There were a lot of incredulous and dismissive comments when SemiAccurate said that Apple would be going Intel free in the next few years. Those sneers soon became begrudging acceptance as people dug in to the details, so we thought it was time for an update to the players and timetables.
A lot has changed in the last year and nearly a half, but the fundamentals of what is going on have not. Apple is going to make their own CPUs in for laptops, and and at that point, Intel becomes an unnecessary supplier of overpriced silicon. This eventuality is quite inevitable, the moves Apple is making lay out an obvious path from the current products to the new ARMBooks.
First up is the A6 CPU, a custom ARM architecture core developed by Apple. This chip happened quite a bit earlier than SemiAccurate expected, we thought the first full core from Cupertino would not arrive until the second generation ARM64/V8 SoC hit the market. That said, the iPhone5 makes it pretty clear that Apple desires to, will, and is more than able of making a custom ARM core. A year ago, people said Apple had no possibility to make a custom core at all, now you can buy one.
Apple has a history of three or four generations of uncore design for their SoCs as well, and the custom core was the last piece in moving from a fruit sticker on a Samsung part to Apple the CPU design house. Given the performance and power consumption of the A6, it is pretty clear that so called first attempt is not a third rate effort. Given the people behind it and the previous Apple SoCs, this is not a surprise. Apple has some world class CPU architects, 4+ generations of ever increasing design complexity, and that knowledge shows in the silicon.
Next up is the OS. Have you noticed that OSX releases of late, well, to be blunt, suck? It’s not that they suck as a stand alone OS, but they take away a lot of the freedoms and flexibility that the Mac desktop OS user has come to expect. Bit by bit Apple is removing all of the parts that make OSX something other than a phone class OS. The UI changes, the App store, and the overall feel of the new OS move it more and more toward a slightly open iOS, not a UNIX core with a slick GUI on top. It is being progressively closed down and phone-ized. Any guesses as to why?
The hardware that is a MacBook is also becoming more device like with each iteration. Upgradeable RAM? Thing of the past. Configurable options? Nope. HD upgrades? Good luck with it if you try. The pretty metal thing that was once a laptop is now very device like in physical attributes, upgradeability, and serviceability. It is nothing like what you would have called a laptop a mere two years ago.
These three main trends, and a few happening in the background make Apple’s intentions pretty obvious. They are making laptops into big iOS devices step by step. No single change is too radical, no one is too quick, just a progressive chipping away at what a MacBook is. What is left is more and more reminiscent of an iOS device with a keyboard and bigger screen.
Slow and steady does not mean there are no serious problems remaining between the present and a uniform unconfigurable nirvana. For Apple to move to a laptop with an internal custom ARM based core, they would need to have a 64-bit OS running on a 64-bit CPU. iOS is not there yet, but going from 32-bit to 64-bit is not a big deal. Moving the other way is. More memory and more instructions are rarely a problem, especially if you have control of the entire code base like Apple does.
The real headache is the hardware, ARM64/V8 chips are not there yet. Nvidia was promising the financial analyst set that they would have a 64-bit Denver out in the market by Q4 of this year, and be the first 64-bit ARM chip on the market. We scoffed. Nvidia was promising the impossible to Wall Street, and internally there was no question that they knew it was impossible. That said, the reason for the Q4/2012 (Or sometimes Q1/2013 depending on the gullibility of the analyst being briefed) date was that other chip makers were expected to have ARM64 devices on the market by mid-2013. Nvidia had executive egos to keep inflated, not to mention stock prices.
In the intervening year and change, it has become quite clear that any realistic ARM64 based phone SoC, be it the stock ARM or custom V8 ISA, will not be out in that time frame. Yes, there are a few that will give it a shot, but nothing polished enough for high volume low power phone use. This entire ecosystem looks to have moved back a year or more. We will see what happens when, dark horses abound. Estimated targets of mid to late 2013 are now firmly in 2014, but work continues at a very fevered pitch.
The fab difficulties that Apple is facing are now much less of a theoretical problem, the TSMC tapeouts have fallen flat, and A6 is still being made at Samsung. Given the love between both companies at the moment, SemiAccurate doesn’t feel this is going to be the case for much longer, but Apple has nowhere to go for the time being. Conversely, Apple is the single volume customer for Samsung’s foundry division, and it is a very profitable relationship for the Korean giant.
Silicon fabrication is a really thorny love/hate relationship that neither can afford to break out of despite a burning desire to do so. Until Apple can come up with a foundry partner that will provide both the quantity and quality Apple needs, an ARM64 core for MacBooks is not a viable option. That said, there is still time, and a lot of deals are being made in the dark corners of the foundry world. Stay tuned.
That brings us back to Intel. Apple and Intel are not exactly on speaking terms right now. Why? Can you say Ultrabooks? Yes, shiny things for the stupid they may be, but original they are not. Intel is subsidizing the heck out of the market to desperately try and fend off tablets while making PCs desirable. A year in Ultrabooks are a miserable flop, but the rhetoric seems to ratchet up proportionally to the depth of the failures. One problem for Intel is that the message behind the rhetoric is basically an anti-Apple marketing campaign. If there was any hope of salvaging the Apple relationship, incessant Ultrabook drumbeating pretty much set the remaining bridges on fire. There are a bunch of other issues there too, but we can’t talk about them for a bit.
In the end, Apple is moving all their hardware and software to an iDevice paradigm. That paradigm progressively moves up the performance ladder, mapping software aside, and the bits already above that baseline already are being sanded down bit by bit. The gating factors now, as they were last May, are a suitable ARM64 CPU/SoC and a place to make it. For now, neither are imminent, and things appear to have slipped a bit for everyone. Timetables may have stretched, but nothing fundamental has changed. As we said, it is a done deal.S|A
Latest posts by Charlie Demerjian (see all)
- More on Intel’s 10nm process problems - Sep 17, 2018
- Intel puts out another 14nm 2020 server platform - Sep 11, 2018
- Why Can’t Intel Supply Enough 14nm Xeons? - Sep 10, 2018
- Intel can’t supply 14nm Xeons, HPE directly recommends AMD Epyc - Sep 7, 2018
- AMD reintroduces the Athlon name with two CPUs - Sep 6, 2018