Somewhat related: I believe RISC-V is going to well in the coming decade. I've been trying to figure out what I can invest in to get exposed to it, but have been unable to find anything. If anybody had got any angles, I'd love to hear them!
They might, on the basis of it being an open unencumbered ISA that will see a lot of investment from parties trying to get rid of Intel/ARM dependencies.
Think of it a bit like what happened to JS. Being the open language the web runs, it saw immense competition from multiple particles and today we have a what should be a shitty dynamically typed prototype based scripting language outperform and outoptimize the likes of Java with its static classes and types in some scenarios.
I would be wary about comparing software to hardware in this way. Open source software is generally only successful in cases when it's a cost center not the actual end product.
I don't see how than can work when developing competitive CPU cores is extremely expensive. Why would anyone who does that make them free? (even licensing seems like a poor idea looking at ARMs business model).
Also the web ecosystem is still basically a pile of shit on a few key metrics compared.to (say) boring Microsoft-land stuff a decade or 3 ago, so I'm not so sure if that's such a good argument.
I meant outperform in a general sense. It could be on cost-effectiveness, especially for low-margin chips. It would mean he'd invest in companies that make this transition, but I expect almost all low-margin producers to make this shift. High-margin we're probably looking at proprietary ISA extensions that they don't open source.
Yeah. My thought is that RISC-V has the potential to be very competitive in terms of compute/watt and compute/$. At the very least versus x86, but hopefully also with arm
I'd expect it would expand the market for customized SoC/microcontroller parts.
We know there's already a lot of "built to purpose" RISC-V products ending up in embedded environments. When you have less obligations to a restrictive ISA grant, you can make something that fits your needs exactly.
So the business case is second-order:
* As has repeatedly been joked, fab companies who aren't locked to a captive/capricious owner (Intel) look appealing. It likely isn't about cutting-edge there, but the flexibility to handle diverse custom orders. Plenty of these products will be like "133MHz and a couple hundred thousand gates" that would be fine on 90nm or bigger.
* Firms that can move from "product" to "consultancy". There's a lot of brilliant work in existing ARM and RISC-V MCUs, but will existing firms be able to move away from "here's a matrix of features, order from the list" and into "Let us meet with your team, build a custom pick-and-mix and add a bespoke unit to meet your need for (smaller number than was historically common) product-specific MCUs?"
Is there something unique to RISC that makes it inexpensive to manufacture custom chips?
Otherwise, I'd imagine the cost of the mask set and validation to far, far outweigh any reduction in unit cost you might see from a custom chip. MCUs are already extremely cheap, especially if you're buying significant quantities.
The ISA is permissionless. You don't have to sign any legal documents, just download the specs, download some chip designs off github, and get going. The model is very different from x86 and Arm.
I'm guessing most of the value will be taken by TSMC. They don't care who wins :-) This move is about Intel Foundry Services (IFS) trying to get some of that action. IFS is now operated separately from the x86 part of the company.
Depends on where you think the value will accrue - Maybe cloud providers who are some of x86’s biggest customers today but would switch if workloads work there? But their bottom line is exposed to so many different things it’s not a clean bet.
For the processors designers themselves, I wouldn’t bet outside value accrues to them since the instruction set is open and competition could be fierce.
Another option would be to figure out who you think will be manufacturing the other peripherals needed to use RISC processors. Something like Supermicro on the server side (which benefitted hugely from a similar relationship making nvidia cards into servers), or find some supplier on whatever market you think will pop because of RISC-V (automotive, military systems, etc). Basically whoever turns these processors into finished systems is probably a good bet, but I haven’t done enough research to know who that would be.
Thank you. Yes, I agree that it's a good idea to look for companies that create divided systems based on RISC-V. It's primarily these I've had trouble to identify :)
There's a vast amount of ocean boiling to do first, unless such a thing just emulates x86 or aarm64.
It took until...tomorrow for me to be able to run my workloads on arm. That only happened (hasn't 100% happened yet) because we had RPi and then the M1 forcing all the dependency folks to ship arm builds. I've never seen a RISC-V docker container...
I believe RISC-V will recreate balkanization and UNIX wars of the past. There will be just to much incentive for corporations to introduce closed/proprietary extension to distinguish themselves from their competitors. So either users would be forced to use only core specification(that would be not as competitive as new extensions) or be forced to choose extended ISA from one of big players.
Yes. People upthread are talking about how corporations want to recreate the open software revolution with open hardware and this will (and already does) play out exactly like it has with open software: big vendors extract all the value from the product, contribute nothing back, and keep any cool features they come up with internal and proprietary.
This is the problem with arm, it’s this foundational technology that underpins everything from Grace to Dojo and TPU and yet arm doesn’t make any fucking money. They lost money during the first year of the pandemic (and the two years prior iirc!) despite the boom in electronics sales!
Nvidia buying arm for the synergy of forcing CUDA as the default arm graphics IP (and BYO for higher price) was the good ending. Nobody else is going to pay $40b for a company that makes 0.5b +/- 1b without some other synergy, and the other possibilities are companies like Samsung, tsmc, or Oracle with their own synergies. CUDA as an ai platform was absolutely a play that would be cashing out big right now if they had succeeded.
