Apple uses an Arm-based architecture for its A-series chips in the iPhone and iPad, and each year, those chips get faster and more efficient. In fact, when introducing the latest A12 and A13 chips, Apple has made it a point to emphasize that these chips are faster than many Intel-based chips in competing devices.
The 2018 iPad Pro models with A12X chips, for example, are close in speed to the 2018 15-inch MacBook Pro models… With Apple closing the speed gap between Arm and x86, there’s no reason why many of the company’s notebook machines (and even desktop machines) couldn’t be powered by Arm-based chips instead of standard Intel chips…
Apple is said to be aiming to transition to its own Arm-based chips starting in 2020, though the transition period could take some time.
MacDailyNews Take: The sooner the better!
Intel is well-past its glory days. Today, Intel’s claim to fame – besides not being able to make modem chips very well – is peddling inefficient, embarrassing, fatally-flawed junk. — MacDailyNews, May 15, 2019
I don’t know.
Until one is put into a full blown personal computer with a desktop OS so we can draw comparisons… I don’t care.
This has always been one of my commenst about an A-series, ARM based processor in a Mac.
How many PCIe lanes does the current A13 support? None. How many and what kind will be required for a top of the line Mac? > 48 and 5.0 (spec has been out for almost a year)
How many USB ports does the current A13 support? None. How many and how fast will be required for a top of the line Mac? > 4 and USB 4.0 (this will include Thunderbolt)
How many Ethernet ports does the current A13 support? None. How many and how fast will be required for the top of the line Mac? at least two and at least 10 Gbps
The list goes on and on.
Sure Apple could develop a separate “sister chip” (what we used to call “glue chips” in the 70s and 80s that sat between the CPU and the rest of the world), but that is extra cost and extra design effort every time Apple changes the A-series processor. Further there would almost certainly be multiple versions of that chip required: one for MacBooks, one for MacBook Pros, one for iMacs, one for Mac Pros (if that machine ever shows up) and maybe more such as for an iMac Pro.
Adding all of this capability into an A-series processor will do two things: 1) require more power and 2) slow it down. These two things are just facts. People can try to dispute either or both of them, but they are living a fantacy if they think either or both of these things won’t happen.
When you add these things on top of requiring a Rosetta like interpreter to run legacy code for many years it will likely be several more years before we see A-series Macs shipping to the general public. However, I am absolutely certain that today Apple is running macOS on A-series processors in a lab way underground under the spaceship that requires retina scans and finger print scans and blood samples and > 32 character passwords to get into it. (Just kidding on the password part. 🙂 )
Please don’t get me wrong. Several years ago I was one of the people saying that we’d all start to see A-series Macs for sale to the public before 2020.
However back then Apple’s A-series processors were making processing speed increases of as much as 50% year over year with no end in sight. Now processing speed increases, on average, of the A-series processors is roughly 10 – 20% year over year. That’s not much more than Intel’s capabilities increases in recent years. So now Apple’s A-series processors are chasing a moving target and just barely gaining on that target. (Don’t even try to say the A-series has already surpasssed the capabilities of the Intel processors. One: the A-series proecessors don’t have the capabilities I listed at first. Two: they have, in limited cases, surpassed the i-3 and in some cases i-5 processors. They have not surpassed any i-7 or i-9 processors.)
Thank you for your valuable insight.
Great post, Shadowself. Lots of things to ponder – things that, in my ignorance, I had not considered.
I, too, was predicting A-series Macs by 2020. And that prediction may not be too far off. It is certainly possible that Apple has already addressed many or all of the I/O issues that you identified in your post. If so, then it is reasonable to expect that Apple has maintained tight security to prevent leaks until an A-series Mac is ready for a public reveal. But Suppose that it is more likely that another year or two or three may pass before an A-series Mac is ready for retail.
Another consideration that interests me in the application of A-series processors for the Mac – is the A-series architecture compatible with parallelization? Ideally, I would like to see MBPs with four or eight A-series processors, iMacs with 16 to 32 processors, and Mac Pros with 64+ processors. Does the A-series architecture support this kind of scaling? Or is this yet another issue which would drive a custom version of the A-series SoC for Macs?
A desktop OS runs several times more threads just by being on. In early Windows days, we often booted straight to DOS to give applications the whole machine.
