This is nothing about ARMv9 the ISA but much more about their new CEO Rene Haas. Arm has always been pricing their design on the lower end, bundling GPU and other designs IP. I have long argued since they enter 64bit era their performance profile and profits does not align well especially when comparing to AMD and Intel.
Even with the increased pricing the Cortex X5 / X925 and upcoming X6 / X930 they are still pretty good value. Unless Apple has something big with A19 / M5 the X6 / X930 should be competitive with M4 already. I just wish they spend a little more money on R&D for the GPU IP side of things.
Hoepfully we have some more news from Nvidia in Computex 2025
AMD and Intel actually fabricate chips for sale to others (outsourced to TSMC in AMD’s case) and take the risks associated with that. ARM on the other hand is just an IP provider. They are not comparable. ARM should have kept its original strategy of aiming to profit from volume that enabled its rise in the first place. Its course change likely looks great to SoftBank’s investors for now, but it will inevitably kill the goose that lays the golden eggs as people look elsewhere for what ARM was.
That said, ARM’s increased license fees are a fantastic advocate for RISC-V. Some of the more interesting RISC-V cores are Tenstorrent’s Ascalon and Ventana’s Veyron V2. I am looking forward to them being in competition with ARM’s X925 and X930 designs.
RISC-V is not immune from license fees, unless you want to design a high performance core from the ground up. If you want something as capable as an M4, there is years of R&D to get to that level. I'm sure a big player could do just that in house, but many would license Si-Five or similar. It will be interesting to see if Qualcomm and the like would make a move towards RISC-V, given their ARM legal issues
There are an incredible number of companies designing their own RISC-V cores right now. Some of them are even are making some of their designs entirely open source so that they are royalty free. The highest end designs are not, but it is hard to imagine their creators not undercutting ARM’s license fees since that is money that they would not have otherwise.
As for Qualcomm, they won the lawsuit ARM filed against them. Changing from ARM to RISC-V would delay their ambition to take marketshare from Intel and AMD, so they are likely content to continue paying ARM royalties because they have their eyes on a much bigger prize. It also came out during the lawsuit that Qualcomm considers their in-house design team to be saving them billions of dollars in ARM royalty fees, since they only need to pay royalties for the ISA and nothing else when they use their own in-house designs.
I doubt open source designs are going to be competitive with closed source. Also, design is just part of the problem. There is a whole lot of other things you need to get a chip out. I do not think RISC-V chips will be cheaper than other architecture when you take everything into account.
It is funny that you should say that, considering that I was wondering this myself earlier today WRT the Hazard3 cores in the RP2350. It turns out someone did benchmarks:
The Hazard3 core was designed by a single person while the ARM Cortex cores were presumably designed by a team of people. The Hazard3 cores mostly outperforms the Cortex-M0+ cores in the older RP2040 and are competitive with the Cortex-M33 cores that share the RP2350 silicon. For integer addition and multiplication, they actually outperform the Cortex-M33 cores. Before you point out that they lost most of the benchmarks against the Cortex-M33 cores, let me clarify that the integer addition and multiplication performance matter far more for microcontrollers than the other tests, which is why I consider them to be competitive despite the losses. The Hazard3 cores are open source:
That said, not all RISC-V designs are open source, but some of the open source ones are performance competitive with higher end closed source cores, such as the SonicBoom core from Berkeley:
As for the other problem you cite, the RP2350 has both RISC-V and ARM cores. It is a certainty that if the ARM cores had not been present, the RP2350 would have been cheaper, since less die area would have been needed and ARM license fees would have been avoided.
So far, patent lawsuits have been more of a problem for those using ARM designs (Qualcomm) than those using RISC-V designs. The Raspberry Pi foundation, Western Digital and Nvidia have successfully put RISC-V designs into their products without any issues. The first two even made their core designs open source (see Hazard3 and SweRV).
My point is that if RISC-V takes off people will struggle to do decent implementations of it without stepping on the toes of the people already in the area.
I'd go so far as to say this is the entire SiFive strategy.
RISC-V already has taken off. There are billions of RISC-V cores shipped in consumer products every year. Adoption outside of the embedded MCU space is slower, but that is natural. Your FUD about SiFive is absurd. Hardware patents related to CPU design are typically ISA independent.
That has not stopped new CPU designs from being made for any architecture and will not stop RISC-V designs from being made. If this were an actual problem, no one could design CPUs.
> Patents tend to expire at different times around the world, plus there is the possibility of submarine patents. Without a declaration from Hitachi, adopting any processor design using their ISA is likely considered a legal risk.
If you combine this with your observation that CPU patents tend to be ISA independent then surely any widespread commercial deployment of RISC-V requires an assertion from everyone else in the semi industry that they do not in fact own patents on your implementation of it or it is likely considered a legal risk.
That or you just hold some things to different standards than others.
There is a history of industry litigation over people implementing others’ ISAs without their full blessing. The Qualcomm ARM lawsuit was the most recent example of this. There is less litigation over people designing CPUs using ISAs whose designers permitted reuse.
You keep trying to spread FUD concerning RISC-V. The issue you are trying to raise is one that if valid, would prevent anyone from designing a CPU, yet many do without legal issues. Hence, the issue you raise is invalid (by modus tollens).
SuperH is owned by Hitachi. You cannot use them without a license from Hitachi as far as I know. RISC-V is unique in that its creator permits anyone to make and use RISC-V cores royalty free. It also supports 64-bit, which SuperH never did.
In any case, you should probably stop writing before you shove your foot any deeper into your mouth.
You should apologize to the people reading your comments for wasting their time. It is clear you are clueless about RISC-V and your foot is well into your mouth.
As for the J2, its creator does not request licensing fees, but Hitachi might require them. Unlike RISC-V, the creator of SuperH (Hitachi) is not known to have declared the ISA to be royalty free. I am not aware of such a declaration and even if there was, it is irrelevant because there is no reason to use SuperH over RISC-V. Nothing about the J2 supports the FUD you are spreading about RISC-V.
> You should apologize to the people reading your comments for wasting their time. It is clear you are clueless about RISC-V and your foot is well into your mouth.
You're absolutely out of line.
> As for the J2, its creator does not request licensing fees, but Hitachi might require them.
"FUD". The whole point of the timing of the release of the J2 was it is based purely on now expired Hitachi patents, so they do not require any licensing fees.
Patents tend to expire at different times around the world, plus there is the possibility of submarine patents. Without a declaration from Hitachi, adopting any processor design using their ISA is likely considered a legal risk. Beyond that, SuperH just is not very interesting. It lacks 64-bit support and there is very little interest in it by the industry, so software support is not that great.
By the way, my comment telling you that you should apologize to the community received an upvote and likely will receive more. You really are wasting people’s time with your anti-RISC-V FUD.
If you take the time to read my comments thoroughly, you will notice that I always spoke to your behavior, and not to you personally. There has been nothing wrong with my behavior, which has been tame compared to how a number of others in the industry react when encountering things that are wrong or even upon mere disagreement. My only fault is that I do not sugarcoat things, which is hardly a fault in a technical forum where facts and logic are valued.
By the way, having one’s foot in one’s mouth is an idiom meaning you said something wrong, which refers to behavior. It being obvious you are clueless is a reference to your writing, which again, refers to behavior. Saying you should apologize to people for wasting their time is similarly a reference to your behavior, and you invited that criticism by demanding an apology in broken English.
Unfortunately the Internet is full of people who are very confident about things they don't actually have a mastered understanding on. It's not necessarily worthwhile to invest time and effort into interacting with everyone who stated their opinions.
China will likely be the country taking forward RISC-V and ditching Arm and x86 completely. With USA trying to stop other countries from using latest Chinese tech they are given more reason to ditch any and all propitiatory US tech. So over the next decade I expect RISC-V architecture to enter and flood all Chinese tech devices from Tvs to cars and everything else that needs a CPU.
