Wednesday, June 24th 2020
Intel Gives its First Comments on Apple's Departure from x86
Apple on Monday formalized the beginning of its departure from Intel x86 machine architecture for its Mac computers. Apple makes up to 4 percent of Intel's annual CPU sales, according to a MarketWatch report. Apple is now scaling up its own A-series SoCs that use Arm CPU cores, up to performance levels relevant to Macs, and has implemented support for not just new and upcoming software ported to the new Arm machine architecture, but also software over form the iOS and iPadOS ecosystems on Mac, starting with its MacOS "Big Sur" operating system. We reached out to Intel for some of its first comments on the development.
In a comment to TechPowerUp, an Intel spokesperson said "Apple is a customer across several areas of our business, and we will continue to support them. Intel remains focused on delivering the most advanced PC experiences and a wide range of technology choices that redefine computing. We believe Intel-powered PCs—like those based on our forthcoming Tiger Lake mobile platform—provide global customers the best experience in the areas they value most, as well as the most open platform for developers, both today and into the future."As we mentioned earlier, Apple only makes up a single-digit percentage of Intel's annual sales, and the loss of sales to Apple could mean more silicon that, in our opinion, Intel could divert to the DIY channel. Over the past two years, Intel was embattled with shortages in the DIY retail channel as the OEM channel soaked up the bulk of its silicon allocation.
Could Intel have done more to retain Apple? To begin answering this question, one must dig into possible reasons behind Apple's departure from Intel. Apple's industrial design, since the revolutionary MacBook (2016), has been toward thinner devices with more battery life, lesser compute on the client-end, and more on the cloud. The company sees a future in devices with iPad-like always-on availability, and battery life running into dozens of hours. Apple wants greater control over what its suppliers provide, to attain these goals.
While Intel has managed to bring SoC TDP down to 7 W thru 15 W with its 10 nm U-segment processors, these processors appear to be falling short of Apple's performance/Watt requirements. An example of chip design control Apple expects from its hardware suppliers can be found with AMD. The Sunnyvale-based firm supplies Apple with its most efficient bins of Radeon GPUs, and in cases such as the "Navi 12," Pro 5500, and Pro W5700X, even reserves graphics SKUs exclusive to Apple (not sold to its AIB channel).
In our opinion, Intel could have done more to retain Apple. The engineering department certainly rose to the occasion, developing "Lakefield." While performance numbers are still under the wraps, "Lakefield" is the kind of chip one would expect in an Apple portable - an extremely power-efficient client-segment processor capable of sub 1-Watt idle, high burst performance, and great customization flexibility thanks to its Foveros Packaging that lets system designers pick and choose the I/O components they want specific to their designs. On the other hand, the business-end of Intel may have fumbled with Apple. If designing chips that match Apple's requirements didn't work, Intel could have used the nuclear option - of pitching the x86 machine architecture itself.
Since its inception, Intel has licensed the x86 machine architecture to over a dozen companies. There are currently only two active licensees - AMD and VIA. The rest either withered away, or consolidated (VIA consolidated CenTaur and CYRIX). There have been no new licensees in at least the past 15 years (not if you don't count the sub-licensing of Zhaoxin by VIA or to THATIC by AMD). Apple could have been the first new x86 licensee in a generation, and with a little assistance, could have developed its own x86 SoCs. Apple pays for Unix in the era of Linux, and it would have surely indulged a well-crafted license deal with Intel.
Apple's departure from x86, despite amounting to a paltry percentage of Intel's sales, could reshape the client-computing segment. In our opinion, the success of the Arm machine architecture on Macs presents a greater threat to Intel than even x86 licensee AMD, as it could trigger other semiconductor firms with deep pockets and Arm licenses, such as NVIDIA and Samsung, to develop "high performance" Arm SoCs of their own, for thin-and-light notebooks. The "Wintel" era is long gone, and Microsoft is only too happy to indulge and grow its Windows-on-Arm ecosystem.
The future of Intel's client-segment silicon looks increasingly similar to that of Arm. Highly modular IP blocks, including from third-parties, integrate on innovative new packaging formats, such as Foveros, with an overwhelming focus on performance/Watt, thanks to hybrid cores, idle power-draw, battery life, and performance in bursts. A lot is riding on the success of the tiny "Lakefield."
