RISC-V set to announce 25% market penetration — open-standard ISA is ahead of schedule, securing fast-growing silicon footprint
Posted by 3G6A5W338E@reddit | hardware | View on Reddit | 43 comments
wintrmt3@reddit
Which market exactly? Even the post they are linking doesn't say.
Tuna-Fish2@reddit
All cpus.
(Well over 90% of all cpus sold are in microcontrollers.)
RISC-V is doing really well in all the markets where you want a passable cpu core at the lowest cost possible. They now have good enough implementations available and sufficient tooling support that for majority of the market there is no sensible reason to pick anything else if designing a new product.
Quatro_Leches@reddit
zero presence in consumer products main processors.
Fit_Flower_8982@reddit
This is very arbitrary. It is entirely understandable and predictable that it takes a long time to reach critical mass and enter the consumer market; that is no reason to dismiss it. In fact, if it weren't for smartphones, ARM would surely have followed the same path.
Quatro_Leches@reddit
no, EU+US companies wont adopt it for mass market for obvious reasons
nanonan@reddit
If the price/performance is there and the software ecosystem is built up I don't see why anyone would avoid it.
NerdProcrastinating@reddit
High performance RISC V cores are on the horizon from Tenstorrent's: Ascalon, and Callandor.
It looks like first product using their own cores will be out in 2026. Callandor in 2027, on paper, appears like it will be extremely competitive with top performing ARM & x86 cores.
bookincookie2394@reddit
Callandor may be IP-complete in 2027, but will probably take another one or two years to show up in any actual hardware.
NerdProcrastinating@reddit
Ah yes, good point.
I would love to see it appear in some Linux compatible consumer hardware, but I think your other comment above is spot on with the difficulty of going from core IP to full chip at volume. Hopefully their chiplet strategy will make it easier if they can partner with someone for the rest of the chip.
bookincookie2394@reddit
Plenty of small startups are designing RISC-V core IP. What's harder is to design chips around those cores that people would be willing to buy at scale.
Exist50@reddit
I think it will be very interesting if Qualcomm ends up switching to RISC-V. If they can get Android to treat it as a first class citizen, that opens many doors. Especially if Google actually follows through with their push for Android as a PC OS.
monocasa@reddit
I don't think Qualcomm is switching anytime soon unfortunately. They just won their main lawsuit with ARM and have already announced next gen nuvia "inspired" cores with the ARM ISA. Not a peep regarding anything RISC-V.
The whole thing seems to be a negotiating tactic that worked, and has since been shelved.
Exist50@reddit
They won the lawsuit, but their license is not perpetual. It extends till 2028 as-is, or 2033 with an additional (~negligible) fee. After that, unless something fundamental changes about ARM's stance, they'll probably once again try to up Qualcomm's rates significantly. Yes, this is a long enough timeline one might be tempted to punt on the issue, but if they're going to switch ISA, they can't really afford to wait till the last minute.
They're been pretty active in RVI. Perhaps most notably for their efforts to kill compressed, but in other areas as well.
monocasa@reddit
I guess we're talking about different time scales. That effort to kill the C extension was almost two years ago; we haven't heard a peep recently, despite new announcements (and releases!) of ARM processors in the same family of cores.
WUT_productions@reddit
Companies that make microcontrollers run on razor thin margins. Having a RISC-V chip let's you make your chips cheaper than anyone else. That's just market efficiency.
Wait_for_BM@reddit
RISCV removed the software side of the equation: compiler, development environment, software and OS, RTOS etc.
The processor cores will still cost money as either the companies develop them with their engineers (takes time + money) or license them already develop (money) from someone else. Or they could license the IP for short term and make their own in the mean while. At least they have a choice.
Exist50@reddit
Using a RISC-V core does still cost some money, but it's still significantly cheaper than ARM. To the point where it sounds like ARM's de facto abandoning the low end embedded market.
There are even open source RISC-V cores, though they're not particularly great.
monocasa@reddit
For its PPA, the Hazard3 is a pretty good core.
iBoMbY@reddit
I wouldn't call Tenstorrent a large company for example:
Quatro_Leches@reddit
Don’t they get billions in investment
MairusuPawa@reddit
Everyone with a Nvidia GPU already has a RISC-V chip in their computer.
symmetry81@reddit
They have a RISC-V core but not a RISC-V main processor.
moofunk@reddit
A "main processor" is going to start meaning something different eventually. Is it just one of many processor cores on a single motherboard or chip that can host the OS?
RISC-V lets you build computers with a variety of CPU cores from not much larger than microcontrollers to full 64-bit RVA23 compliant cores and hook individual cores up to a variety of accelerators.