I personally hope China get's competitive in the node size as well as I want gpu and cpus start getting cheaper every generation again as once TSMC got big lead over Intel/Samsung and Nvidia got a big lead over AMD prices have stopped coming down generation to generation for CPU's and GPU's
RISC-V is definitely gaining traction in China, but it does not have a monopoly on Chinese CPU core design:
* Loongson is pushing a MIPS derivative forward.
* Sugon is pushing a x86 derivative (originally derived from AMD Zen) forward
* Zhaoxin is pushing a x86 derivative (derived from VIA’s chips) forward.
There was Shenwei with its Alpha processor derivative, but that effort has not had any announcements in years. However, there is still ARM China. Tianjin Phytium and HiSilicon continue to design ARM cores presumably under license from ARM China. There are probably others I missed.
There is also substantial RISC-V development outside of China. Some notable ones are:
* SiFive - They are the first company to be in this space and are behind many of the early/current designs.
* Tenstorrent - This company has Jim Keller and people formerly from Apple’s chip design team and others. They have high performance designs up to 8-wide.
* Ventana - They claim to have a high performance core design that is 15-wide.
* AheadComputing - they hired Intel’s Oregon design team to design high performance RISC-V cores after the Royal Core project was cancelled last year.
* The Raspberry Pi foundation - their RP2350 contains Hazard3 RISC-V cores designed by one of their engineers.
* Nvidia - They design RISC-V cores for the microcontrollers in their GPUs, of which the GPU System Processor is the most well known. They ship billions of RISC-V cores each year as part of their GPUs. This is despite using ARM for the high end CPUs that they sell to the community.
* Western Digital - Like Nvidia, they design RISC-V cores for use in their products. They are particularly notable because they released the SweRV Core as open source.
* Meta - They are making in-house RISC-V based chips for AI training/inference.
This is a short list. It would be relatively easy to assemble a list of dozens of companies designing RISC-V cores outside of China if one tried.
USA has now started banning companies of other countries from using Chinese tech if the Chinese tech has US components its a big over reach but it will move Chinese tech companies to move away from any US propitiatory tech.
That is not what your link says, but regardless of the details, Chinese companies are free to do whatever they want if they have no interest in exporting their products outside of China. Many do not care about markets outside of China. It is unlikely that China will drop all other ISAs in favor of RISC-V, especially since x86 and ARM are just as dominant in China as they are in other countries.
But that is the thing China wants to move on to exporting high value items themselves instead of manufacturing it for others and letting them take most of the profits. The bans and stuff has just started but this will result in China moving towards RISC-V the same way export of latest node tech has resulted in China doing it themselves and rapidly catching up. If you read my original comment what I said was over the next decade China will move away from Arm and x86 for RISC-V. It takes years to plan and built devices 5-6 years from now we will find out what I am predicting comes true or not.
You should not reason about China as a monolithic entity. China has a population of 1.4 billion people. Some look outward while others look inward. Those looking outward are interested in RISC-V for certain things since it is not subject to U.S. export controls (so far).
China is unlikely to move away from x86 and ARM internally even in a 10 year span. The only way that would happen is if RISC-V convinces the rest of the world to move away from those architectures in such a short span of time. ISA lock-in from legacy software is a deterrent for migration in China just as much as it is in any other country.
By the way, RISC-V is considered a foreign ISA in China, while the MIPS-derived LoongArch is considered (or at least marketed as) a domestic ISA. If the Chinese make a push to use domestic technology, RISC-V would be at a disadvantage, unless it is rebranded like MIPS was.
Correct me if I am wrong, but in RISC-V's case, you would be licensing the core design alone, not a license for the ISA plus the core on top.
Right now, AFAIK only Apple is serious about designing their own ARM cores, while there are multiple competing implementations for RISC-V (which are still way behind both ARM and x86, but slooowly making their way).
VERY long-term, I expect RISC-V to become more competitive, unless whoever-owns-ARM-at-the-time adjusts strategy.
Either way, I'm glad to see competition after decades of Intel/x86 dominance.
Qualcomm has a serious development effort in their Oryon CPU cores. Marvel had ThunderX from the Cavium acquisition, but they seem to have discontinued development.
MediaTek and others using ARMv9 design and pricing, heck even Qualcomm are selling their SoC on Windows PC at cheaper price compared to Intel or AMD.
Even at a higher IP price their final product are cheaper, faster and competitive. There may be a strategy about leaving money on the table, but it is another thing about leaving TOO much money on the table. If Intel and AMD's pricing is so far above ARM, there is nothing wrong with increasing your highest performance core 's pricing.
I would not be surprised in a 2 - 3 years time the highest PC performance CPU / SoC is coming from Nvidia with ARM CPU Core rather than x86. But knowing Nvidia I know they will charge similar pricing to Intel :D
So far, Qualcomm is not paying the royalty rate hikes since they are selling ARM hardware using cores covered under the ARMv8 architectural license that they obtained before SoftBank started pushing ARM to improve profitability.
It is interesting that you should mention MediaTek. They joined the RISC-V Software Ecosystem in May 2023:
It seems reasonable to think that they are considering jumping ship. If they are designing their own in-house CPU cores, it will likely be a while before we see them as part of a mediatek SoC.
In any case, people do not like added fees. They had previously tolerated ARM’s fees since they were low, but now that they are raising them, people are interested in alternatives. At least some of ARM’s partners are paying the higher for now, but it is an incentive to move to RISC-V, which is no fee for the ISA and either no fee or low fee for IP cores. For example, the hazard3 cores that the Raspberry Pi Foundation adopted in the RP2350 did not require them to pay royalty fees to anyone.
Huawei is probably the one that possibly might move away from arm because they have their own operating system. The US could tighten their controls more and ban them from arm altogether- so far they are prohibited from arm 9.
I am not sure Huawei would go for riscv - they could easily go for their own isa or an arm fork.
As an almost exclusively microcontroller user of Arm's products, a big meh from me. v8 is still slowly rolling out. M33 is making headway but I was really hoping for M55 to be the bigger driver.
That folks are still making new Cortex-A7 (2011) designs is wild. A-35 doesn't seem to be very popular or better.
Cortex-M33 (2016) derives–as you allude to–from ARMv8 (2015). But yeah it barely seems only barely popular, even now.
Having witnessed some of the 9p's & aughts computing, I never in a million years would have guessed microcontrollers & power efficient small chips would see so little change across a decade of time!!
Isn’t there some dynamic at play where STM will put one of these on a board, that board becomes a “standard” and then it’s cloned by other manufacturers, lowering cost? (legality aside)
Some chips that have come out in the past 3 years with Cortex A7:
Microchip SAMA7D65 and SAMA7G54. Allwinner V853 and T113-S3.
It's not like a massive stream of A7's. But even pretty big players don't really seem to have any competitive options to try. The A-35 has some adoption. There is an A34 and A32 that I don't see much of, don't know what they'd bring above the A7. All over a decade old now and barely seen.
To be fair, just this year ARM announced Cortex-A320 which I don't know much about, but might perhaps be a viable new low power chip.
You can get a lot of mileage out of a Cortex-M7. NXP has some which run up to 1 GHz - that's a ridiculous amount of power for a "microcontroller". It'd easily outperform an early-to-mid-2000s desktop PC.
There are no similarities between Cortex-M7 and Cortex-A7 from the POV of obsolescence.
Cortex-M7 belongs to the biggest-size class of ARM-based microcontrollers. There is one newer replacement for it, Cortex-M85, but for now Cortex-M7 is not completely obsolete, because it is available in various configurations from much more vendors and at lower prices than Cortex-M85.
