Qualcomm to Bring RISC-V Based Wearable Platform to Wear OS by Google | Hacker News Reader

Qualcomm to Bring RISC-V Based Wearable Platform to Wear OS by Google

qualcomm.com

102 pointsfork-bomber43 comments

RV needs an ecosystem to survive. Google can afford the hockey-stick dev costs, but 95% of the rest of OEMs cannot. Hopefully Google makes their tooling open source, that would really help shave years of RV adoption.

RISC-V is rapidly building the strongest ecosystem.

Yet it’s extremely weak compared to ARM and x86 atm

Roughly a decade ago, that's what people said about ARM too: "A licensed RISC ISA? Why use that when MIPS and PowerPC and x86 exist?"

For the most part, all it took to put ARM on the board was to get it's core designs on newer silicon. RISC-V will likely be the same, with the added advantage that most software projects already support RISC arches through ARM.

I don't think you realize that we WANT RiscV to succeed. Stop being so defensive. It really doesn't have a solid ecosystem yet and the ISA is still in its infancy. Debug is awful and security is still in rev0.x stage. If you want it to be real, stop playing the internet "i win the argument" game, understand where RISCV is deficient, and use your energy to actually help.

I wanted ARM to succeed too. I have the original Rev2 RasPi board sitting 10 feet away from me, for all the good an ARMv6 SOC does me in 2023. Even back at release it was little more than a glorified embedded SOC that could boot Linux for giggles. I'm willing to acknowledge goofy trash for what it is.

RISC-V feels like it's coming around to the same track. Much like how Broadcom had to shoulder the software support of the Raspberry Pi, Qualcomm et. al are going to have to do roughly the same thing now. If the support on top of those cores is good enough, it won't really matter who makes the chip (a-la ARM's ecosystem). The current cores are certainly on the simple side, but that's the name of the RISC game. Simply having a capable microcontroller spec is enough to dent ARM's market share.

> have the original Rev2 RasPi board sitting 10 feet away from me,

ARM was already a magnitude or two more successful than RISC-V before the first Raspberry Pi even came out.

It looks unlikely[0].

But if you are still going to make that claim about ARM success in 2000 vs RISC-V now, then better source the data to support it.

0. https://riscv.org/news/2022/07/europe-steps-up-as-risc-v-shi...

ARM was the default and pretty much the only option for mobile devices for 20+ years. It was probably in a better position ~2000 than RISC-V is now yet look how long it took for it to become (somewhat) competitive in the PC and server markets.

>It was probably in a better position ~2000 than RISC-V is now

The subject was software ecosystem, and there was not much of an ecosystem for ARM until Raspberry Pi happened and the community jumped on it.

RISC-V ecosystem might not have ARM parity yet, but it is way further along than ARM was at the time Raspberry Pi launched.

I would argue RISC-V ecosystem is where ARM was barely 5 years ago, and catching up fast.

My prediction? ARM's software ecosystem will be in the rear mirror within 2 years.

If you only consider desktop/server software to be actual “software” then yes. Otherwise I don’t think that’s accurate

> My prediction? ARM's software ecosystem will be in the rear mirror within 2 years.

Why? To be honest I find the somewhat religious level fanaticism related to RISC-v somewhat baffling

>If you only consider desktop/server software to be actual “software” then yes. Otherwise I don’t think that’s accurate

I was replying to:

>yet look how long it took for it to become (somewhat) competitive in the PC and server markets.

As for:

>Why? To be honest I find the somewhat religious level fanaticism related to RISC-v somewhat baffling

I won't entertain the question, due to the loaded tail.

Issue is not dev cost, it's existing ecosystem. To drive a smartwatch ("smart" as in running a smartphone OS giving it 2 days battery life tops)[0], you need a full already existing ecosystem, so either iOS or Google/Android. Only Google can get Google to support RISC-V in Google/Android.

[0] there might alteady be less smart watches that are already running RISC-V but we can't know since they are closed.

It’s a supported dev target in the simulator, for what it’s worth.

Related/dupe from earlier today:

Snapdragon X Series chips are heading to PCs to take on Apple's processors

https://news.ycombinator.com/item?id=37899676

Qualcomm announces first-ever mass-market RISC-V Android SoC

https://news.ycombinator.com/item?id=37919907

In my opinion, all the RISC-V showing lately from Qualcomm is related to their Nuvia acquisition and the ensuing lawsuit with ARM regarding licensing Nuvia core designs. Qualcomm is most likely doing this to put pressure on ARM.

If/when the lawsuit is finished, I expect Qualcomm is deprioritize RISC-V projects.

But why? What does ARM offer that is superior?

Not really where OP was headed, but...

I would say RISC-V is still unmature in some areas. ARM is far more than ISA, for example the interrupt controller in a modern ARM SoC is more complex than a RISC-V core. Can anyone provide a viable RISC-V alternative with proper Linux support soon?

Another issue is performance. We simply don't know how RISC-V performs compared to latest ARMs.

That's why I believe it's a smart move to start with Android Wear as target-platform.

It narrows the scope to less I/O, smaller screen-resolutions, lower cost, no cellular modem, yet still the same OS-architecture and moderate/high sales-volume.

Good target to ramp up a new platform...

>for example the interrupt controller in a modern ARM SoC is more complex than a RISC-V core.

RISC-V has the AIA[0], a pretty good interrupt controller.