Well, if it’s not nvidia (and it seems any other buyout would be equally unlikely to get approval) then SoftBank is just gonna have to make the existing licensees pay up, and cash out the business that way (revenue and then IPO). So there’s a lot of licensees who are going to find an extra zero on their TCO of ARM usage.
That’s the reason you saw Qualcomm pitch a fit and start lying about the terms of their licensing deal last year. And google and meta and Amazon are equally upset. And they’re gonna start funding a competitor.
Those companies (and the bulk android market) mostly don’t care about performance, this is the “cheapest thing that works” market. The accelerator (TPU etc) is the thing that really matters now.
There may be a minimal amount of collaboration (ala Minix or CDDL) to build that common platform but that’s all they really care about, is some basic P-core and E-core designs. There is no incentive for Amazon to build processors that are way faster than googles, actually everyone is still selling vCPUs based on sandy bridge processing power. They’d rather make cheaper smaller CPUs (this is why zen4c and zen5c exist, in large part) than compete on performance or features. That’s just not where the secret sauce is, and it’s not where the revenue is.
So we are going to see the same thing as the software space, where vendors take the fruit of risc-v, build their own stuff around it, and pay nothing and contribute nothing back. Everything important will live in proprietary, non-portable IP that doesn’t need to be shared under open terms. The same problems as always when you BSD/MIT license.
The future is a google TPU that you are not allowed to purchase yourself, stamping on a human face endlessly. But the boot will be based on a core of open-source/open-license designs, how wonderful! (with the interesting parts living inside proprietary parts of the core ofc)
Nvidia/CUDA dominance is way better than google platform-as-service that you can’t even run on-prem if you wanted to. And for all you can say about them, nvidia has never stopped pushing to the next thing, even when they had an insurmountable lead, and even when nobody else believed in the goal or saw where they were trying to go. Google and Amazon don’t benefit from pushing the limit like this, they just want cheap arm.micro instances and a host for their proprietary accelerators.
The difference between BSD/MIT and GPL is fundamentally about who gets the freedom - BSD/MIT let developers do what they want with it, and that sometimes comes at the expense of user freedoms. And when google/Amazon/meta say they want free and open cores, they certainly don’t mean they’ll be giving away their IP.
That’s arm’s predicament in a nutshell. Their business model lets other companies extract all the value and revenue. They’re this foundational technology that google/Amazon/meta make billio...
Since RISC-V already has standard extensions for pretty much every feature that any other mainstream ISA has, it is pretty hard to see what is left for generally-applicable custom extensions that make enough difference to be worth getting locked in to.
The only Balkanization that has occurred so far is companies implementing draft versions of standard extensions because they don't want to wait years for the final version: e.g. the Kendryte K210 with priv ISA 1.9.1 [1], THead C906 and C910 implementing draft 0.7.1 of the Vector specification [2], THead implementing Physical Memory Attributes which didn't even have an official draft spec yet, THead and Andes implementing equivalents to (mostly) the eventual Zb or Zc extensions [3]
None of that has much in the way of lock-in as affected code can be either simply recompiled using the relevant official extension or else trivially ported.
[1] they appear to have simply used the current Berkeley Rocket snapshot at the time, 1.10 is what was ratified
[2] which document said "we think we're really really close to being ready to ratify this" but there turned out to be three more major revisions over 1.5 years. Nevertheless, many important library functions such as memcpy() are binary compatible between them, and forward-porting other 0.7.1 code to 1.0 is trivial.
[3] e.g. clz and popcount, shift-and-add for addressing calculations, byteswap, short opcodes for simple forms of byte load/store
I disagree. Its a totally different situation. With Unix you had lots of companies pushing Unix but not actually that much software.
We are in totally different world. For almost all compute task there are large amount of existing products, both open and closed. And all of them already target specific profiles.
If you develop something that is not in a standard profile you have a huge amount of work ahead of you to maintain a gigantic amount of software in forked state. Any many of those projects will simply not accept any proposals to up stream. And with the commercial software its worse because unless you achieve huge market dominance, those companies simply wont make a release for you.
RISC-V is in the process of standardizing pretty much everything that is commonly used. And any company pushing an alternative to those standards is pretty much dead in the water. The waste majority of software in the world is already fine with what's standard now. And over the next 1-2 years lots of stuff that some software needs will be standard.
Beyond that we are getting into highly specialized extensions. In those there will be competition and balkanization. But this simply wont effect the waste majority of users. Those things will be mostly about proprietary AI acceleration functions, special data types and so on. This has already happened but its simply not relevant to the waste majority of software in the world.
But all those companies understand that if their thing isn't gone be standard, they will have issues in the long run. So in reality what we are seeing is that all these companies are involved in the standards discussion and try to find one standard. They don't want to fight each other when Nvidia, Intel, AMD are much better targets.
Invest in companies making esoteric accelerators, like Fujitsu, Marvell, Broadcomm, probably many more I don't know about.
RISC-V is going to show up more and more in heavily customized stuff like HPC chips or networking accelerators before it starts replacing Xeons, as (right now) thats what its best suited for. Look up what Jim Keller said about RISC-V usage for Tenstorrent and how ARM just doesn't allow for the kind of customization these specialized designs need from a standardized ISA.