Cortex-M7 and its successor Cortex-M85 have similar die sizes and instructions-per-clock performance with the Cortex-R8x and Cortex-A5xx cores (Cortex-M5x, Cortex-R5x and Cortex-A3x are smaller and slower cores), but while the Cortex-M8x and Cortex-R8x cores have short instruction pipelines, suitable for maximum clock frequencies around 1 GHz, the Cortex-A5xx cores have longer instruction pipelines, suitable for maximum clock frequencies around 2 GHz (allowing greater throughput, but also greater worst-case latency).
Unlike Cortex-M7, Cortex-A7 is really completely obsolete. It has been succeeded by Cortex-A53, then by Cortex-A55, then by Cortex-A510, then by Cortex-A520.
For now, Cortex-A55 is the most frequently used among this class of cores and both Cortex-A7 and Cortex-A53 are truly completely obsolete.
Even Cortex-A55 should have been obsolete by now, but the inertia in embedded computers is great, so it will remain for some time the choice for cheap embedded computers where the price of the complete computer must be well under $50 (above that price Cortex-A7x or Intel Atom cores become preferable).
For cores included in FPGAs, sadly there are none better than Cortex-A53 and Cortex-A72, because there have been no significant upgrades to the families of bigger FPGAs for many years. However in that case you buy the chip mainly for the FPGA and you have to be content with whatever CPU core happens to be included.
On the other hand, for the CPUs intended for cheap embedded computers there are a very large number of companies that offer products with Cortex-A55, or with Cortex-A76 or Cortex-A78, so there is no reason to accept anything older than that.
Texas Instruments is not really representative for embedded microcontrollers or computers, because everything that it offers is based on exceedingly obsolete cores.
Even if we ignore the Chinese companies, which usually have more up-to-date products, there are other companies, like Renesas and NXP, or for smaller microcontrollers Infineon and ST, all of which offer much less ancient chips than TI.
Unfortunately, the US-based companies that are active in the embedded ARM-based computer segment have clearly the most obsolete lines of products, with the exception of NVIDIA and Qualcomm, which however target only the higher end of the automotive and embedded markets, by having expensive products. If you want something made or at least designed in USA, embedded computers with Intel Atom CPUs are likely to be a better choice than something with an archaic ARM core.
For the Intel Atom cores, Gracemont has similar performance to Cortex-A78, Tremont to Cortex-A76 and Goldmont Plus to Cortex-A75; moreover, unlike the CPUs based on Cortex-A78, which are either scarce or expensive (like Qualcomm QCM6490 or NVIDIA Orin), the CPUs based on Gracemont, i.e. Amston Lake (Atom x7000 series) or Twin Lake (N?50 series), are both cheap and ubiquitous.
The latest Cortex-A7xx cores that implement the Armv9-A ISA are better than any Intel Atom core, but for now they are available only in smartphones from 2022 or more recent or in some servers, not in embedded computers (with the exception of a product with Cortex-A720 offered by some obscure Chinese company).
Does anyone but Renesas even offer a Cortex M85 based MCU? Afaik the all the other high performance ARM based microcontrollers still use a Cortex M7 except for a few M55 based chips.
In general Renesas offers more modern microcontrollers than all the other vendors of MCUs, which have decreased a lot the rate of new product launches during recent years, but unfortunately they are also among the most expensive.
I have also not seen Cortex-M85 except from Renesas. Cortex-M55 is seldom an alternative to Cortex-M7, because Cortex-M55 is smaller and slower than Cortex-M7 (but faster than the old Cortex-M4 or than the newer Cortex-M33).
Cortex-M55 implements the Helium vector instruction set, so for an application that can use Helium it may match or exceed the speed of a Cortex-M7, in which case it could replace it. Cortex-M55 may also be used to upgrade an old product with Cortex-M7, if the M7 was used only because Cortex-M4 would have been too slow, but the full speed of Cortex-M7 was not really needed.
At a trade show I saw a chip coming out with DDR3L. Imagine a 2025 chip with RAM from 15? years ago. They said it's all that they needed. Probably have a perpetual license or something.
> I never in a million years would have guessed microcontrollers & power efficient small chips would see so little change across a decade of time
It's because the software ecosystem around them is so incredibly lousy and painful.
Once you get something embedded to work, you never want to touch it again if you can avoid it.
I was really, really, really hoping that the RISC-V folks were going to do better. Alas, the RISC-V ecosystem seems doomed to be repeating the same levels of idiocy.
Switching microcontrollers means you have a lot of work to do to redo the HW design, re-run all of your pre-production testing, update mfg/QA with new processes and tests, possibly rewrite some of your application.. and you need to price in a new part to your BOM, figure out a secure supply for some number of years... And that just assumes you don't want to do even more work to take advantage of the new chip's capabilities by rewriting even more of your code. All while your original CPU probably still does fine because this is embedded we're talking about and your product already does what it needs to do.
The RP2040 and RP2350 are fairly big changes from the status quo, although they are not very energy efficient compared to other MCUs. Coincidentally, the RP2350 is part of the RISC-V ecosystem. It has both RISC-V and ARM cores and lets you pick which to use.
RISC-V is even worse: The Cortex-M series have standardized interrupt handling and are built so you can avoid writing any assembly for the startup code.
Meanwhile the RISC-V spec only defines very basic interrupt functionality, with most MCU vendors adding different external interrupt controllers or changing their cores to more closely follow the faster Cortex-M style, where the core itself handles stashing/unstashing registers, exit of interrupt handler on ret, vectoring for external interrupts, ... .
The low knowledge/priority of embedded of RISC-V can be seen in how long it took to specify an extension tha only includes multiplication, not division.
Especially for smaller MCUs the debug situation is unfortunate: In ARM-World you can use any CMSIS-DAP debug probe to debug different MCUs over SWD. RISC-V MCUs either have JTAG or a custom pin-reduced variant (as 4 pins for debugging is quite a lot) which is usually only supported by very few debug probes.
RISC-V just standardizes a whole lot less (and not sensibly for small embedded) than ARM.
Being customizable is one of RISC-V’s strengths. Multiplication can be easily done in software by doing bit shifts and addition in a loop. If an embedded application does not make heavy use of multiplication, you can omit multiplication from the silicon for cost savings.
That said, ARM’s SWD is certainly nice. It appears to be possible to debug the Hazard3 cores in the RP2350 in the same way as the ARM cores:
> If an embedded application does not make heavy use of multiplication, you can omit multiplication from the silicon for cost savings.
The problem was that the initial extension that included multiplication also included division[1]. A lot of small microcontrollers have multiplication hardware but not division hardware.
Thus it would make sense to have a multiplication-only extension.
IIRC the argument was that the CPU should just trap the division instructions and emulate them, but in the embedded world you'll want to know your performance envelopes so better to explicitly know if hardware division is available or not.
I don't think that library refutes anything of what I said.
First of, that library requires you to fundamentally change the code, by moving some precomputation outside loops.
Of course I can do a similar trick to move the division outside the loop without that library using simple fixed-point math, something which is a very basic optimization technique. So any performance comparison would have to be against that, not the original code.
It is also much, much slower if your denominator changes for each invocation:
In terms of processor time, pre-computing the proper magic number and shift is on the order of one to three hardware divides, for non-powers of 2.
If you care about a fast hardware divider, then you're much more likely to have such code rather than the trivially-optimized code like the library example.
> It appears to be possible to debug the Hazard3 cores in the RP2350 in the same way as the ARM cores:
It is, but (as far as I understood it), they're using ARM SWD IP (which is a fine choice). But since their connection between the SWD IP and RISC-V DM is custom, you're going to need your adjust your debug probe software quite a bit more than between different Cortex MCUs.
Other vendors with similar issues (for example WCH) build something similar but incompatible, requiring their own debug probe. This is a solved problem for ARM cortex.
> It's because the software ecosystem around them is so incredibly lousy and painful.