In a comment to TechPowerUp, an Intel spokesperson said "Apple is a customer across several areas of our business, and we will continue to support them. Intel remains focused on delivering the most advanced PC experiences and a wide range of technology choices that redefine computing. We believe Intel-powered PCs—like those based on our forthcoming Tiger Lake mobile platform—provide global customers the best experience in the areas they value most, as well as the most open platform for developers, both today and into the future."As we mentioned earlier, Apple only makes up a single-digit percentage of Intel's annual sales, and the loss of sales to Apple could mean more silicon that, in our opinion, Intel could divert to the DIY channel. Over the past two years, Intel was embattled with shortages in the DIY retail channel as the OEM channel soaked up the bulk of its silicon allocation.
Could Intel have done more to retain Apple? To begin answering this question, one must dig into possible reasons behind Apple's departure from Intel. Apple's industrial design, since the revolutionary MacBook (2016), has been toward thinner devices with more battery life, lesser compute on the client-end, and more on the cloud. The company sees a future in devices with iPad-like always-on availability, and battery life running into dozens of hours. Apple wants greater control over what its suppliers provide, to attain these goals.
While Intel has managed to bring SoC TDP down to 7 W thru 15 W with its 10 nm U-segment processors, these processors appear to be falling short of Apple's performance/Watt requirements. An example of chip design control Apple expects from its hardware suppliers can be found with AMD. The Sunnyvale-based firm supplies Apple with its most efficient bins of Radeon GPUs, and in cases such as the "Navi 12," Pro 5500, and Pro W5700X, even reserves graphics SKUs exclusive to Apple (not sold to its AIB channel).
In our opinion, Intel could have done more to retain Apple. The engineering department certainly rose to the occasion, developing "Lakefield." While performance numbers are still under the wraps, "Lakefield" is the kind of chip one would expect in an Apple portable - an extremely power-efficient client-segment processor capable of sub 1-Watt idle, high burst performance, and great customization flexibility thanks to its Foveros Packaging that lets system designers pick and choose the I/O components they want specific to their designs. On the other hand, the business-end of Intel may have fumbled with Apple. If designing chips that match Apple's requirements didn't work, Intel could have used the nuclear option - of pitching the x86 machine architecture itself.
Since its inception, Intel has licensed the x86 machine architecture to over a dozen companies. There are currently only two active licensees - AMD and VIA. The rest either withered away, or consolidated (VIA consolidated CenTaur and CYRIX). There have been no new licensees in at least the past 15 years (not if you don't count the sub-licensing of Zhaoxin by VIA or to THATIC by AMD). Apple could have been the first new x86 licensee in a generation, and with a little assistance, could have developed its own x86 SoCs. Apple pays for Unix in the era of Linux, and it would have surely indulged a well-crafted license deal with Intel.
Apple's departure from x86, despite amounting to a paltry percentage of Intel's sales, could reshape the client-computing segment. In our opinion, the success of the Arm machine architecture on Macs presents a greater threat to Intel than even x86 licensee AMD, as it could trigger other semiconductor firms with deep pockets and Arm licenses, such as NVIDIA and Samsung, to develop "high performance" Arm SoCs of their own, for thin-and-light notebooks. The "Wintel" era is long gone, and Microsoft is only too happy to indulge and grow its Windows-on-Arm ecosystem.
The future of Intel's client-segment silicon looks increasingly similar to that of Arm. Highly modular IP blocks, including from third-parties, integrate on innovative new packaging formats, such as Foveros, with an overwhelming focus on performance/Watt, thanks to hybrid cores, idle power-draw, battery life, and performance in bursts. A lot is riding on the success of the tiny "Lakefield."
52 Comments on Intel Gives its First Comments on Apple's Departure from x86
Apple is going to be a nice test case really for how such a shift might happen.
Apple's been patching Intel's shit for years and it's no surprise they're making moves to get out from under that burden.
In the form of THE most generic corporate statement imaginable.
This for sure will hurt Intel unless other big customers absorb the now available quantity.