Look into the fabs that make RISC-V processors. At some point Intel Foundry Service will be spun off as a separate company, so invest in that when it happens.
To be honest, it’s a little hard to really take this seriously. The number of times I’ve watched Intel make Serious Declarations of Significant Strategic Realignments in the last decade or so only to watch them inevitably shed all the new divisions and product lines to Focus On Core Products makes it hard to really give this the attention it might warrant.
just looks like centrino or the multimedia PC standard: a basket of technologies (and like centrino/ultrabook, all intel ones) that define a standard for media PCs and set-top hardware that would support a certain distribution model.
now of course a lot of people didn't get what centrino was either (or what vpro is today!). it's a little hard to market these hardware standards when it's not something obvious like "the multimedia PC standard" that consumers grok. But it's the same kind of idea: we're going to need X, Y and Z hardware and feature support to do this workload or customer use-case. If your PC has it, then it's Centrino/Ultrabook/whatever!
But again I think it's true that people respond more intuitively to the idea of branding the product and not the platform. "Ultrabook" or “Multimedia PC Level 3” makes more sense as a concept to people than "Centrino" or "Viiv".
Centrino was an absolute slam-dunk success that did significantly bring x86 laptop computers forward (it may not be complete hyperbole to say that the Pentium M is the most important Intel CPU of this century), but in some ways it became obsolete by its own success: As all laptop manufacturers followed and gave us Ultrabooks that were amazing, Centrino as a brand didn't become a differentiator anymore, and Viiv just wasn't offering anything important enough - though the brand made perfect sense, given how hot Media Center PCs were for a short while.
And yet they are still delaying releases, canceling promising products, and making some questionable technical choices (like overhauling AVX512 and bifurcating the new standard, instead of uniting it like ARM and AMD did, among others).
I like Pat, I like what Pat says, I get plans are long term, but it still feels like the rubber is not meeting the road.
Falcon Shores, which was partially canceled/delayed. This is a big one.
Nervana, canceled in favor of Havana's Gaudi (though fortunately Gaudi 3 missed the axe).
And there are smaller things here and there, like the cancellation of iGPU-heavy designs like Broadwell or Skylake with eDRAM, and some rumors like the cancellation of the Arc Battlemage big die (albeit from a questionably reliable source).
I was at Intel when Krzanich was shown the door. After some time Bob Swan was announce as the new CEO. A bean counter, that was a first for Intel leadership. On the internal employee bulletin board the company posted how Bob is the right choice to which someone replied "If Bob was the right choice why did it take seven months?"
they are making only one bet now, (two if you account discrete gpu), making the production of the chip different division than the design of the x86 processors, and selling as service , most of this chips are arm or risc-v base so this is more a necessity to change the brand awardness and say "we will produce your non x86 chips", and we are in the committees of this technologies.
There is a rubicon coming up for Intel. At some point the fab side will be separated out into a separate company. Intel is currently leaving money on the table on the fab side under its current structure as nobody is going to trust Intel to make their latest and greatest while the chip design side is under the same roof.
The alternative is that the fab side goes down with Intel when the bottom drops out of x86.
This is exactly the reason. There are now (in all but corporate legality) two Intels, the one that makes x86 and the one that runs the foundries. Most RISC-V was going to be made on TSMC, and Intel Foundry Services (IFS) wants a piece.
No, I do not want Intel to touch anything associated with RISC-V. I do not want to see Intel MI cloned into RISC-V (and other backdoors), nevermind all the SMT Issues Intel is having.
Will Nvidia do something similar? If not dig themselves, could they actually design and license derivatives of Risc-v? I don't know how licensing works here.
ARM is preparing to IPO next week, their main business is collecting license fees. Here are last year’s income statement:
Revenue: $2.67B (down from 2.7)
Gross profit margin: 96%
Net Profit: $524m
And they are planning to IPO at a valuation of 60-70 billion dollars! Thats a P/E of 130 and a P/S of 25. Frankly this is ridiculous for a company that collects license fees, there is no potential for massive growth, all smartphones are using ARM chips already and their biggest client has a perpetual license already. Why is it valued like a tech startup?
This is Softbank dumping on retail and they somehow got Apple and Intel to buy a small piece of the IPO (700m combined).
They rely on people wanting to design their own ~~ships~~ chips
So if we assume that their ecosystem is healthly and mature at this moment,
then there's potential that customers wanting to design their own stuff will go to them.
So until RISC-V matures, then they'll be making profits, yup?
But is there growth potential? Are companies really interested in designing their own chips except a few startups that already license from them and probably google/amazon/whatever that need highly specialized chips, but already license from them?
Generally speaking, ARM sells chip blueprints in return for a royalty on every chip manufactured using their blueprints (or designed to be compatible). Their growth potential is more chips being made using their blueprints or getting a higher royalty per chip manufactured.
> then there's potential that customers wanting to design their own stuff will go to them.
There is plenty of evidence that anyone 'wanting to design their own stuff' are targeting RISC-V. Licencing ARM for that sort of thing is both costly and a pain in the arse and a distraction from actual processor development.
You dismiss startups, but the ripest growth market for processors is in AI accelerators and they are predominantly RISC-V based in the startup world.