This is reaching breaking point entirely because of how powerful modern MCUs are too. You simply cannot develop and maintain software of scale and complexity to exploit those machines using the mainstream practices of the embedded industry.
The RP2350 lets you choose between 2 RISC-V cores and 2 ARM cores. I believe it even supports 1 RISC-V core and 1 ARM core for those who like the idea of their microcontrollers using two different ISAs simultaneously.
Microchip Technology has a number of RISC-V options.
Depends on the task. My favorite example is a chip that has a lot more than a microcontroller onboard, but it's an old v7m. I need the rest and have to struggle with what they give. If it was RV, power PC, mips, whatever, I'd have to use it.
ARM used to be UK owned until Conservative government lack of foresight allowed it to be sold to Softbank, leaving AIM (UK's NASDAQ, part of LSE) despite being in the national interests, and security, to keep it British. Thanks Mrs May (ex-PM) for approving that one (it was the last regulatory hurdle, that it was not in national security interests, so had to go past her).
Of course Boris Johnson (the next PM) __tried to woo ARM back to LSE__ because they realised they fucked up, and of course what huge foreign company would refloat on the LSE when you have NASDAQ, or bother floating on both?
Can you imagine if America had decided to allow Intel or Apple to be sold to a company in another country? Same sentiment.
- Yep I'm a pissed off ex-ARM shareholder forced out by the board's buyout decision and Mrs May waving it through.
Before reading article: I would like to know if this architecture will help Linux close to Apple architecture efficiencies....
After reading article: I suddenly realize that CPUs will probably no longer pursue making "traditional computing" any faster/efficient. Instead, everything will be focused on AI processing. There are absolutely no market/hype forces that will prompt the investment in "traditional" computing optimization anymore.
I mean, yeah, there's probably three years of planning and execution inertia, but any push to close the gap with Apple by ARM / AMD / Intel is probably dead, and Apple will probably stop innovating the M series.
The 128- and 256-core ARM server chips (like from Ampere) are pushing server performance in interesting ways. They're economically viable now for trivially parallelizable things like web servers, but possibly game-changing if your problem can put that many general-purpose cores to work.
The thing is, there aren't that many HPC applications for that level of parallelism that aren't better served by GPUs.
It makes sense to focus. Efficiencies in CPU design are not going to see as large of an impact on user workloads as focused improvements on inference workloads. The average phone user will be happier for the longer battery life as the onslaught of ai workloads from software companies is likely not going to slow and battery life will be wrecked if nothing changes.
You think so? I posit that the deliverance of AI/ML (LLM/genAI) services and experiences are predicated upon "traditional computing" - so, there will be some level of improvement in this domain for at least quite some time longer.
M4 still has >2x better performance per watt than either of those chips. Of course they are pretty much ignoring desktop so they can’t really compete with AMD/Intel when power is not an issue but that’s not exactly new
M4 has ">2x better performance per watt" than either Intel or AMD only in single-threaded applications or applications with only a small number of active threads, where the advantage of M4 is that it can reach the same or a higher speed at a lower clock frequency (i.e. the Apple cores have a higher IPC).
For multithreaded applications, where all available threads are active, the advantage in performance per watt of Apple becomes much lower than "2x" and actually much lower than 1.5x, because it is determined mostly by the superior CMOS manufacturing process used by Apple and the influence of the CPU microarchitecture is small.
While the big Apple cores have a much better IPC than the competition, i.e. they do more work per clock cycle so they can use lower clock frequencies, therefore lower supply voltages, when at most a few cores are active, the performance per die area of such big cores is modest. For a complete chip, the die area is limited, so the best multithreaded performance is obtained with cores that have maximum performance per area, so that more cores can be crammed in a given die area. The cores with maximum performance per area are cores with intermediate IPC, neither too low, nor too high, like ARM Cortex-X4, Intel Skymont or AMD Zen 5 compact. The latter core from AMD has a higher IPC, which would have led to a lower performance per area, but that is compensated by its wider vector execution units. Bigger cores like ARM Cortex-X925 and Intel Lion Cove have very poor performance per area.
I guess that depends on your definition of “desktop”.
What that really means (I think) is they aren’t using the power and cooling available to them in traditional desktop setups. The iMac and the Studio/Mini and yes, even the Mac Pro, are essentially just laptop designs in different cases.
Yes, they (Studio/Pro) can run an Ultra variant (vs Max being the highest on the laptop lines) but the 2x Ultra chip so far has not materialized. Rumors say Apple has tried it but rather could get efficiencies to where they needed to be or ran into other problems connecting 2 Ultras to make a ???.
The current Mac Pro would be hilarious if it wasn’t so sad, it’s just “Mac Studio with expansion slots”. One would expect/hope that the Mac Pro would take advantage of the space in some way (other than just expansion slots, which most people have no use for aside from GPUs which the os can’t/won’t leverage IIRC).
I think this largely misses the point. Power users, so most of the users on HN, are a niche market. Most people don't need a hundred gigs of RAM, they need their laptop to run Powerpoint and a browser smoothly and for the battery to last a long time. No other manufacturer is anywhere close to Apple in that segment as far as I'm concerned.
> but i really want to have atleast 96GB in notebook, tablet.
in notebooks it's been possible for years. a friend of mine had 128gb (4x32gb ddr4) in his laptop about 4-6 years ago already. it was a dell precision workstation (2100 euros for the laptop alone, core i9 cpu, nothing fancy).
Nowadays you can get 64gb individual ddr5 laptop ram sticks. as long as you can find a laptop with two ram sockets you can easily get 128b memory on laptops.
regarding tablets... it's unlikely to be seen (<edit> in the near future</edit>). tablet OEMs tip their hats to the general consumer markets, where <=16gb ram is more than enough (and 96gb memory would cost more than the rest of the hardware for no real user/market/sales advantage)
This is nothing about ARMv9 the ISA but much more about their new CEO Rene Haas. Arm has always been pricing their design on the lower end, bundling GPU and other designs IP. I have long argued since they enter 64bit era their performance profile and profits does not align well especially when comparing to AMD and Intel.
Even with the increased pricing the Cortex X5 / X925 and upcoming X6 / X930 they are still pretty good value. Unless Apple has something big with A19 / M5 the X6 / X930 should be competitive with M4 already. I just wish they spend a little more money on R&D for the GPU IP side of things.
Hoepfully we have some more news from Nvidia in Computex 2025
AMD and Intel actually fabricate chips for sale to others (outsourced to TSMC in AMD’s case) and take the risks associated with that. ARM on the other hand is just an IP provider. They are not comparable. ARM should have kept its original strategy of aiming to profit from volume that enabled its rise in the first place. Its course change likely looks great to SoftBank’s investors for now, but it will inevitably kill the goose that lays the golden eggs as people look elsewhere for what ARM was.
That said, ARM’s increased license fees are a fantastic advocate for RISC-V. Some of the more interesting RISC-V cores are Tenstorrent’s Ascalon and Ventana’s Veyron V2. I am looking forward to them being in competition with ARM’s X925 and X930 designs.
RISC-V is not immune from license fees, unless you want to design a high performance core from the ground up. If you want something as capable as an M4, there is years of R&D to get to that level. I'm sure a big player could do just that in house, but many would license Si-Five or similar. It will be interesting to see if Qualcomm and the like would make a move towards RISC-V, given their ARM legal issues
There are an incredible number of companies designing their own RISC-V cores right now. Some of them are even are making some of their designs entirely open source so that they are royalty free. The highest end designs are not, but it is hard to imagine their creators not undercutting ARM’s license fees since that is money that they would not have otherwise.
As for Qualcomm, they won the lawsuit ARM filed against them. Changing from ARM to RISC-V would delay their ambition to take marketshare from Intel and AMD, so they are likely content to continue paying ARM royalties because they have their eyes on a much bigger prize. It also came out during the lawsuit that Qualcomm considers their in-house design team to be saving them billions of dollars in ARM royalty fees, since they only need to pay royalties for the ISA and nothing else when they use their own in-house designs.
