Apple's Internal Hardware Team Is Working On Modems That Will Likely Replace Intel (arstechnica.com) 81
An anonymous reader quotes a report from Ars Technica: Apple will design its own modems in-house, according to sources that spoke with Reuters. In doing so, the company may hope to leave behind Intel modems in its mobile devices, which Apple has used since a recent falling out with Qualcomm. According to the sources, the team working on modem design now reports to Johny Srouji, Apple's senior vice president of hardware technologies. Srouji joined Apple back in 2004 and led development of Apple's first in-house system-on-a-chip, the A4. He has overseen Apple silicon ever since, including the recent A12 and A12X in the new iPhone and iPad Pro models.
Before this move, Apple's modem work ultimately fell under Dan Riccio, who ran engineering for iPhones, iPads, and Macs. As Reuters noted, that division was heavily focused on managing the supply chain and working with externally made components. The fact that the team is moving into the group focused on developing in-house components is a strong signal that Apple will not be looking outside its own walls for modems in the future. In recent years, Apple has been locked in a costly and complex series of legal battles with Qualcomm, the industry's foremost maker of mobile wireless chips. While Apple previously used Qualcomm's chips in its phones, the legal struggles led the tech giant to turn instead to Intel in recent iPhones.
Before this move, Apple's modem work ultimately fell under Dan Riccio, who ran engineering for iPhones, iPads, and Macs. As Reuters noted, that division was heavily focused on managing the supply chain and working with externally made components. The fact that the team is moving into the group focused on developing in-house components is a strong signal that Apple will not be looking outside its own walls for modems in the future. In recent years, Apple has been locked in a costly and complex series of legal battles with Qualcomm, the industry's foremost maker of mobile wireless chips. While Apple previously used Qualcomm's chips in its phones, the legal struggles led the tech giant to turn instead to Intel in recent iPhones.
Best of luck (Score:5, Insightful)
If nothing else, I suspect that there are some Qualcomm on Intel employees who work on these designs that are about to receive some job offers with very attractive salaries.
Re: (Score:3)
Motorola doesn't make anything.
Fixed that for you. Motorola of old is now just a name slapped on a Chinese box.
Re: (Score:1)
They still own IP which was the original point.
Re: (Score:2)
Most of it already cross-licensed to Apple, I think. :-D
Re: (Score:2)
Re: (Score:2, Insightful)
I think the even bigger question is: can they do it while avoiding the patent minefield that Qualcomm and, presumably, Intel (and the other big players, like Samsung) have laid surrounding cellular modems.
RAND (Score:1)
The whole point of the RAND patent pool around standards for things like wireless modems, is exactly so there will not be a minefield - just a tollbooth of relatively known quantity.
Re: (Score:2, Insightful)
Wireless communication is a pretty tech and research intensive field. Then again, so is microprocessor design.
There is immense demand for the technology and Qualcomm won't be on the top of the pile for long with the way they keep behaving.
Intel made a good stab at it, but Intel has severe structural problems as a company that dooms their side ventures to failure.
Apple has been good at developing first party parts for their products and if anyone has a chance they will- They'll also have the good sense to ab
Re:Best of luck (Score:4, Informative)
I suspect a large part of this success is not due to technical superiority, but getting priority at newer (lower power) fabs due to Apple's large volume of orders. Their Ax processors were the first to use the smaller lithography available from fabs, which meant a corresponding reduction in power consumption for the same level of performance (or alternatively, better performance at the same power consumption). Back when they still used Samsung as a fab, even Samsung Semiconductor prioritized Apple's order ahead of Samsung Mobile's own Exynos SoCs.
If you remember the whole mess over the Nvidia Maxwell GPUs (the 8xx and 9xx series) not performing as expected, it's for the same reason. Nvidia assumed they'd be able to use TSMC's new 16nm and Samsung's new 14nm processes to manufacture Maxwell, and designed Maxwell assuming the thermal limits of those lithographies. But Apple's order for Ax processors bumped them down in the queue. That forced Nvidia to manufacture Maxwell on 28nm, leading it to overheat until they redesigned it with fewer cores, meaning it under-performed. It was so bad they actually re-used the Kepler architecture (700x series) for their higher-end mobile 8xx GPUs, since it was already optimized for the thermals of 28nm. Pascal (10xx series) was manufactured on 14nm and 16nm as expected, and was a success.
Re: (Score:2)
NVIDIA never manufactured anything at a Samsung fab and the processes are not even compatible.
You would have to redesign the chip to manufacture it there.
Re: (Score:3)
Re: (Score:2)
"Intel hasn't had a lot of success outside of x86"
Intel's greatest success outside of x86 has been... their cellular modems. Until the XMM 7560 their modems used ARM cores for the communications processor and were manufactured on TSMC processes.
Re: (Score:3)
Apple supposedly opened an office in San Diego right next to where Qualcomm designs its baseband processors.
You can guess why.