RISC-V is growing fast and has a structural advantage to ARM's ISA monopoly, in that sense it is a real danger to ARM's business model.
The largest ecosystems that are deployed as production hardware right now, sure.
That's because designing a high performance core takes several years, and then getting it into a chip, tested, into mass production, and in to computers in stores or data centres takes another three or four years.
RISC-V has only existed as a frozen (ratified) base ISA for four years and a couple of months, and a lot of extra stuff that you want in applications processors for phones or PCs was ratified in November 2021, less than two years ago.
You're not going to see a lot of products with the late 2021 stuff until 2025 or 2026.
In terms of people starting projects now, or recently, that is indeed mostly RISC-V.
purely from investment perspective the scaling RISC-V development provides a hard cap on the how aggressive their pricing policy can be. At this point all major SoC companies have some interest/activity in RISC-V development so I doubt ARM is is in any form of monopoly position. with basically negates any significant stock growth opportunity. Given that era of cheap/easy money is done I really doubt anything good will come out for the IPO holders. this is just finding a bag-holder.
There is the PC business, but it's slowly dying. Seems a bit late in the game to drag it to a new architecture.
Datacenters are another place ARM might see growth, but Intel and AMD already competitive there. The biggest growth area might be Nvidia's Grace Hopper, but you might as well buy (already overpriced?) NVDA.
The reasons for IPOs are for investors to cash out (Saudi Aramco) or to raise money for growth (tech before ~2010). This feels like a cash-out. It's not that investors will necessarily lose money, it's that they're buying a much more mature company than they think.
The strange thing is that even if Softbank is effective at floating at this valuation (apparently by having 28 banks hawking the IPO and having a small float), there is almost no prospect of it remaining at that valuation, and Softbank is left holding the vast majority of the shares.
I don't see how this doesn't end in tears. It seems to be mostly wishful thinking.
> somehow got Apple and Intel to buy a small piece of the IPO (700m combined).
Well Apple makes more sense considering M* chips are ARM based and they have signed until 2040, though that could change at some point.
Intel might be using this as more a way to get "intel" so to speak. Not a bad idea to have some investment leverage.
I could see ARM after going public increasing the cost of licensing and some of the larger players would be ok with that to prevent competitors. Seems like lots of leverage plays here.
This may seem like a dumb question: Do we know how much Intel is paying? I can understand Intel wanting to invest.... but at what price and how much? Because I could imagine Softbank (who seem pretty desparate at this point) punting off big chunks of the company at a discount simply to try and trick retail into inesting. It's certainly not ethical, is it legal?
Anyone have some insight into why RISC-V seems to be getting traction?
Has there not been prior attempts to make an open source cpu?
Is there a lot of skill in making the ISA? It seems to me (naively) that most talented EE students could probably come up with their own ISA, or is there some "magic" in the RISC-V one?
I assume that there are a reference implementations in VHDL/Verilog and Cadence and good support in compilers. Is this what pushed through, where others failed?
The big part is royalty-free open-source licenses. If you don't want to pay ARM for an ISA and RISC-V is good enough for your needs then RISC-V might be for you.
Plus there are limitations on which countries the latest ARM designs can be shipped to. RISC-V has no such limitation. The RISC-V Foundation is based in Switzerland and so isn't subject to US trade sanctions.
ARM also license chip designs that implement the architecture. Not quite sure, but I think they'll do (for a fee) much of the process of making a chip that doesn't involve the fab.
ARM had some blunders with licensing changes for one, and then the split with ARM China was bad. Price plays a part. The next thing is that several open designs, including the reference design from Berkley, did very well at optimizing the most common instructions on which easily 90% of computation actually happens. As a result, very early RISC-V silicon showed seriously good performance at extremely low cost. As things are now, RISC-V is becoming increasingly competitive with ARM and x86 with amazing rapidity. There will be a few more hurdles in making RISC-V as performant as x86 or ARM, but given the ISA's open source nature, many eyes will make all bugs shallow.
RISC-V is not an open source CPU, it is an open and license-free Instruction Set.
A number of open source RISC-V CPU core implementations do exist -- including the THead C910, currently the highest performance RISC-V CPU you can buy at retail, used in the quad core TH1520 chip (Sipeed Lichee Pi 4A, BeagleBoard Ahead, Milk-V Meles, Roma laptop) and 64 core SG2042 chip (Milk-V Pioneer, an as yet unnamed dual socket board from PerfXLab, others coming soon). But there is no requirement to make RISC-V core open source.
There have certainly been other attempts to create an open ISA or CPU, either from scratch or else relicensing a previously proprietary design.
The problems with them which caused the RISC-V guys to make something new rather than use something existing included:
- unsuitable license e.g. GPL
- 32 bit only (OpenRISC, SPARC)
- all the opcode space already used, no room for future extension
- poor code density or other ISA problems e.g. delay slots
A number of existing ISAs have been made more open since RISC-V was developed, including POWER and MIPS, but have been basically too late and also simply too restrictive and even not irrevocably open e.g. MIPSOPen which was not all that open and was re-closed after just six months (and MIPS make RISC-V now).
It's true that it's very easy to make up your own ISA, but it's hard to get it to be both good and also not similar enough to something else to get you sued.