I doubt open source designs are going to be competitive with closed source. Also, design is just part of the problem. There is a whole lot of other things you need to get a chip out. I do not think RISC-V chips will be cheaper than other architecture when you take everything into account.
It is funny that you should say that, considering that I was wondering this myself earlier today WRT the Hazard3 cores in the RP2350. It turns out someone did benchmarks:
https://icircuit.net/benchmarking-raspberry-pi-pico-2/3983
The Hazard3 core was designed by a single person while the ARM Cortex cores were presumably designed by a team of people. The Hazard3 cores mostly outperforms the Cortex-M0+ cores in the older RP2040 and are competitive with the Cortex-M33 cores that share the RP2350 silicon. For integer addition and multiplication, they actually outperform the Cortex-M33 cores. Before you point out that they lost most of the benchmarks against the Cortex-M33 cores, let me clarify that the integer addition and multiplication performance matter far more for microcontrollers than the other tests, which is why I consider them to be competitive despite the losses. The Hazard3 cores are open source:
https://github.com/Wren6991/Hazard3
That said, not all RISC-V designs are open source, but some of the open source ones are performance competitive with higher end closed source cores, such as the SonicBoom core from Berkeley:
https://adept.eecs.berkeley.edu/wiki/_media/eop/adept-eop-je...
As for the other problem you cite, the RP2350 has both RISC-V and ARM cores. It is a certainty that if the ARM cores had not been present, the RP2350 would have been cheaper, since less die area would have been needed and ARM license fees would have been avoided.
RISC-V implementations are going to prove to be absolute patent minefields.
Just because something is open source will not stop you from being stung during manufacturing, rather like how Android deployments are not free.
So far, patent lawsuits have been more of a problem for those using ARM designs (Qualcomm) than those using RISC-V designs. The Raspberry Pi foundation, Western Digital and Nvidia have successfully put RISC-V designs into their products without any issues. The first two even made their core designs open source (see Hazard3 and SweRV).
How are SiFive going to protect their IP when everyone is free to copy it?
Patents.
You're not free to copy SiFive's IP cores.
Open ISA != all implementations of it are free (although in RISC-V case, many are).
Sorry, that was poorly worded.
My point is that if RISC-V takes off people will struggle to do decent implementations of it without stepping on the toes of the people already in the area.
I'd go so far as to say this is the entire SiFive strategy.
RISC-V already has taken off. There are billions of RISC-V cores shipped in consumer products every year. Adoption outside of the embedded MCU space is slower, but that is natural. Your FUD about SiFive is absurd. Hardware patents related to CPU design are typically ISA independent.
> Hardware patents related to CPU design are typically ISA independent.
So that is merely the entire semiconductor industry patent portfolio that you will have to avoid.
That has not stopped new CPU designs from being made for any architecture and will not stop RISC-V designs from being made. If this were an actual problem, no one could design CPUs.
To quote you elsewhere in this thread:
> Patents tend to expire at different times around the world, plus there is the possibility of submarine patents. Without a declaration from Hitachi, adopting any processor design using their ISA is likely considered a legal risk.
If you combine this with your observation that CPU patents tend to be ISA independent then surely any widespread commercial deployment of RISC-V requires an assertion from everyone else in the semi industry that they do not in fact own patents on your implementation of it or it is likely considered a legal risk.
That or you just hold some things to different standards than others.
There is a history of industry litigation over people implementing others’ ISAs without their full blessing. The Qualcomm ARM lawsuit was the most recent example of this. There is less litigation over people designing CPUs using ISAs whose designers permitted reuse.
You keep trying to spread FUD concerning RISC-V. The issue you are trying to raise is one that if valid, would prevent anyone from designing a CPU, yet many do without legal issues. Hence, the issue you raise is invalid (by modus tollens).
Anyone is free to make a RISC-V CPU without infringing on SiFive’s IP.
Which in practice will mean free to make simplistic implementations using the lessons of twenty years ago.
If this was a winning strategy those open source implementations of SuperH cores would have been incredibly popular instead of dying in obscurity.
Not so simplistic, see the XiangShan HotChips presentation:
https://hc2024.hotchips.org/assets/program/conference/day2/2...
SuperH is owned by Hitachi. You cannot use them without a license from Hitachi as far as I know. RISC-V is unique in that its creator permits anyone to make and use RISC-V cores royalty free. It also supports 64-bit, which SuperH never did.
In any case, you should probably stop writing before you shove your foot any deeper into your mouth.
https://j-core.org/
> In any case, you should probably stop writing before you shove your foot any deeper into your mouth.
Apology expected.
You should apologize to the people reading your comments for wasting their time. It is clear you are clueless about RISC-V and your foot is well into your mouth.
As for the J2, its creator does not request licensing fees, but Hitachi might require them. Unlike RISC-V, the creator of SuperH (Hitachi) is not known to have declared the ISA to be royalty free. I am not aware of such a declaration and even if there was, it is irrelevant because there is no reason to use SuperH over RISC-V. Nothing about the J2 supports the FUD you are spreading about RISC-V.
> You should apologize to the people reading your comments for wasting their time. It is clear you are clueless about RISC-V and your foot is well into your mouth.
You're absolutely out of line.
> As for the J2, its creator does not request licensing fees, but Hitachi might require them.
"FUD". The whole point of the timing of the release of the J2 was it is based purely on now expired Hitachi patents, so they do not require any licensing fees.
Patents tend to expire at different times around the world, plus there is the possibility of submarine patents. Without a declaration from Hitachi, adopting any processor design using their ISA is likely considered a legal risk. Beyond that, SuperH just is not very interesting. It lacks 64-bit support and there is very little interest in it by the industry, so software support is not that great.
By the way, my comment telling you that you should apologize to the community received an upvote and likely will receive more. You really are wasting people’s time with your anti-RISC-V FUD.
> By the way, my comment telling you that you should apologize to the community received an upvote and likely will receive more.
I too was upvoted for asking for your apology.
I will not apologize for speaking facts, and nor should you, but it is your random unnecessary insults that are unacceptable.
That's me done with this. You clearly have your opinions, but your behavior has been a discredit to the community you apparently represent.
If you take the time to read my comments thoroughly, you will notice that I always spoke to your behavior, and not to you personally. There has been nothing wrong with my behavior, which has been tame compared to how a number of others in the industry react when encountering things that are wrong or even upon mere disagreement. My only fault is that I do not sugarcoat things, which is hardly a fault in a technical forum where facts and logic are valued.
By the way, having one’s foot in one’s mouth is an idiom meaning you said something wrong, which refers to behavior. It being obvious you are clueless is a reference to your writing, which again, refers to behavior. Saying you should apologize to people for wasting their time is similarly a reference to your behavior, and you invited that criticism by demanding an apology in broken English.
Unfortunately the Internet is full of people who are very confident about things they don't actually have a mastered understanding on. It's not necessarily worthwhile to invest time and effort into interacting with everyone who stated their opinions.
China will likely be the country taking forward RISC-V and ditching Arm and x86 completely. With USA trying to stop other countries from using latest Chinese tech they are given more reason to ditch any and all propitiatory US tech. So over the next decade I expect RISC-V architecture to enter and flood all Chinese tech devices from Tvs to cars and everything else that needs a CPU.
