Intel's plan to license x86 cores for chips with Arm, RISC-V and more inside

Intel has shed some light on how it hopes customers will create single-package processors in which x86, Arm and RISC-V cores work together. For that, Intel will license its most important asset, the x86 architecture, to those who want to make custom silicon. Depending on the application, customers will be able to mix up x86, …

[Anonymous Coward]

Intel

You remind me of a person who is about to drown. their arms flailing in a desperate effort to remain afloat.

Apple has dumped you. They will not be the only ones to do so. Other companies are just as capable as Apple at creating their own chips.

Your future efforts? Too late. Your own complacency will be your final downfall.

[John Brown (no body)]

Re: Intel

Not so sure about that. I don't see MS moving away from x86 anytime soon. Although I'm struggling to see where Intel are going with this. I don't really see any mass market for CPUs with multiple incompatible architecture cores. Are there people out there already combining ARM, Risc-V and x86 boxes that could benefit from having it all in one box/package? I see people using GPUs for data processing offloaded from the CPU, but not on Intel silicon. It's usually Nvidea GFX cards. It'll be interesting to see if anyone runs with this, why they choose the route and what benefits, if any, it brings.

[Michael Habel]

Re: Intel

Microsoft are just another dinosaur, about the only thing they are still capable of is to make S0NY look like a bunc of jerks. Which to be homnest, when you have the same cat running the place that killed Sega some twenty-odd years ago. doesnt seem to hard to do. While MicroSofts off somewhere being Chad, buing up Bio/Zeni & Acti/Bliz Cerny gets it into his head to buy... BUT NOT OWN Bungie --ouch

Well I'm not a fan of the Failbox, and S0NYs Managment is set to make the PSV as succsessful as the PS Vita, at this point. At this point my trust on / in S0NY is that they will continue to fisk this up. Why they don't mortgage the Farm to aquire, and in turn make all future Take2 / Rockstar Games PS Exclusive, is frankly lost on me. Perhaps S0NY are holding out some hope that Brandons DoJ will block this deal? I'd gladly place my money on exactly that NOT happining.

That said the Windows OS has been sliding down ever increasing rungs of hell, with its embedded spyware, and Candy Cursh (Which it now owns... so YEAH!), have been such worthwhile upgrades since XP/Vista/7.

No when the OD/EMs finally get 'round to bending the arm of the Redmond Giant, they will probably cave in and go Arm, as well. In fact, I was asking myself if this was a way to re-woo Apple to reimplement the x86 instruction set on the Mx Chip. Thinking about it here. This seems more like its going the other way, and how x86 finally dies, once the legacy software dies off.

[Down not across]

Re: Intel

Well I'm not a fan of the Failbox, and S0NYs Managment is set to make the PSV as succsessful as the PS Vita, at this point. At this point my trust on / in S0NY is that they will continue to fisk this up. Why they don't mortgage the Farm to aquire, and in turn make all future Take2 / Rockstar Games PS Exclusive, is frankly lost on me. Perhaps S0NY are holding out some hope that Brandons DoJ will block this deal? I'd gladly place my money on exactly that NOT happining.

Nor am I. However Sony pushed me to it with the PS3 debacle back in the day. Constant delays, and then in the end sold a crippled console in Europe. Not long after they then started stripping features off too (other OS). The new Series X is actually a decent console.

Exclusives are so last decade. If Sony did that they'd be even more dead to me than they already are. There is no game (as much as I enjoy GTA) that would make me go spend 500 quid on a console just for that game (or possibly couple more depending on the studio in question).

it's not just the console debacle either, they have have had other bugles and the few times I've tried their hardware it has been substandard quality and had to be sent back.

Quite sad really. You used to have to pay premium for Sony (Trinitron etc) but it was worth it. They've lost their way (like to be fair pretty much everyone these days).

[John Robson]

Re: Intel

"You remind me of a person who is about to drown. their arms flailing in a desperate effort to remain afloat."

Whilst that is the popular presentation for TV the reality is usually very different - people don't have the capacity to wave their arms in the air when they are really drowning.

https://rnli.org/magazine/magazine-featured-list/2017/march/how-to-recognise-drowning-its-not-like-the-movies

There is a panic phase before they start drowning when they will, but that doesn't last long (since raising your arms tends to push your head underwater).