Re: (Score:3)
Only now? (Score:2)
I'm surprised they taken until now to begin this. Having your own silicon design team, who have already produced the W2 chip for the ear buds (yeah I know different ball park but still a toe dipped in wireless), you'd think you'd have started this back when they picked a fight with Qualcomm.
Also with the perpetual drive to put everything on a single chip and use less space this seems like an obvious move.
I'm guessing it's their habit of focusing on just a few things (CPU, Graphics & AI silicon) that has
Inevitable, yet interesting (Score:5, Interesting)
After they moved chip production in-house with their A-series, which has routinely been benchmarking far ahead of contemporary chips from Qualcomm, people started wondering when they'd do the same with cellular modems. Their recent spat with Qualcomm and their reliance on a lesser product from Intel may have hastened their decision to take it on themselves.
It'll be interesting to see if, when, and how it pans out. With their processors and GPUs, they were able to build on top of ARM and Imagination tech, respectively (though they're now doing their own thing with the latter), but what do they build on with modems? And how will they avoid Qualcomm's prodigious patent portfolio covering the technology?
Re:Inevitable, yet interesting (Score:5, Insightful)
they're trying to compete with hardware companies on hardware chips. This will not end well.... for Apple.
Given that it’s already going well for Apple with their A-series chips, I’m inclined to think you’re a bit out of touch with reality. Please re-sync with reality at your earliest convenience.
Re: (Score:3)
What they are doing is licensing a 3rd party technology from someone else and bolting it together like Lego while at the same time using their market power and huge orders to be the first in line for the latest technologies fabs can provide which gives them a nice edge.
This is very different from designing a modem, a task which far more experienced companies than Apple are doing poorly at.
Re:Inevitable, yet interesting (Score:4, Informative)
Sorry, no. Their cores are fully custom and designed from scratch. They haven't licensed other people's technology and bolted them together since the A5 over seven years ago.
The idea that first access to fabs is the only reason their ARM cores beat everyone else's is laughable, considering that Huawei's core was out on the same process as the A12 within a month or two but only half as fast. In fact, Apple's two year old A10 is pretty comparable to the Huawei's latest, and Qualcomm's upcoming Snapdragon 855 (which also uses the same process as the A12)
Re: (Score:3)
Sorry, no. Their cores are fully custom and designed from scratch.
There's an entire universe of difference between "fully custom and designed from scratch" and licensing the entire architecture from ARM.
Re:Inevitable, yet interesting (Score:4, Informative)
There's an entire universe of difference between "fully custom and designed from scratch" and licensing the entire architecture from ARM.
Apple needs a license to build anything using the ARM aarch64 instruction set. The implementation is pure Apple.
Re: (Score:2, Interesting)
All they are licensing is the iSA. That's a definition of what the opcodes are and what the opcode is supposed to do, it is just text. Apple's engineers laid out the entire CPU, deciding where every transistor and wire goes, they didn't get any of that from ARM. Neither did Samsung for their latest Exynos, it also had fully custom cores. Qualcomm, on the other hand, licenses an ARM core like the A76 and makes a few modifications, and Huawei just licenses the straight core and doesn't make any changes of the
Re: (Score:2)
What they are doing is licensing a 3rd party technology from someone else and bolting it together like Lego
If it were that easy, wouldn't everyone else be doing it too?
Re: (Score:2)
Everyone IS doing it. SoCs are dime a dozen from basically every manufacturer remotely related to technology. The difference is in their market power which provides them an edge when it comes to getting it fabbed.
Re: (Score:2)
What they are doing is licensing a 3rd party technology from someone else and bolting it together like Lego while at the same time using their market power and huge orders to be the first in line for the latest technologies fabs can provide which gives them a nice edge.
What you just said doesn’t have the ring of truth to it, nor does it hold up to scrutiny. For instance, you do realize that they use the same fabs as everyone else, right, namely Samsung and TSMC? And that until this last year or so, everyone was pretty much held up at 10nm, so there was no major advantage to be had? Moreover, if what you were saying was true, we’d expect the Snapdragon line to be even better, given that it’s licensing the same tech, built on the same fabs, but should be b
Re: (Score:2)
For instance, you do realize that they use the same fabs as everyone else, right, namely Samsung and TSMC?
Indeed. Did you read to the end of my sentence. Just because I walk up to TSMC doesn't mean I'll get access to what Apple has.
And that until this last year or so, everyone was pretty much held up at 10nm, so there was no major advantage to be had?
Well given the rounding error differences between top of the line ARM SoCs on the market in terms of performance I fully agree with you.
Moreover, if what you were saying was true, we’d expect the Snapdragon line to be even better, given that it’s licensing the same tech, built on the same fabs, but should be benefitting from larger market power and orders than what even Apple can bring to bear ... snip ... Likewise, if that’s all it took, we’d expect Samsung by itself to have done it already and achieved similar results, but they haven’t.