It is a requirement in the platform spec.

https://github.com/riscv/riscv-aia

High-end cores you can license?

Besides that ARM still seems to be superior to RISC-V in pretty much every way besides the fact that you have to deal with ARM itself.

Which might also be a positive thing in some cases i.e. designing your own cores from scratch should be much more expensive than licensing from ARM.

>High-end cores you can license?

Yet Ascalon and Alastor can be licensed, today, and ARM has nothing with comparable performance. LG already licensed Ascalon.

Nevermind ARM, Apple's current M2 CPUs doesn't compare to Ascalon.

I’m not sure hypothetical/imaginary (until one can prove otherwise) CPUs count

The context was:

>High-end cores you can license?

And they are cores you can license -and LG has licensed already[0]- i.e. not in any way imaginary or hypothetical.

Let's try and not move goalposts.

0. https://www.eetimes.com/jim-keller-on-ai-risc-v-tenstorrents...

An actual cpu running actual benchmarks or real world benchmarks seems like a very reasonable goalpost, not sure who is supposedly moving them

>not sure who is supposedly moving them

You, obviously.

Your goalpost was:

>High-end cores you can license?

And now you're trying to turn it into:

>An actual cpu running actual benchmarks or real world benchmarks

Which constitutes a (failed) attempt to move the goalpost, right after having been given exactly what you asked for.

It was perfectly obvious that I was talking about functioning CPUs.

But sure congratulations on winning your imaginary argument

Well, they might announce this publicly to help their license negotiations with ARM (rumor is that their license is up for renewal in 2024).

But Qualcomm is also a multi billion-dollar company which has no second supplier for their core technology. They used to license the ARM architecture to build their own designs (Scorpion, Krait), but lately the CPU-cores are mostly ARM-design. RISC-V would be a good candidate to reduce their business risk.

Whereas when trying to remove the litigation-PR from the Nuvia lawsuit, the case is about Qualcomm trying to transfer everything Nuvia has developed under a specific narrow ARM-license to the broad license of Qualcomm:

ARM originally gave favorable conditions to Nuvia, a startup which aimed to develop tech and compete in a field where ARM is weak (servers). Now Qualcomm, the biggest customer of the strongest revenue-market of ARM (mobile devices) has acquired that startup to repurpose their development and use it in mobile devices, PCs, automotive, VR,...

ARM has filed a lawsuit that this was never in scope of the license and that the tech should be destroyed: https://s3.documentcloud.org/documents/22273195/arm-v-qualco...

I want to ask this ignorant question

Other than being OS, cheaper... For the consumer do they need to know that it's risc v? Would a laptop instead of saying "Intel" or "ryzen" it would say risc v on the sticker... Otherwise it's still windows or ChromeOS?

Comparing apples and oranges. Ryzen is a product name, Intel a brand name. It would be similar to Snapdragon and Qualcomm, respectively.

You wouldn't see riscv on the sticker the same way you don't see x86-64 or arm64 on the sticker.

Intel put in a lot of legwork into ensuring that x86 is a first-class citizen on android, and this has far-reaching effects.

Beyond just building against the android runtime, it is now trivial to support additional architectures at every stage of the pipeline, even for native code.

Although x86 has the advantage that all dev workstations use it for VMs, at one point Google even had a MIPS version of the playstore, not to mention how different V7 and V8a are within ARM architectures.

> Although x86 has the advantage that all dev workstations use it for VMs

Mine is based on ARM

Even post-2020, I've seen vanishingly few production stacks that "just work" using ARM virtualization.

The most likely way you'd get this is by buying some cheap smartwatch. All of those are currently powered by Arm chips and it's unlikely that the general consumer knows anything about that.

Your laptop or desktop might well have RISC-V cores in it already. They are present in at least WD drives and some nVidia GPUs. But any time you need a cheap embedded core and don't want to pay licensing fees or deal with contracts is a case where you might choose RISC-V.

At the moment, there are only downsides to riscv as a consumer. It is largely used in areas which are otherwise invisible to the average consumer. Lower performance, less software availability, less reliable companies producing chips, etc. Eventually it may improve the diversity of the hardware ecosystem which will in term mean more competition and related benefits for consumers. Either way, software compatibility is the biggest reason a consumer may care directly.

Speaking of RISC-V. Earlier today I was thinking that I’d like to buy a desktop machine with a RISC-V CPU and came across one that seems aimed at computationally demanding stuff: https://www.crowdsupply.com/milk-v/milk-v-pioneer

Anyone had any experiences with the company behind MILK-V?

Seems nice. Would like to order.

I'd wait for tomorrow's announcement (SG2038) before making any decisions.

We already know SG2038 uses SiFive P670+X280.

That's the ratified Vector 1.0 spec, unlike SG2042.

If you have money to burn then sure. We have some on the way. However performance isn't that great, the core (C910) has a bunch of hardware bugs, and as the sibling comment notes Sophgo have just announced the successor.

Here's the announcement of the successor chip (not a successor to the Milk-V although I suppose that will come later): https://news.ycombinator.com/item?id=37927575 It's interesting although not surprising that they ditched T-Head.

This is such a big news for RISC-V adoption.

Are they designing the riscv cores themselves?

It looks like it, yes.

Qualcomm is hitting it out of the park.

Reply on news.ycombinator.com