And once you have your own ISA where does the software come from? Making or porting compilers and operating systems and everything is a task many orders of magnitude larger than designing an ISA or CPU core.
> Anyone have some insight into why RISC-V seems to be getting traction?
Because its a a change in business model.
Its simple, in the past you had to pick a chip vendor and then make your software work on that. So you decide to get MIPS or ARM and then put your software on that.
RISC-V is different, you decide on RISC-V and you can start working on that. Then you can decide what vendor you want to get, or go with open source or do it yourself.
So in the past it was 'vendor -> isa', no its 'isa -> vendor'.
So bunch of guys in a big company lab can simply take an open source chip, throw it on an FPGA, develop the software and tell their manager "We can ship this, we just need to select a vendor for the actual chip". And then after 2 years if the vendor sucks, just switch vendor.
If you go to ARM. First there are lawyers and it can take months. Then once your software is on ARM, well now they have you by the balls.
RISC-V is free and vendors compete on an open playing field.
> Has there not been prior attempts to make an open source cpu?
As others have said, its not a CPU. Its an ISA specification. The RISC-V Foundation does not design chips.
As the other commenter has explained, there were other open ISA but they all had various technical and historical issues.
If you want actual open implementation of those specification, look at Chips Alliance, LowRisc or CORE-V for example.
> Is there a lot of skill in making the ISA?
Yes. Not for the base ISA, but if you want to have Crypto, Vectors, Atomics and so on and so on, its a huge effort. Yes some students designed the core, but RISC-V is far more then that by now.
But the real cost is actually that you have to port an insane amount of software, huge amount of man-hours to port everything and keep up with new extensions.
> I assume that there are a reference implementations in VHDL/Verilog and Cadence and good support in compilers. Is this what pushed through, where others failed?
There was just demand for something like it. Other efforts had historical issues or technical issues.
Once Berkley pushed it, even before the standard, various people started adopting it and it took of like a rocket after that.
Yes. If you know nothing about investing, the best thing you can do is read "A Random Walk Down Wall Street". It's quick and easy and explains why beating the market by picking stocks is not a great way to invest.
Softbank are dumping their shares in ARM while at peak valuation, before RISC-V eats its market share (they saw it coming, that's why they are rushing to IPO before rise of RISC-V).
If you don't know anything about finance or stocks then you are EXACTLY the retail customer they are looking for!
SoftBank paid $32 billion for Arm in 2016. People at the time were saying that was waaay too much, and that was even without 99.9% of people in the industry having RISC-V on their radar yet.
A fair valuation of Arm, assuming that they're going to grow at a kind of industrial average rate from here on, would be 15 to 20 times annual profits. That's something like $8 to $12 billion. $50 billion is insanity.
Investing a sizeable chunk of your savings is gambling. In a couple of years you will know whether it was a good idea or not. And some so-called experts will tell you, that's what we said. And some other so-called experts will be very quiet.
95 comments
[ 3.1 ms ] story [ 162 ms ] threadThink of it a bit like what happened to JS. Being the open language the web runs, it saw immense competition from multiple particles and today we have a what should be a shitty dynamically typed prototype based scripting language outperform and outoptimize the likes of Java with its static classes and types in some scenarios.
I don't see how than can work when developing competitive CPU cores is extremely expensive. Why would anyone who does that make them free? (even licensing seems like a poor idea looking at ARMs business model).
Also the web ecosystem is still basically a pile of shit on a few key metrics compared.to (say) boring Microsoft-land stuff a decade or 3 ago, so I'm not so sure if that's such a good argument.
We know there's already a lot of "built to purpose" RISC-V products ending up in embedded environments. When you have less obligations to a restrictive ISA grant, you can make something that fits your needs exactly.
So the business case is second-order:
* As has repeatedly been joked, fab companies who aren't locked to a captive/capricious owner (Intel) look appealing. It likely isn't about cutting-edge there, but the flexibility to handle diverse custom orders. Plenty of these products will be like "133MHz and a couple hundred thousand gates" that would be fine on 90nm or bigger.
* Firms that can move from "product" to "consultancy". There's a lot of brilliant work in existing ARM and RISC-V MCUs, but will existing firms be able to move away from "here's a matrix of features, order from the list" and into "Let us meet with your team, build a custom pick-and-mix and add a bespoke unit to meet your need for (smaller number than was historically common) product-specific MCUs?"
Otherwise, I'd imagine the cost of the mask set and validation to far, far outweigh any reduction in unit cost you might see from a custom chip. MCUs are already extremely cheap, especially if you're buying significant quantities.
Other companies are already public, like Andes
Look at names of companies that give talks at https://events.linuxfoundation.org/riscv-summit/
For the processors designers themselves, I wouldn’t bet outside value accrues to them since the instruction set is open and competition could be fierce.
Another option would be to figure out who you think will be manufacturing the other peripherals needed to use RISC processors. Something like Supermicro on the server side (which benefitted hugely from a similar relationship making nvidia cards into servers), or find some supplier on whatever market you think will pop because of RISC-V (automotive, military systems, etc). Basically whoever turns these processors into finished systems is probably a good bet, but I haven’t done enough research to know who that would be.
disclaimer: this is pure speculation, I am not privy to any information, I am just a rando engineer.