I personally hope China get's competitive in the node size as well as I want gpu and cpus start getting cheaper every generation again as once TSMC got big lead over Intel/Samsung and Nvidia got a big lead over AMD prices have stopped coming down generation to generation for CPU's and GPU's
RISC-V is definitely gaining traction in China, but it does not have a monopoly on Chinese CPU core design:
There was Shenwei with its Alpha processor derivative, but that effort has not had any announcements in years. However, there is still ARM China. Tianjin Phytium and HiSilicon continue to design ARM cores presumably under license from ARM China. There are probably others I missed.There is also substantial RISC-V development outside of China. Some notable ones are:
This is a short list. It would be relatively easy to assemble a list of dozens of companies designing RISC-V cores outside of China if one tried.USA has now started banning companies of other countries from using Chinese tech if the Chinese tech has US components its a big over reach but it will move Chinese tech companies to move away from any US propitiatory tech.
https://www.bis.gov/media/documents/general-prohibition-10-g...
That is not what your link says, but regardless of the details, Chinese companies are free to do whatever they want if they have no interest in exporting their products outside of China. Many do not care about markets outside of China. It is unlikely that China will drop all other ISAs in favor of RISC-V, especially since x86 and ARM are just as dominant in China as they are in other countries.
But that is the thing China wants to move on to exporting high value items themselves instead of manufacturing it for others and letting them take most of the profits. The bans and stuff has just started but this will result in China moving towards RISC-V the same way export of latest node tech has resulted in China doing it themselves and rapidly catching up. If you read my original comment what I said was over the next decade China will move away from Arm and x86 for RISC-V. It takes years to plan and built devices 5-6 years from now we will find out what I am predicting comes true or not.
You should not reason about China as a monolithic entity. China has a population of 1.4 billion people. Some look outward while others look inward. Those looking outward are interested in RISC-V for certain things since it is not subject to U.S. export controls (so far).
China is unlikely to move away from x86 and ARM internally even in a 10 year span. The only way that would happen is if RISC-V convinces the rest of the world to move away from those architectures in such a short span of time. ISA lock-in from legacy software is a deterrent for migration in China just as much as it is in any other country.
By the way, RISC-V is considered a foreign ISA in China, while the MIPS-derived LoongArch is considered (or at least marketed as) a domestic ISA. If the Chinese make a push to use domestic technology, RISC-V would be at a disadvantage, unless it is rebranded like MIPS was.
They've already exfiltrated Arm's IP and began designing their own Arm cores. Is there a need for them to switch?
Correct me if I am wrong, but in RISC-V's case, you would be licensing the core design alone, not a license for the ISA plus the core on top.
Right now, AFAIK only Apple is serious about designing their own ARM cores, while there are multiple competing implementations for RISC-V (which are still way behind both ARM and x86, but slooowly making their way).
VERY long-term, I expect RISC-V to become more competitive, unless whoever-owns-ARM-at-the-time adjusts strategy.
Either way, I'm glad to see competition after decades of Intel/x86 dominance.
Qualcomm has a serious development effort in their Oryon CPU cores. Marvel had ThunderX from the Cavium acquisition, but they seem to have discontinued development.
Yes, but the playing field is different. Anyone can become a Risc-V IP provider and many such companies have already been created.
MediaTek and others using ARMv9 design and pricing, heck even Qualcomm are selling their SoC on Windows PC at cheaper price compared to Intel or AMD.
Even at a higher IP price their final product are cheaper, faster and competitive. There may be a strategy about leaving money on the table, but it is another thing about leaving TOO much money on the table. If Intel and AMD's pricing is so far above ARM, there is nothing wrong with increasing your highest performance core 's pricing.
I would not be surprised in a 2 - 3 years time the highest PC performance CPU / SoC is coming from Nvidia with ARM CPU Core rather than x86. But knowing Nvidia I know they will charge similar pricing to Intel :D
So far, Qualcomm is not paying the royalty rate hikes since they are selling ARM hardware using cores covered under the ARMv8 architectural license that they obtained before SoftBank started pushing ARM to improve profitability.
It is interesting that you should mention MediaTek. They joined the RISC-V Software Ecosystem in May 2023:
https://riseproject.dev/
It seems reasonable to think that they are considering jumping ship. If they are designing their own in-house CPU cores, it will likely be a while before we see them as part of a mediatek SoC.
In any case, people do not like added fees. They had previously tolerated ARM’s fees since they were low, but now that they are raising them, people are interested in alternatives. At least some of ARM’s partners are paying the higher for now, but it is an incentive to move to RISC-V, which is no fee for the ISA and either no fee or low fee for IP cores. For example, the hazard3 cores that the Raspberry Pi Foundation adopted in the RP2350 did not require them to pay royalty fees to anyone.
After watching the Qualcomm-ARM lawsuit in December, I have very little respect for ARM's relentless pursuit of profit for profit's sake:
https://www.tantraanalyst.com/ta/qualcomm-vs-arm-trial-day-1...
https://www.tantraanalyst.com/ta/qualcomm-vs-arm-trial-day-2...
https://www.tantraanalyst.com/ta/qualcomm-vs-arm-trial-day-3...
Doesn't ARM have a problem with RISC-V and Chinese CPUs? Long term seems they're bound to loose most of the market by simply being priced out.
Most of the high end Chinese chips are based on ARM as of now.
What about in 5 years? riscv isn't competitive at the top end but it's closing in fast.
Huawei is probably the one that possibly might move away from arm because they have their own operating system. The US could tighten their controls more and ban them from arm altogether- so far they are prohibited from arm 9.
I am not sure Huawei would go for riscv - they could easily go for their own isa or an arm fork.
Timothy Prickett Morgan is a fantastic writer and analyst. Love reading his stuff.
As an almost exclusively microcontroller user of Arm's products, a big meh from me. v8 is still slowly rolling out. M33 is making headway but I was really hoping for M55 to be the bigger driver.
That folks are still making new Cortex-A7 (2011) designs is wild. A-35 doesn't seem to be very popular or better.
Cortex-M33 (2016) derives–as you allude to–from ARMv8 (2015). But yeah it barely seems only barely popular, even now.
Having witnessed some of the 9p's & aughts computing, I never in a million years would have guessed microcontrollers & power efficient small chips would see so little change across a decade of time!!
Isn’t there some dynamic at play where STM will put one of these on a board, that board becomes a “standard” and then it’s cloned by other manufacturers, lowering cost? (legality aside)
STM32H5 in 2023 (M33): https://newsroom.st.com/media-center/press-item.html/p4519.h...
GD32F5 in 2024: https://www.gigadevice.com/about/news-and-event/news/gigadev...
STM32N6 in 2025 (M55): https://blog.st.com/stm32n6/
i.e. it takes some time for new chips to hit cost targets, and most applications don’t need the latest chips?
I don’t know a lot about the market, though, and interested to learn more
Some chips that have come out in the past 3 years with Cortex A7:
Microchip SAMA7D65 and SAMA7G54. Allwinner V853 and T113-S3.
It's not like a massive stream of A7's. But even pretty big players don't really seem to have any competitive options to try. The A-35 has some adoption. There is an A34 and A32 that I don't see much of, don't know what they'd bring above the A7. All over a decade old now and barely seen.
To be fair, just this year ARM announced Cortex-A320 which I don't know much about, but might perhaps be a viable new low power chip.
The way you put it makes it sound like STM have a serious security problem with their manufacturing.
It’s a serious problem for their lawyers, primarily: https://olimex.wordpress.com/2015/11/09/chinese-clones-attac...
cheap Chinese clones of US IP is a broader problem that affects more than just STM
You can get a lot of mileage out of a Cortex-M7. NXP has some which run up to 1 GHz - that's a ridiculous amount of power for a "microcontroller". It'd easily outperform an early-to-mid-2000s desktop PC.
There are no similarities between Cortex-M7 and Cortex-A7 from the POV of obsolescence.
Cortex-M7 belongs to the biggest-size class of ARM-based microcontrollers. There is one newer replacement for it, Cortex-M85, but for now Cortex-M7 is not completely obsolete, because it is available in various configurations from much more vendors and at lower prices than Cortex-M85.