[Greg D]

Re: Intel

I doubt very much that Intel are going anywhere. Nice try though.

The x86 doomsayers have been on about the death of this instruction set for years, but it wont happen. Not as long as there are demands for the workloads - and there will ALWAYS be demands for the workloads while classical computing is still king - RISC is a nice alternative for lower power consumption, but thats about it. This move only shows the strength x86 still has, and its still dominant in the home computing and cloud computing spaces. How will Intel possibly cope with all that demand for their licensed IP?

I dont think they will go anywhere bud.

[Michael Habel]

Re: Intel

Ahh yes but, Apple had yet to show the world+dog the way to how ARM based Macs were smoking those with intel 5/7/9 cores. the cats been loosed. Arm Chips are likely from a cost point cheaper to buy in bulk than x86.

Thats why I susspect that this move is less Apple, and more of a slow goodbye to x86, As the Arm-ie parts of the Chip will soon be takeing over, while the x86 Core, keeps getting derecated down the leagacy shoot.

I guess time will tell if there will be a 13 Series line. I susspect this new "Chip" would debut around such a time. Probably in some ultrabook somewhere. Undoubtedly one thats 96% Mac-compat.*

*Sans the T1 Security Chip.

[jetjet]

No worries, chinese will give them a hand.

[mark l 2]

So it looks like Intel sees its future in being more like TSMC in producing chips for others rather than just relying on using its fabs for its own kit?

This makes sense since building fabs is expensive and if the PC and server market could take a down turn and the demand for X86 drops off.

[bigtreeman]

open sauce

All the proprietary processor makers are shitting themselves.

Just as Linux moved the goalposts so will risc-v.

It won't happen overnight, but it will happen.

Intel is making each way bets, plowing money into risc-v and offering x86.

They have a fucking lot to loose.

Taking a strong direction and going with the new tech would be much wiser.

[Anonymous Coward]

Gravy, gravy, graviton

Cloud hyperscalers are causing server workloads to run on Arm - cheaper.

Apple made desktop workloads run on custom Arm.

Mobile on Arm.

x86 no longer has a monopoly on anything.

What will Intel do?

Arm is yo future x86 his past!

[Anonymous Coward]

Job?

Having a quick browse here : https://en.wikipedia.org/wiki/List_of_Intel_manufacturing_sites Shows the majority of chip fabs are in the US. If this is successful, it will have a huge impact on 'merica jobs...

[Michael Habel]

Sounds like a last chance ditch attempt at trying to keep x86 relevent for a little while longer to me. I say its time to man up and just yank that band-aid off. The sooner x-86 get relegated to history the better....

But, the real question, is who are they trying to woo with this?, and why should Apple take them up on it?

[Anonymous Coward]

While I'm not a fan of x86, the world is currently benefiting from the Intel-AMD duopoly where the common ISA allows data centres to shop around.

Although RISC architectures are inherently more elegant, it seems like everything is limited by throughput of the memory hierarchy so from a software perspective it doesn't matter hugely whether the code is running on x86, ARM, or RISC-V. Although the RISCs are better for low-powered applications, and really the only game in town for embedded.

That might explain the interest from Intel on non-x86 architectures - the same principle that makes the Alder Lake "efficiency" cores a good idea can be applied to the even more efficient RISC architectures.

[3arn0wl]

With open source chip designs beginning to appear, it looks like an astute move to pivot to fabbing - something Intel are already set up to do.

Technology has been powered by processors based on different ISAs for a while, so there's nothing particularly new there. I guess: rightly or wrongly Intel thinks that there are things that one ISA can do that another can't... or at least, not so well. Otherwise they would have plumped for one of them... Or is this the gracious admission that X86 has been bested?

Will we now begin to see powerful RISC-V processors with an application compatibility layer for the very survival of Intel's wealthy clients?

[naive]

Combining ARM and x86 cores perhaps has a great potential

A mobile phone with ARM and x86 cores would allow native Android, Chrome OS and Windows to run.

Maybe we finally get in an age where we can simply insert the mobile phone into a standardized cradle, press an icon, boot windows and continue to work with a keyboard, mouse and screen combined with the standard features of Android.