From what I hear Tenstorrent is doing with their Blackhole Galaxy server is that it doesn't have a main CPU like their older Wormhole Galaxy, but uses the CPU cores on the accelerator chips to self host Linux. I could see this concept scaling down to desktop computers, where you buy a PCI accelerator card that can host its own OS, and that's all you need or a mobile system with 300-500 RISC-V cores on a single chip.
symmetry81@reddit
I think the relevant distinction here is whether you can install your own code to run on a given core or whether it only runs code provided to you by the device manufacturer. The Blackhole Galaxy would certainly count as something with a RISC-V main core, but NVidia GPU power management cores wouldn't.
waiting_for_zban@reddit
It depends how you define "Consumer Products". For example the Orange Pi has an SBC with RISC-V.
marmarama@reddit
The Orange Pi RV2 benchmarks about as fast as a Raspberry Pi 3, which is getting on for 10 years old, based on the Cortex-A53 which is a low-end core from 2012.
That's fast enough for a modest home router and plenty of embedded devices, but would be very disappointing for a lot of consumer devices.
RISC-V is for sure coming for ARM and x86 eventually, but there seems to be a real issue with RISC-V scaling up in terms of performance. Lots of promises, marketing claims and "theoretically" fast cores that then benchmark really badly.
It's been like that since the first commercial RISC-V application processor cores a decade ago, and the gap between them and ARM/x86 is, if anything, getting wider, not narrower.
Exist50@reddit
The flagship RISC-V cores are not in the Orange Pi, they're in stuff like Alibaba servers. And iirc, Rivos was rumored to have silicon with a proper high-perf RISC-V design, but we probably won't see it publicly given the Meta acquisition. Shame, given the team's Apple history.
reveil@reddit
How can it not benefit small companies and startups more? Large companies can design a custom ISA or just buy designs from ARM because they have the resources. RISC-V existing makes the barrier of entry smaller not larger. Of course the big companies will use it as well. Just like Godot engine being free can benefit big casinos and gambling it also benefits small indie developers.
Exist50@reddit
Well, they do have a point in that large companies benefit because they have the volume for the ARM licensing costs to be significant. But this is very much a "rising tide lifts all boats" kind of thing. Unless you're ARM.
3G6A5W338E@reddit (OP)
This simply wasn't going to happen before RVA23 hardware, which is expected to show up in early 2026.
SecretTop1337@reddit
For now, RISC-V is coming for ARM and AMD64 though.
pc0999@reddit
BTW
https://github.com/OpenXiangShan/XiangShan
LessonStudio@reddit
One other key part of RISCV is that it potentially end runs US export laws.
In many chips (STM32 included) there are little bits here and there licences from US companies; not just ARM. Things like the encryption. This then puts the product under "arms control" and other US bullying.
RISCV is a huge opportunity to end run this bullying.
Also, I really don't trust anything where the US and encryption overlap. I don't really trust anything where china and encryption overlap either; so my question is: How long before companies like ST start making RISCV chips?
CJKay93@reddit
How? ISAs aren't typically listed under export control lists; designs are.
LessonStudio@reddit
No, it doesn't matter if a US source contributes. If it is openly licensed, that is the end of that. Where it is a problem, even for STM32, as an example, is they sign licenses with organizations which do come under the bevy of US export rules. Those companies then pass those licensing restrictions onto companies like ST.
But, if, say, a French company was making RISCV based chips using open source designs, they are free and clear of the US bullying and assh*lery.
Also, keep in mind, with these licences, they can insist on reporting, etc. Which means an EU company may be reporting sales effectively to the US government. This isn't only about security in the form of military stuff, but that this data can easily end up in the hands of US companies. The USG has been caught passing SWIFT data on to corporations about what their foreign competitors are up to.
The less influence the US has on tech, the better off the world will be. Their BS about it not going to cuba, etc is nonsense. Those countries get all the "banned" tech they want. It is about control, and information. Instead of ordering directly from digikey, they just order from someone who orders from digikey.
CJKay93@reddit
Why would the design be open-source? Very few of the RISC-V designs out there are open source because it costs an extraordinary amount of money to design and fabricate a competitive one.
LessonStudio@reddit
More and more of the basics are available open source. Their final design might be closed, but there are lots of resources, which traditionally were licensed. Things like a sha256 implementation, or an AES one. These would come under all kinds of US "munitions" regulations.
symmetry81@reddit
Wouldn't RISC-V designs for, e.g., cellphone SoCs also have little bits here and there that have proprietary encryption that do the same thing? It's the ISA that's open, any particular RISC-V design will generally have the same copyright protections, bits of included IP, etc that any similar design would have.
LessonStudio@reddit
RISCV can be as open as the designer wants. I suspect there are more and more hardware encryption options being cooked up in open source. These would have no license controlled by anyone.
iBoMbY@reddit
I guess Europe is sitting this one out as well.
BlueGoliath@reddit
Is this it? Is it finally the year of the RISC-V CPUs?
hackenclaw@reddit
Just having China alone can help RISC-V adoption a lot faster. It is a huge market that is now constantly under US export laws.
AutoModerator@reddit
Hello 3G6A5W338E! Please double check that this submission is original reporting and is not an unverified rumor or repost that does not rise to the standards of /r/hardware. If this link is reporting on the work of another site/source or is an unverified rumor, please delete this submission. If this warning is in error, please report this comment and we will remove it.
I am a bot, and this action was performed automatically. Please contact the moderators of this subreddit if you have any questions or concerns.