Cortex-M7 and its successor Cortex-M85 have similar die sizes and instructions-per-clock performance with the Cortex-R8x and Cortex-A5xx cores (Cortex-M5x, Cortex-R5x and Cortex-A3x are smaller and slower cores), but while the Cortex-M8x and Cortex-R8x cores have short instruction pipelines, suitable for maximum clock frequencies around 1 GHz, the Cortex-A5xx cores have longer instruction pipelines, suitable for maximum clock frequencies around 2 GHz (allowing greater throughput, but also greater worst-case latency).
Unlike Cortex-M7, Cortex-A7 is really completely obsolete. It has been succeeded by Cortex-A53, then by Cortex-A55, then by Cortex-A510, then by Cortex-A520.
For now, Cortex-A55 is the most frequently used among this class of cores and both Cortex-A7 and Cortex-A53 are truly completely obsolete.
Even Cortex-A55 should have been obsolete by now, but the inertia in embedded computers is great, so it will remain for some time the choice for cheap embedded computers where the price of the complete computer must be well under $50 (above that price Cortex-A7x or Intel Atom cores become preferable).
In embedded old Cortex-A53 (and A72) are the most "new" cores used compared to A9 (and A15). E.g. TI AMxxxx [1] and Xilinx UltraScale+ vs Zynq.
[1] https://www.ti.com/microcontrollers-mcus-processors/arm-base...
For cores included in FPGAs, sadly there are none better than Cortex-A53 and Cortex-A72, because there have been no significant upgrades to the families of bigger FPGAs for many years. However in that case you buy the chip mainly for the FPGA and you have to be content with whatever CPU core happens to be included.
On the other hand, for the CPUs intended for cheap embedded computers there are a very large number of companies that offer products with Cortex-A55, or with Cortex-A76 or Cortex-A78, so there is no reason to accept anything older than that.
Texas Instruments is not really representative for embedded microcontrollers or computers, because everything that it offers is based on exceedingly obsolete cores.
Even if we ignore the Chinese companies, which usually have more up-to-date products, there are other companies, like Renesas and NXP, or for smaller microcontrollers Infineon and ST, all of which offer much less ancient chips than TI.
Unfortunately, the US-based companies that are active in the embedded ARM-based computer segment have clearly the most obsolete lines of products, with the exception of NVIDIA and Qualcomm, which however target only the higher end of the automotive and embedded markets, by having expensive products. If you want something made or at least designed in USA, embedded computers with Intel Atom CPUs are likely to be a better choice than something with an archaic ARM core.
For the Intel Atom cores, Gracemont has similar performance to Cortex-A78, Tremont to Cortex-A76 and Goldmont Plus to Cortex-A75; moreover, unlike the CPUs based on Cortex-A78, which are either scarce or expensive (like Qualcomm QCM6490 or NVIDIA Orin), the CPUs based on Gracemont, i.e. Amston Lake (Atom x7000 series) or Twin Lake (N?50 series), are both cheap and ubiquitous.
The latest Cortex-A7xx cores that implement the Armv9-A ISA are better than any Intel Atom core, but for now they are available only in smartphones from 2022 or more recent or in some servers, not in embedded computers (with the exception of a product with Cortex-A720 offered by some obscure Chinese company).
Does anyone but Renesas even offer a Cortex M85 based MCU? Afaik the all the other high performance ARM based microcontrollers still use a Cortex M7 except for a few M55 based chips.
In general Renesas offers more modern microcontrollers than all the other vendors of MCUs, which have decreased a lot the rate of new product launches during recent years, but unfortunately they are also among the most expensive.
I have also not seen Cortex-M85 except from Renesas. Cortex-M55 is seldom an alternative to Cortex-M7, because Cortex-M55 is smaller and slower than Cortex-M7 (but faster than the old Cortex-M4 or than the newer Cortex-M33).
Cortex-M55 implements the Helium vector instruction set, so for an application that can use Helium it may match or exceed the speed of a Cortex-M7, in which case it could replace it. Cortex-M55 may also be used to upgrade an old product with Cortex-M7, if the M7 was used only because Cortex-M4 would have been too slow, but the full speed of Cortex-M7 was not really needed.
Rumor has it that ST has one on the way (STM32V8).
At a trade show I saw a chip coming out with DDR3L. Imagine a 2025 chip with RAM from 15? years ago. They said it's all that they needed. Probably have a perpetual license or something.
> I never in a million years would have guessed microcontrollers & power efficient small chips would see so little change across a decade of time
It's because the software ecosystem around them is so incredibly lousy and painful.
Once you get something embedded to work, you never want to touch it again if you can avoid it.
I was really, really, really hoping that the RISC-V folks were going to do better. Alas, the RISC-V ecosystem seems doomed to be repeating the same levels of idiocy.
Switching microcontrollers means you have a lot of work to do to redo the HW design, re-run all of your pre-production testing, update mfg/QA with new processes and tests, possibly rewrite some of your application.. and you need to price in a new part to your BOM, figure out a secure supply for some number of years... And that just assumes you don't want to do even more work to take advantage of the new chip's capabilities by rewriting even more of your code. All while your original CPU probably still does fine because this is embedded we're talking about and your product already does what it needs to do.
The RP2040 and RP2350 are fairly big changes from the status quo, although they are not very energy efficient compared to other MCUs. Coincidentally, the RP2350 is part of the RISC-V ecosystem. It has both RISC-V and ARM cores and lets you pick which to use.
RISC-V is even worse: The Cortex-M series have standardized interrupt handling and are built so you can avoid writing any assembly for the startup code.
Meanwhile the RISC-V spec only defines very basic interrupt functionality, with most MCU vendors adding different external interrupt controllers or changing their cores to more closely follow the faster Cortex-M style, where the core itself handles stashing/unstashing registers, exit of interrupt handler on ret, vectoring for external interrupts, ... .
The low knowledge/priority of embedded of RISC-V can be seen in how long it took to specify an extension tha only includes multiplication, not division.
Especially for smaller MCUs the debug situation is unfortunate: In ARM-World you can use any CMSIS-DAP debug probe to debug different MCUs over SWD. RISC-V MCUs either have JTAG or a custom pin-reduced variant (as 4 pins for debugging is quite a lot) which is usually only supported by very few debug probes.
RISC-V just standardizes a whole lot less (and not sensibly for small embedded) than ARM.
Being customizable is one of RISC-V’s strengths. Multiplication can be easily done in software by doing bit shifts and addition in a loop. If an embedded application does not make heavy use of multiplication, you can omit multiplication from the silicon for cost savings.
That said, ARM’s SWD is certainly nice. It appears to be possible to debug the Hazard3 cores in the RP2350 in the same way as the ARM cores:
https://gigazine.net/gsc_news/en/20241004-raspberry-pi-pico-...
> If an embedded application does not make heavy use of multiplication, you can omit multiplication from the silicon for cost savings.
The problem was that the initial extension that included multiplication also included division[1]. A lot of small microcontrollers have multiplication hardware but not division hardware.
Thus it would make sense to have a multiplication-only extension.
IIRC the argument was that the CPU should just trap the division instructions and emulate them, but in the embedded world you'll want to know your performance envelopes so better to explicitly know if hardware division is available or not.
[1]: https://docs.openhwgroup.org/projects/cva6-user-manual/01_cv...
Software division is often faster than hardware division, so your performance remark seems to be a moot point:
https://libdivide.com/
I don't think that library refutes anything of what I said.
First of, that library requires you to fundamentally change the code, by moving some precomputation outside loops.
Of course I can do a similar trick to move the division outside the loop without that library using simple fixed-point math, something which is a very basic optimization technique. So any performance comparison would have to be against that, not the original code.
It is also much, much slower if your denominator changes for each invocation:
In terms of processor time, pre-computing the proper magic number and shift is on the order of one to three hardware divides, for non-powers of 2.