The fact that x86 architecture will be a bit power hungry due to its complex instruction set is not really an issue since running a desktop OS would allow the device to charge in the cradle.

This would free us from clutter, a plethora of devices and having to choose between things like google drive or one drive.

It is interesting to watch which products will result from this decision of Intel.

[iron]

Re: Combining ARM and x86 cores perhaps has a great potential

Why? Laptops are already crippled in performance why would you want to downgrade further? Toys are fine for web browsing but if there's a job to be done give me a desktop with lots of cores and RAM.

[Claverhouse]

Re: Combining ARM and x86 cores perhaps has a great potential

So very true.

[heyrick]

Re: Combining ARM and x86 cores perhaps has a great potential

"A mobile phone with ARM and x86 cores would allow native Android, Chrome OS and Windows to run.

Brilliant! Now let's talk battery life......

"since running a desktop OS would allow the device to charge in the cradle"

Aha, so you can't really use x86 on the go because it'll shit on the battery, not to mention likely being painful to use desktop Windows on a 5 inch screen. So, uh, what's the point exactly?

Sounds like a solution in desperate search of a problem.

[Michael Habel]

Re: Combining ARM and x86 cores perhaps has a great potential

I might have missed the boat but, hasn't someone _coof_ Samsung _coof_ already done this with DeX?

*Granted it probably wasn't Redmonds finest that was running behind DeX though.

[Michael Habel]

Re: Combining ARM and x86 cores perhaps has a great potential

Ahh if only it were that simple... When the slightst workload (Browsing, YouTube, or <Insert Game>), sets the phondelslab off like a torch. Let alone trying to run Windows, as an app. So perhaps Samsung could hire Linus (Tech Tips), to work out a watercooling soulution to that AIO phablet.

[adam 40]

Wot no Atom?

Whatever happened to the Atom CPU? Not mentioned at all...

[hammarbtyp]

It feels a bit like strapping an electric motor to a steam engine, just to keep steam power relevant

What apples M1 has shown is that ARM processors outperform Intel's brightest and best when power consideration is important. And that is 90% of applications today. Laptops are more convenient and can do 95% of what a desktop can do, and then we consider tablets and phones, Intel best in this area has always come up short.

Apart from some niche applications, the x86 architecture is just showing its age. The only thing really ARM from full adoption is that windows still have not managed a working solution, but it cannot be far away

The x86 architecture has had a good run, but we may well be seeing its end run, which is why Intel is flailing around trying to show it is still relevant

[Brewster's Angle Grinder]

People keep using the M1 as evidence ARM is better than x86. But you can't buy a licence for an M1 chip from ARM; and you won't get all that performance goodness if you buy a stock Cortex. That's before we get to all the reasons the M1 might not run your workload as fast.

[hammarbtyp]

Firstly its not about pure execution speed, its about a trade off between power/thermal efficiency and execution speed. In that respect the M1 beats Intel's best hands down.

And yes, it is Apple proprietary technology, however what it show is what is possible and you can be sure that other companies are following on behind.

[Rob Davis]

Exciting. They're having to evolve their business model just like Microsoft.

Microsoft have changed tack in recent years: Visual Studio Code, cross platform. WSL2 (which from experience is very good). Integration with Android. Open sourcing some other software too. Enhanced developer relations. I think they're saying: "we still have Windows, and Office but we do lots of other things too that can run on other platforms."

Now, Intel are doing something like this: "Yes we have x86 but we recognise other platforms and will work with them." Even, integrate them together. Provide solutions - answers to questions, rather than just answers without a question. The hybrid concept of x86 and ARM on one package has always interested me, the ability to run x86 Windows and APPs alongside Android apps natively, in tablets or even some phones. Good to see RISC-V on the table, they're not missing a trick there.

It's what a business should do, move with the times. Enlightened Pragmatism.

No affiliation from me.

[luis river]

That Intel initiative, too late.

The X86 Kingdom its in trouble, menace from other´s CPU actors, today it is one real fact, perhaps future to help X86 Intel ISA to come from somewhere that will be total unexpected.