It took until...tomorrow for me to be able to run my workloads on arm. That only happened (hasn't 100% happened yet) because we had RPi and then the M1 forcing all the dependency folks to ship arm builds. I've never seen a RISC-V docker container...
This is the problem with arm, it’s this foundational technology that underpins everything from Grace to Dojo and TPU and yet arm doesn’t make any fucking money. They lost money during the first year of the pandemic (and the two years prior iirc!) despite the boom in electronics sales!
Nvidia buying arm for the synergy of forcing CUDA as the default arm graphics IP (and BYO for higher price) was the good ending. Nobody else is going to pay $40b for a company that makes 0.5b +/- 1b without some other synergy, and the other possibilities are companies like Samsung, tsmc, or Oracle with their own synergies. CUDA as an ai platform was absolutely a play that would be cashing out big right now if they had succeeded.
Well, if it’s not nvidia (and it seems any other buyout would be equally unlikely to get approval) then SoftBank is just gonna have to make the existing licensees pay up, and cash out the business that way (revenue and then IPO). So there’s a lot of licensees who are going to find an extra zero on their TCO of ARM usage.
That’s the reason you saw Qualcomm pitch a fit and start lying about the terms of their licensing deal last year. And google and meta and Amazon are equally upset. And they’re gonna start funding a competitor.
Those companies (and the bulk android market) mostly don’t care about performance, this is the “cheapest thing that works” market. The accelerator (TPU etc) is the thing that really matters now.
There may be a minimal amount of collaboration (ala Minix or CDDL) to build that common platform but that’s all they really care about, is some basic P-core and E-core designs. There is no incentive for Amazon to build processors that are way faster than googles, actually everyone is still selling vCPUs based on sandy bridge processing power. They’d rather make cheaper smaller CPUs (this is why zen4c and zen5c exist, in large part) than compete on performance or features. That’s just not where the secret sauce is, and it’s not where the revenue is.
So we are going to see the same thing as the software space, where vendors take the fruit of risc-v, build their own stuff around it, and pay nothing and contribute nothing back. Everything important will live in proprietary, non-portable IP that doesn’t need to be shared under open terms. The same problems as always when you BSD/MIT license.
The future is a google TPU that you are not allowed to purchase yourself, stamping on a human face endlessly. But the boot will be based on a core of open-source/open-license designs, how wonderful! (with the interesting parts living inside proprietary parts of the core ofc)
Nvidia/CUDA dominance is way better than google platform-as-service that you can’t even run on-prem if you wanted to. And for all you can say about them, nvidia has never stopped pushing to the next thing, even when they had an insurmountable lead, and even when nobody else believed in the goal or saw where they were trying to go. Google and Amazon don’t benefit from pushing the limit like this, they just want cheap arm.micro instances and a host for their proprietary accelerators.
The difference between BSD/MIT and GPL is fundamentally about who gets the freedom - BSD/MIT let developers do what they want with it, and that sometimes comes at the expense of user freedoms. And when google/Amazon/meta say they want free and open cores, they certainly don’t mean they’ll be giving away their IP.
That’s arm’s predicament in a nutshell. Their business model lets other companies extract all the value and revenue. They’re this foundational technology that google/Amazon/meta make billio...
The only Balkanization that has occurred so far is companies implementing draft versions of standard extensions because they don't want to wait years for the final version: e.g. the Kendryte K210 with priv ISA 1.9.1 [1], THead C906 and C910 implementing draft 0.7.1 of the Vector specification [2], THead implementing Physical Memory Attributes which didn't even have an official draft spec yet, THead and Andes implementing equivalents to (mostly) the eventual Zb or Zc extensions [3]
None of that has much in the way of lock-in as affected code can be either simply recompiled using the relevant official extension or else trivially ported.
[1] they appear to have simply used the current Berkeley Rocket snapshot at the time, 1.10 is what was ratified
[2] which document said "we think we're really really close to being ready to ratify this" but there turned out to be three more major revisions over 1.5 years. Nevertheless, many important library functions such as memcpy() are binary compatible between them, and forward-porting other 0.7.1 code to 1.0 is trivial.
[3] e.g. clz and popcount, shift-and-add for addressing calculations, byteswap, short opcodes for simple forms of byte load/store
We are in totally different world. For almost all compute task there are large amount of existing products, both open and closed. And all of them already target specific profiles.
If you develop something that is not in a standard profile you have a huge amount of work ahead of you to maintain a gigantic amount of software in forked state. Any many of those projects will simply not accept any proposals to up stream. And with the commercial software its worse because unless you achieve huge market dominance, those companies simply wont make a release for you.
RISC-V is in the process of standardizing pretty much everything that is commonly used. And any company pushing an alternative to those standards is pretty much dead in the water. The waste majority of software in the world is already fine with what's standard now. And over the next 1-2 years lots of stuff that some software needs will be standard.
Beyond that we are getting into highly specialized extensions. In those there will be competition and balkanization. But this simply wont effect the waste majority of users. Those things will be mostly about proprietary AI acceleration functions, special data types and so on. This has already happened but its simply not relevant to the waste majority of software in the world.