If you care about a fast hardware divider, then you're much more likely to have such code rather than the trivially-optimized code like the library example.
Good point. I withdraw my remark.
> It appears to be possible to debug the Hazard3 cores in the RP2350 in the same way as the ARM cores:
It is, but (as far as I understood it), they're using ARM SWD IP (which is a fine choice). But since their connection between the SWD IP and RISC-V DM is custom, you're going to need your adjust your debug probe software quite a bit more than between different Cortex MCUs.
Other vendors with similar issues (for example WCH) build something similar but incompatible, requiring their own debug probe. This is a solved problem for ARM cortex.
> It's because the software ecosystem around them is so incredibly lousy and painful.
This is reaching breaking point entirely because of how powerful modern MCUs are too. You simply cannot develop and maintain software of scale and complexity to exploit those machines using the mainstream practices of the embedded industry.
[dead]
I am surprised more uC use cases have not moved to RISC-5. What do you see keeping you on ARM for what you work on?
The RP2350 lets you choose between 2 RISC-V cores and 2 ARM cores. I believe it even supports 1 RISC-V core and 1 ARM core for those who like the idea of their microcontrollers using two different ISAs simultaneously.
Microchip Technology has a number of RISC-V options.
Depends on the task. My favorite example is a chip that has a lot more than a microcontroller onboard, but it's an old v7m. I need the rest and have to struggle with what they give. If it was RV, power PC, mips, whatever, I'd have to use it.
ESP-32C are the only mainstream ones I've encountered.
Good documentation and support.
Same. On v7 still for most things, even on newer MCUs. The v8 ones, for the use cases I've encountered, primarily add IOT features like secured Flash.
Right. And I'm not IoT so I don't care. Could be fun to play with helium, though.
ARM used to be UK owned until Conservative government lack of foresight allowed it to be sold to Softbank, leaving AIM (UK's NASDAQ, part of LSE) despite being in the national interests, and security, to keep it British. Thanks Mrs May (ex-PM) for approving that one (it was the last regulatory hurdle, that it was not in national security interests, so had to go past her).
Of course Boris Johnson (the next PM) __tried to woo ARM back to LSE__ because they realised they fucked up, and of course what huge foreign company would refloat on the LSE when you have NASDAQ, or bother floating on both?
Can you imagine if America had decided to allow Intel or Apple to be sold to a company in another country? Same sentiment.
- Yep I'm a pissed off ex-ARM shareholder forced out by the board's buyout decision and Mrs May waving it through.
They did the world a favor by indirectly helping riscv. So arguably it's a net positive move.
Before reading article: I would like to know if this architecture will help Linux close to Apple architecture efficiencies....
After reading article: I suddenly realize that CPUs will probably no longer pursue making "traditional computing" any faster/efficient. Instead, everything will be focused on AI processing. There are absolutely no market/hype forces that will prompt the investment in "traditional" computing optimization anymore.
I mean, yeah, there's probably three years of planning and execution inertia, but any push to close the gap with Apple by ARM / AMD / Intel is probably dead, and Apple will probably stop innovating the M series.
The 128- and 256-core ARM server chips (like from Ampere) are pushing server performance in interesting ways. They're economically viable now for trivially parallelizable things like web servers, but possibly game-changing if your problem can put that many general-purpose cores to work.
The thing is, there aren't that many HPC applications for that level of parallelism that aren't better served by GPUs.
It makes sense to focus. Efficiencies in CPU design are not going to see as large of an impact on user workloads as focused improvements on inference workloads. The average phone user will be happier for the longer battery life as the onslaught of ai workloads from software companies is likely not going to slow and battery life will be wrecked if nothing changes.
You think so? I posit that the deliverance of AI/ML (LLM/genAI) services and experiences are predicated upon "traditional computing" - so, there will be some level of improvement in this domain for at least quite some time longer.
apple M4 vs Intel Core Ultra 9 285K. apple m4 vs AMD Ryzen AI 9 365
apple has to do something.
im not sure intel cpus can have 196GB ram, or it is some mobile ram manufacturing limit, but i really want to have atleast 96GB in notebook, tablet.
M4 still has >2x better performance per watt than either of those chips. Of course they are pretty much ignoring desktop so they can’t really compete with AMD/Intel when power is not an issue but that’s not exactly new
M4 has ">2x better performance per watt" than either Intel or AMD only in single-threaded applications or applications with only a small number of active threads, where the advantage of M4 is that it can reach the same or a higher speed at a lower clock frequency (i.e. the Apple cores have a higher IPC).
For multithreaded applications, where all available threads are active, the advantage in performance per watt of Apple becomes much lower than "2x" and actually much lower than 1.5x, because it is determined mostly by the superior CMOS manufacturing process used by Apple and the influence of the CPU microarchitecture is small.
While the big Apple cores have a much better IPC than the competition, i.e. they do more work per clock cycle so they can use lower clock frequencies, therefore lower supply voltages, when at most a few cores are active, the performance per die area of such big cores is modest. For a complete chip, the die area is limited, so the best multithreaded performance is obtained with cores that have maximum performance per area, so that more cores can be crammed in a given die area. The cores with maximum performance per area are cores with intermediate IPC, neither too low, nor too high, like ARM Cortex-X4, Intel Skymont or AMD Zen 5 compact. The latter core from AMD has a higher IPC, which would have led to a lower performance per area, but that is compensated by its wider vector execution units. Bigger cores like ARM Cortex-X925 and Intel Lion Cove have very poor performance per area.
> where all available threads are active, the advantage in performance per watt of Apple becomes much lower
Perhaps. But that’s an edge case, very few people run their laptop at 100% for any extended period of time.
Apple is ignoring desktop?
I guess that depends on your definition of “desktop”.
What that really means (I think) is they aren’t using the power and cooling available to them in traditional desktop setups. The iMac and the Studio/Mini and yes, even the Mac Pro, are essentially just laptop designs in different cases.
Yes, they (Studio/Pro) can run an Ultra variant (vs Max being the highest on the laptop lines) but the 2x Ultra chip so far has not materialized. Rumors say Apple has tried it but rather could get efficiencies to where they needed to be or ran into other problems connecting 2 Ultras to make a ???.
The current Mac Pro would be hilarious if it wasn’t so sad, it’s just “Mac Studio with expansion slots”. One would expect/hope that the Mac Pro would take advantage of the space in some way (other than just expansion slots, which most people have no use for aside from GPUs which the os can’t/won’t leverage IIRC).
I meant that Apple doesn’t really design or make desktop CPUs. Only oversized laptop ones (even if they are very good).
Their most performant chip, the M3 Ultra is only a bit faster than the 14900k which you can gee for $400 these days.
I think this largely misses the point. Power users, so most of the users on HN, are a niche market. Most people don't need a hundred gigs of RAM, they need their laptop to run Powerpoint and a browser smoothly and for the battery to last a long time. No other manufacturer is anywhere close to Apple in that segment as far as I'm concerned.
> but i really want to have atleast 96GB in notebook, tablet.
in notebooks it's been possible for years. a friend of mine had 128gb (4x32gb ddr4) in his laptop about 4-6 years ago already. it was a dell precision workstation (2100 euros for the laptop alone, core i9 cpu, nothing fancy).
Nowadays you can get 64gb individual ddr5 laptop ram sticks. as long as you can find a laptop with two ram sockets you can easily get 128b memory on laptops.
regarding tablets... it's unlikely to be seen (<edit> in the near future</edit>). tablet OEMs tip their hats to the general consumer markets, where <=16gb ram is more than enough (and 96gb memory would cost more than the rest of the hardware for no real user/market/sales advantage)
Some Intel chips have a max of 192GiB. Others 4TiB. It depends on the chip, but there are definitely machines running terabytes of memory.