[martinusher]

Been there, done that

Back in the early days of x86 Intel was required to have a second source by its major customer so they licensed x86 to AMD. AMD was a smaller player but they actually made better parts than Intel's -- the parts had more predictable timings and ran a lot cooler. The result was that when Intel brought out the 386, the first 32 bit processor, in 1985 they froze AMD out. AMD still had a license to make x86 architecture so they started working on their own 386 design, causing Intel to set up IP roadblocks, claiming IP infringement (in the floating point unit, if I recall correctly) and so on.

AMD persisted and despite immense obstacles brought out their own advanced x86 parts. They really got back on the map when they brought out the first 64 bit processor of this series and they've continued to this day to to be a viable competitor.

Given this track record a company would be unwise to get involved with Intel like AMD did unless they had some really smart lawyers. However, since the x86 architecture is really not needed any more -- you support it because its popular, not because its particularly good -- then it might be worth not bothering. Worst case you can emulate the instructions.

[that one in the corner]

Re: Been there, done that

> Worst case you can emulate the instructions

Which is precisely what the Intel parts have done for a long time now anyway, so guess what a lot of their IP describes; looks like you'd get involved with them after all.

[linpengcheng]

Intel plans to fulfill my prediction 8 months ago

2021-06-14, I published ["Prediction: Intel will use 'RISC-V plus x86 compatibility layer' or 'RISC-V plus x86 heterogeneous computing architecture' to develop a new generation of `warehouse/workshop model` CPU"](https://github.com/linpengcheng/PurefunctionPipelineDataflow/blob/master/doc/Intel_RISC_V.md).

I think Intel should use my solution, Because if Intel wants to implement this solution it has to use my "Warehouse/Shop Model", Because Intel's 10nm process is not as precise as the Apple M1's 5nm process, So it can't integrate as much memory as the Apple M1, ie: can't use the standard warehouse/shop model, can only use variant "dispatch center (virtual global unified warehouse, integration layer or platform, the parent company of the enterprise group)" model, and improve the shortcomings of not integrating large memory by increasing the cache.

Reference:

- 2022-02-14, Agam Shah, [Intel's plan to license x86 cores for chips with Arm, RISC-V and more inside](https://www.theregister.com/2022/02/14/intel_x86_licensing/)

- 2021-06-14, Lin Pengcheng, [Prediction: Intel will use "RISC-V plus x86 compatibility layer" or "RISC-V plus x86 heterogeneous computing architecture" to develop a new generation of "warehouse/workshop model" CPU](https://github.com/linpengcheng/PurefunctionPipelineDataflow/blob/master/doc/Intel_RISC_V.md)

[martinusher]

Re: Intel plans to fulfill my prediction 8 months ago

I think you'll find that Intel's processors are just that, a RISC with microcode that emulates the x86 architecture. As for "Which RISC?" Intel already had a RISC product, the i960, which was actually a pretty good processor. Intel dumped it in favour of an ARM variant licensed from DEC; Intel had some great ideas and initiatives during the 80s and 90s but I suspect that because the corporate focus was on the huge income stream from the PC platform they were just not interested in anything else, no matter how good it was.

(I experienced something similar in the 90s. I was happily using AMC29000 series RISCs when the company abruptly cut the product. The justification for this from the AMD rep was that "the market for x86 was so huge that even with their comparatively small share and a complete monopoly in RISCs there's no economic justification for continuing with that product".)(I suspect the 29K lives on in one form or another in their graphics processors.)

As for RISC in general, its logical. In the 1970s I was working at ICL in England in a group designing a new processor. Their instruction set was huge, complex and capable but it had already been noted that the vast bulk of instructions were simple loads, stores and inter-register operations. This was pre-RISC, processors in general weren't complicated enough to have that distinction (although they'd been microcoded since at least the 1960s) but the idea that fast execution of regular instructions runs rings around complex, capable, instructions was well understood even then -- and the idea was even carried into a product by one ICL group with their 'micos engine'.

[Richard 12]

Surely this is for embedded?

A lot of embedded applications need a "powerful" application processor plus one or more microcontrollers running real-time peripherals - motor control and the like.

I've designed a few of these using separate chips and a serial interconnect.

There are now several manufacturers making chips that place an ARM Cortex-A class with a Cortex-M or RISC-V into the same package with some amount of shared SRAM, which makes these much simpler and cheaper - smaller PCB, reduced BoM etc.

This looks like Intel wanting to get in on that market. We'd certainly have considered it for those products, if the price was reasonable.