But all those companies understand that if their thing isn't gone be standard, they will have issues in the long run. So in reality what we are seeing is that all these companies are involved in the standards discussion and try to find one standard. They don't want to fight each other when Nvidia, Intel, AMD are much better targets.
RISC-V is going to show up more and more in heavily customized stuff like HPC chips or networking accelerators before it starts replacing Xeons, as (right now) thats what its best suited for. Look up what Jim Keller said about RISC-V usage for Tenstorrent and how ARM just doesn't allow for the kind of customization these specialized designs need from a standardized ISA.
Disclaimer: Not investment advice.
Does anyone remember ViiV, or explain what exactly it was?
now of course a lot of people didn't get what centrino was either (or what vpro is today!). it's a little hard to market these hardware standards when it's not something obvious like "the multimedia PC standard" that consumers grok. But it's the same kind of idea: we're going to need X, Y and Z hardware and feature support to do this workload or customer use-case. If your PC has it, then it's Centrino/Ultrabook/whatever!
But again I think it's true that people respond more intuitively to the idea of branding the product and not the platform. "Ultrabook" or “Multimedia PC Level 3” makes more sense as a concept to people than "Centrino" or "Viiv".
https://en.wikipedia.org/wiki/Intel_Viiv
https://en.wikipedia.org/wiki/Centrino
https://en.wikipedia.org/wiki/Intel_vPro
https://en.wikipedia.org/wiki/Multimedia_PC
https://en.wikipedia.org/wiki/Ultrabook
I like Pat, I like what Pat says, I get plans are long term, but it still feels like the rubber is not meeting the road.
Like? I hope you aren't talking about Optane
I believe that was/is impressive tech.
You could have disk and ram in one stick aka kinda unified memory.
Think of it: instead of purchasing RAM+Disk, you purchase one stick and configure it as e.g 40% of the capacity works as a disk and 60% as a "RAM".
But saying that they "cancelled" doesn't tell the whole story.
IIRC: they couldnt find a fab which would produce them after Micron sold their fab.
Nervana, canceled in favor of Havana's Gaudi (though fortunately Gaudi 3 missed the axe).
And there are smaller things here and there, like the cancellation of iGPU-heavy designs like Broadwell or Skylake with eDRAM, and some rumors like the cancellation of the Arc Battlemage big die (albeit from a questionably reliable source).
The alternative is that the fab side goes down with Intel when the bottom drops out of x86.
They have a long way to go (it will require a herculean effort by the leadership) but nonetheless the upside potential is huge.
Revenue: $2.67B (down from 2.7)
Gross profit margin: 96%
Net Profit: $524m
And they are planning to IPO at a valuation of 60-70 billion dollars! Thats a P/E of 130 and a P/S of 25. Frankly this is ridiculous for a company that collects license fees, there is no potential for massive growth, all smartphones are using ARM chips already and their biggest client has a perpetual license already. Why is it valued like a tech startup?
This is Softbank dumping on retail and they somehow got Apple and Intel to buy a small piece of the IPO (700m combined).
They rely on people wanting to design their own ~~ships~~ chips
So if we assume that their ecosystem is healthly and mature at this moment,
then there's potential that customers wanting to design their own stuff will go to them.
So until RISC-V matures, then they'll be making profits, yup?
But is there growth potential? Are companies really interested in designing their own chips except a few startups that already license from them and probably google/amazon/whatever that need highly specialized chips, but already license from them?
What I'm missing?
There is plenty of evidence that anyone 'wanting to design their own stuff' are targeting RISC-V. Licencing ARM for that sort of thing is both costly and a pain in the arse and a distraction from actual processor development.
You dismiss startups, but the ripest growth market for processors is in AI accelerators and they are predominantly RISC-V based in the startup world.
RISC-V is growing fast and has a structural advantage to ARM's ISA monopoly, in that sense it is a real danger to ARM's business model.
At the moment there isn’t though. Really there is basically none. The largest ecosystems making their own chips are using ARM right now, not RISCV.
That's because designing a high performance core takes several years, and then getting it into a chip, tested, into mass production, and in to computers in stores or data centres takes another three or four years.
RISC-V has only existed as a frozen (ratified) base ISA for four years and a couple of months, and a lot of extra stuff that you want in applications processors for phones or PCs was ratified in November 2021, less than two years ago.
You're not going to see a lot of products with the late 2021 stuff until 2025 or 2026.
In terms of people starting projects now, or recently, that is indeed mostly RISC-V.
There is the PC business, but it's slowly dying. Seems a bit late in the game to drag it to a new architecture.
Datacenters are another place ARM might see growth, but Intel and AMD already competitive there. The biggest growth area might be Nvidia's Grace Hopper, but you might as well buy (already overpriced?) NVDA.
The reasons for IPOs are for investors to cash out (Saudi Aramco) or to raise money for growth (tech before ~2010). This feels like a cash-out. It's not that investors will necessarily lose money, it's that they're buying a much more mature company than they think.
I don't see how this doesn't end in tears. It seems to be mostly wishful thinking.
There's a war for the next platform. Meta uses ARM, while Magic Leap 2 for instance uses x64.
https://www.eenewseurope.com/en/meta-details-its-first-custo...
Well Apple makes more sense considering M* chips are ARM based and they have signed until 2040, though that could change at some point.