Except no. How many units you crank out and what you pay at the fab are two different hings. Qualcomm live on tight margins, Apple is the exact opposite and past stories have shown they pay such a nice premium for access to fabs that even Samsung pri
Re: (Score:2)
What you seem to be unaware of is the monumental difference between modifying an existing licensed platform, and going out on your own and building something new. Who will Apple pay for the base design? Qualcomm? Intel? Huawei?
Actually, I’m not. You apparently forgot how I started this thread:
It'll be interesting to see if, when, and how it pans out. With their processors and GPUs, they were able to build on top of ARM and Imagination tech, respectively (though they're now doing their own thing with the latter), but what do they build on with modems?
And at this point, I’m so confused by what you’re trying to argue. You acknowledge the “success of their SoCs” and claim it’s based on access to better fabrication techniques, yet you’re saying there’s a “rounding error difference” in terms of top of the line performance, which seems to contradict what you just said of their success and what the benchmarks have been saying for years
Re: (Score:2)
Well one thing they might want to do is make cellular connectivity common, even is it's not the latest and greatest standard. Right now you get WiFi+BT on almost everything but cellular is always a big extra. Yes, you can use the phone as a hotspot but a laptop has a bigger battery, better space for an antenna and you're more likely to have that plugged in. If it's an in house design they can throw in there for the cost of die space it could be a mass market feature instead of a road warrior niche.
Re: (Score:2)
cellular connectivity is only good when it's not locked in any way and has a dual sim as well.
But DO NOT REPLACE e-net or wifi with cell only. We do not need to pay $10-40/mo outlet fee per device or be stuck with 15-20G caps.
Re: (Score:2)
After they moved chip production in-house with their A-series
Production, really ? I think that they only design them.
Re: (Score:2)
Yes, it’s just design, not fabrication. I misspoke. Thanks for the correction.
Repair? (Score:2)
If a surface mount chip goes on your phone, you toss the whole board. Reworking an SMC part would be cost-comparable to buying a new phone.
Re: (Score:2)
How will a modem replace a chip company? (Score:1, Offtopic)
Wish authors would learn to write headlines...
I hope Apple fixes bufferbloat in LTE & 5G (Score:2)
Re: (Score:2)
What makes you think the end user device is in any way in control of this?
Re: (Score:2)
Re: (Score:3)
In wifi, now, at least, uplinks are easily controlled at the user device (phone), which is what I was mostly measuring.
And how does that fix the monumental amount of equipment that results in buffer bloat between your phone and the target you're accessing?
The user is not in control in this in the slightest. All they can do is not make it worse than the many hops already in the system. Apple won't fix your buffer bloat. They can't. All you can ask them to do is not make it worse.
Re: (Score:2)
Re: (Score:2)
No it doesn't, it occurs on any bottlenecked link including the backhaul. It doesn't need to be the slowest link in the chain.
let me guess 5Gi - a new standard! (Score:2)
This is not the same as Apple and its A-series SoC (Score:5, Interesting)
This is not the same as Apple and its A-series SoCs. Apple purchased licenses from ARM Holdings to produce their own ARM-based CPUs. There is no one single company Apple can go to to license the technology to produce its own baseband chipsets. Additionally, Apple is in a very public battle with Qualcomm about the very patents and licenses that underpin 3G/4G/5G baseband technology.
The patent licensing aside, one of the other Qualcomm lawsuits involves the violation of NDAs and Apple violating Qualcomm's baseband technology trade secrets. In order to integrate Qualcomm's chipsets into the iPhone, Apple entered into NDAs with Qualcomm for detailed technical information. Qualcomm alleges that Apple shared these secrets with Intel after Apple dumped Qualcomm chipsets. Even if Qualcomm cannot prove that Apple did this, it is going to be impossible for Apple to prove that they somehow did not use this same information to produce their own baseband chipsets. I believe this is a much bigger issue for Apple than the patent licensing issue. Undoubtedly there will be direct non-compete clauses in these NDAs.
Short of actually purchasing Qualcomm, or some other baseband chipset manufacturer, it will be impossible for Apple to show that they have come up with their own cleanroom implementation of baseband chipsets that is unencumbered by some kind of patent or licensing issue or NDA contractual issue.
Evolution (Score:1)
A simple cost saving equation, keep the simple stuff in house.
Re: (Score:2)
Undoubtedly, Apple would have their own chips already running in a lab somewhere. They can make all the chips they want in-house, as long as they are for internal use only in their labs, but the moment they incorporate those chips into a commercial device then they immediately become subject to patents and licensing issues. Even if Apple could somehow claim they have come up with a completely clean room implementation (which is rather impossible given their past technical collaboration with Qualcomm, and no
Apple is not going to make money on modems (Score:2)
It will come at huge cost, increasing price of phones (gotta absorb R&D and production cost). Only companies specializing in modems make money from their production.
And good luck with entering that market and not infringing on some patents.
They'd be better off with making a joint venture with Intel.
You have successful platform (iOS), that you make obscene money off. So gotta let your suppliers profit too. I get that Broadcom wanted big piece of pie, but thet's where competition come into picture.
It is g
Apple copying other's innovation again? (Score:2)