Ampere’s Product List: 80 Cores, up to 3.3 GHz at 250 W; 128 Core in Q4
There's also Huawei with basically the entire CCP backing them, if you think Intel's struggles will be over just by a simple transition to 7nm(++) then you're dead wrong. The real pain could well be ahead, especially as more serious contenders with deep(er) pockets emerge in the server space!
It was inevitable since it's Apple's modus operandi and Intel has been a nice stop gap since the times of PowerPC. There was nothing Intel could do stop this.
There were a lot of things Intel could do "a long time ago" to enter the mobile space but their lack of foresight led to the need for and birth of ARM and you know the rest.
And a custom design has one more advantage: you won't be able to compare systems and point out Apple is the more expensive choice, because they now have pixie dust in the CPU :P
But the fact of the matter is this isn't about performance, or thermals, or anything that Intel did wrong or ARM did right; it's about control. Apple wants to control every aspect of their ecosystem, and people who buy overpriced Apple products are stupid enough that they'll happily drink whatever Kool-Aid Apple whips up to convince them that ARM-powered laptops are somehow superior to x86-powered ones. After all, it's not like they'll notice the speed difference anyway - there are only about a dozen real apps in the Apple ecosystem, the rest are app store trash that will run just as well on an ARM potato as they would on x86. I mean, what are Intel gonna say? They almost certainly knew this was coming months if not years ago, and they don't want to burn any bridges with Apple in case this grand experiment flops and Apple comes crawling back, so best not to say anything at all.
I can also guarantee that said deal was renegotiated to give Intel even more money once Apple announced their own ARM IP. At that point Intel had no incentive to help Apple anymore while Apple still needed Intel for laptop CPUs.
And Apple likely wouldn't have been too opposed to paying through its teeth for Intel's CPUs. After all, they just pass the costs downstream to their idiot customers, so it really doesn't affect them in any way.
The issue at large is, they are currently marketing for things that are not presently launched. They are hyping. CNT transistors are all so wonderful, just not in circulation.
Intel instead choosed to milk the market and make profits because they thought they were unmovable from their leader position. Things can change very quickly indeed. And in the technology market, doing catch up against a competitor that is moving foward quickly is really hard.
This is my opinion, and not a fact, but i think that the greatest Intel advantages was their Fabs and process. They seem to have lost that advantages to TMSC. Doesn't matter now if they are on par of behind, they are no longer in front of the competition. On the Architecture side, their process allowed them to still be competitive even when they had crappy arch.(like P4). I am not sure if all intel Arch were so better than the competition or just they had the better process to use. On the other side, AMD, didn't had a chance when they got their crappy arch (Bulldozer) because Crappy Arch + average process is a good receipe for failure.
As for Apple, they would have make the move at some point no matter how Intel performed. They just made their hands on CPU design with Iphones and IPads and they were confident that it was a good time to do the move.
It's just a really bad time for Intel. (and a good one for Apple).
And Intel didn't sit on their laurels either, they designed Ice Lake. Their problem is called 10nm so they can't actually sell Ice Lake.
If the move to ARM wasn't this reachable to apple, they would have long ago made a deal with AMD for Ryzen based Macs. But there might be a potential to do even better with ARM on Mac, who knows? They could more likely scale those iPad chips to a much more powerful form than even AMD could supply.
They could literally flip the whole damn table with this move and nobody will be able to compete. I think a lot of people should be scared like hell right now. Because if Apple can pull this off, all the Apple haters and people who call other people idiots for buying Apple products will be running to Apple for their performance. Nobody cares about the DIY segment outside of this forum, especially not for notebooks which most people already use as their primary device.
I wonder what you folks think about Apple opening up the SDK, packaging, and signing of system and kernel level extensions (is drivers for you Windows folks)?
This is a strange departure for Apple, where MacOS has always discouraged installing kernel extensions, even though there is a thriving community of “mods,” driver enhancements, and even custom/ported drivers for MacOS. For example, the github for 64-bit LSI RAID drivers has regular commits and new features for over a decade.
Part of me thinks they have to do this to ease the transition, but in the past they’ve only offered the SDK to licensed partners.
I don’t know anything about RISC vs... TRISK or whatever, and what that means for devices like USB storage devices. What does all of this mean?