before the breakup of the USSR, let it go.

[panick21_]

Europe Supercomputer project is well funded and they are investing quite a bit. Large European industrials are also getting into RISC-V because they are building things like trains that they will have to maintain for 50+ years.

[throwway120385]

Yeah I came here to say something similar. This is ARM's game to lose, and they need to remember that their architecture was in the same situation as RISC-V is at some point. The only thing that stops RISC-V right now is that everyone is focused on ARM. If ARM gives their IP users a reason to switch by raising IP costs or making bad architecture decisions then RISC-V will take advantage of that to make inroads. History is full of incumbents that go on to lose their entire market when they get complacent.

[o_m]

Right now it looks like China will the one dominating RISC-V

[brucehoult]

Right now they are the most enthusiastic and putting in a lot of work, yes.

That's other people being short-sighted, not China doing anything wrong or sinister.

There are in fact quite a lot of exciting non-Chinese developments being announced recently, including at the RISC-V Summit that is on now, but those things will take several years to make their way in to the market.

[amelius]

> it isn't an apples to apples comparison

It's not about how great the teams behind these CPUs are.

It's about how great the CPUs are.

[AnimalMuppet]

Fine, but the CPU with more money to work on the architecture often winds up being the better CPU.

[exe34]

right now, because one of them had a lot of investment and the other less so.

[perihelions]

- "I was going to say it isn't an apples to apples comparison but I will restrain myself"

That's the ignoble rhetorical device of applephasis