Intel might be using this as more a way to get "intel" so to speak. Not a bad idea to have some investment leverage.
I could see ARM after going public increasing the cost of licensing and some of the larger players would be ok with that to prevent competitors. Seems like lots of leverage plays here.
https://www.reuters.com/markets/deals/softbanks-arm-ipo-curr...
Has there not been prior attempts to make an open source cpu?
Is there a lot of skill in making the ISA? It seems to me (naively) that most talented EE students could probably come up with their own ISA, or is there some "magic" in the RISC-V one?
I assume that there are a reference implementations in VHDL/Verilog and Cadence and good support in compilers. Is this what pushed through, where others failed?
Plus there are limitations on which countries the latest ARM designs can be shipped to. RISC-V has no such limitation. The RISC-V Foundation is based in Switzerland and so isn't subject to US trade sanctions.
A number of open source RISC-V CPU core implementations do exist -- including the THead C910, currently the highest performance RISC-V CPU you can buy at retail, used in the quad core TH1520 chip (Sipeed Lichee Pi 4A, BeagleBoard Ahead, Milk-V Meles, Roma laptop) and 64 core SG2042 chip (Milk-V Pioneer, an as yet unnamed dual socket board from PerfXLab, others coming soon). But there is no requirement to make RISC-V core open source.
There have certainly been other attempts to create an open ISA or CPU, either from scratch or else relicensing a previously proprietary design.
The problems with them which caused the RISC-V guys to make something new rather than use something existing included:
- unsuitable license e.g. GPL
- 32 bit only (OpenRISC, SPARC)
- all the opcode space already used, no room for future extension
- poor code density or other ISA problems e.g. delay slots
A number of existing ISAs have been made more open since RISC-V was developed, including POWER and MIPS, but have been basically too late and also simply too restrictive and even not irrevocably open e.g. MIPSOPen which was not all that open and was re-closed after just six months (and MIPS make RISC-V now).
It's true that it's very easy to make up your own ISA, but it's hard to get it to be both good and also not similar enough to something else to get you sued.
And once you have your own ISA where does the software come from? Making or porting compilers and operating systems and everything is a task many orders of magnitude larger than designing an ISA or CPU core.
Because its a a change in business model.
Its simple, in the past you had to pick a chip vendor and then make your software work on that. So you decide to get MIPS or ARM and then put your software on that.
RISC-V is different, you decide on RISC-V and you can start working on that. Then you can decide what vendor you want to get, or go with open source or do it yourself.
So in the past it was 'vendor -> isa', no its 'isa -> vendor'.
So bunch of guys in a big company lab can simply take an open source chip, throw it on an FPGA, develop the software and tell their manager "We can ship this, we just need to select a vendor for the actual chip". And then after 2 years if the vendor sucks, just switch vendor.
If you go to ARM. First there are lawyers and it can take months. Then once your software is on ARM, well now they have you by the balls.
RISC-V is free and vendors compete on an open playing field.
> Has there not been prior attempts to make an open source cpu?
As others have said, its not a CPU. Its an ISA specification. The RISC-V Foundation does not design chips.
As the other commenter has explained, there were other open ISA but they all had various technical and historical issues.
If you want actual open implementation of those specification, look at Chips Alliance, LowRisc or CORE-V for example.
> Is there a lot of skill in making the ISA?
Yes. Not for the base ISA, but if you want to have Crypto, Vectors, Atomics and so on and so on, its a huge effort. Yes some students designed the core, but RISC-V is far more then that by now.
But the real cost is actually that you have to port an insane amount of software, huge amount of man-hours to port everything and keep up with new extensions.
> I assume that there are a reference implementations in VHDL/Verilog and Cadence and good support in compilers. Is this what pushed through, where others failed?
There was just demand for something like it. Other efforts had historical issues or technical issues.
Once Berkley pushed it, even before the standard, various people started adopting it and it took of like a rocket after that.
I've never invested before and I'm planning on purchasing ARM stock the day they IPO.
Is this a bad idea? I was planning on investing a sizeable chunk of my savings.
Most big ARM clients are dumping ARM for RISC-V.
Including, Qualcomm:
https://arstechnica.com/gadgets/2023/08/qualcomm-one-of-arms...
Google (Converting whole Android system to support RISC-V):
https://arstechnica.com/gadgets/2023/01/google-announces-off...
And Samsung :
https://research.samsung.com/news/Samsung-Electronics-Partic...
When you want to invest look for the underdog, not a company that has it's market share ready to be eaten alive by RISC-V in the coming 5-10 years.
SoftBank paid $32 billion for Arm in 2016. People at the time were saying that was waaay too much, and that was even without 99.9% of people in the industry having RISC-V on their radar yet.
A fair valuation of Arm, assuming that they're going to grow at a kind of industrial average rate from here on, would be 15 to 20 times annual profits. That's something like $8 to $12 billion. $50 billion is insanity.
Consider reading the "RISK FACTORS" section of their IPO filing: https://www.sec.gov/Archives/edgar/data/1973239/000119312523...
If you don't know much, then don't try to be smart. Invest in broad index bonds rather then individual stocks.
For my money, ARM is sinking in the water. RISC-V will eat its lunch far faster then anybody believes. ARM overall trajectory is down not up.