The A13 chip has multiple cores. The A-series CPUs have had multiple cores for many years. Clearly Apple has the ability to put multiple cores into a single processor. Besides iOS is multi threaded. (It’s just not anywhere near as multi threaded as say macOS Catalina or Windows10.) That is not the issue. It’s just scaling. Could Apple do a 16 core A-series processor? Absolutely. That chip would require more power than the current A13, but Apple could do it.
With regard to booting into a single thread system: A great example was how MIT was monitoring one of their sites back in the 90s. They had it on an old (old even back in those days) Mac Plus that was running, IIRC, System 3.0. That version of the OS allowed only one application/thread to run at one time. In fact when the OS loaded the application the majority of the OS itself unloaded leaving the application running virtually by itself with just a few core services available to support the application. The people in charge of the MIT site found that this particular configuration was more unasailable than anything else they tested, and you can bet the MIT guys tested lots of variants before deciding on that particular Mac configuration.
What I was contrasting is that it would need those cores for a desktop system due to the need to service more threads of execution. Or we disable parts of the machine, such as multitasking (my DOS example) to allow a given CPU to perform a given task faster, for the task to have more of the CPU’s attention.
That, and as you correctly noted, any given calculation requires a theoretical minimum of energy to be performed. A not heat generating chip still requires power. You correctly point out, more cores, more tasks, but more power.
You do realize these apple chips are for embedded mobile hardware yet they are outperforming Intel desktop chips. They could put 5 of these things in a MacBook Pro to run more ports than a Mac Pro if they wanted.
Did you even read my post?
If an A-series processor supports zero PCIe lanes then putting five of them into a MacBook Pro supports how many PCIe lanes?
Do the arithmetic that you likely learned back when you were seven or eight years old. (It’s not even math.)
That’s right. It’s still zero PCIe lanes.
The same goes for USB. The same goes for any other set of ports other than Lightning.
I think it’s inevitable. Before, when Macs switched to Intel, there was a need to keep up with Windows machines running on Intel (or compatible). The PowerPC alliance wasn’t cutting it. There was also the advantage of running Windows applications directly (or more directly) without the inefficiency of emulation. Back then, some software didn’t exist for Mac.
Today, there’s much less need to run Windows on a Mac. For most users, there’s less need for specialized apps in a web-based world. Efficiency is more important than pure power. I’ll bet there are prototype ARM-based Macs in Apple’s secret lab, running the current version of MacOS and key apps, right now. Just as there were secret Intel-based Macs well before MacOS X Tiger (10.4.x) when the actual PowerPC-to-Intel switch was announced.
ARM-based Macs will start as the new “MacBook” (currently open spot). It’s user interface will look and feel like any current Mac, like Intel-based Macs versus PowerPC-based Macs during that transition. The “pro” Macs will be the last to transition, because those users rely more on specialized and complex applications.
Ken, you’re correct about intel Mac in Apple’s lab years before Jobs announced them. Back in the day I was working for Apple as an engineer and I saw them running in the lab. I had to sign an NDA of course.
I read that some applications ran faster in emulation mode.
I was just noticing what is happening with he new Surface Pro X with “full Windows 10” running on an ARM processor. People keep running into incompatible software.
Will be interesting to see what Apple’s approach will be, particularly in the initial stages of this move. Will it be a new product category or a MacBook or an Air that sports Arm based chipsets to start with. The decision will be important in determining whether indeed it will be a full Mac device or effectively a beefed up iOS device masquerading as a Mac. Apples work particularly with Adobe to get ‘real’ versions of Photoshop and now we hear illustrator on iPad OS, plus web/cloud versions of many applications and the ongoing work to make iOS apps work on Macs and possibly to a degree vice versa too, does muddy the exact route they are travelling with this concept as clearly so much is not presently even in the public domain to make precictions clearer. There are bound to be initial incompatibilities whatever they decide so the best solution is beyond my pay grade to determine with both engineering and marketing aspects to consider. So what might make sense for a low end Air might be totally unworkable for a high end workstation as others have alluded too.
I also believe it is inevitable, but ” there’s no reason why many of the company’s notebook machines (and even desktop machines) couldn’t be powered by Arm-based chips,” is certainly ignoring the reality of software compatibility (as others, including theloniousmac, point out). I don’t see Apple putting ARM in Macs until/unless there is a Rosetta equivalent to enable legacy x86 compiled software. We may see AMD chip replacing Intel before ARM is ready for a true takeover.
Why?
Being dependent on TSMC is a bad idea. Intel fabs are in the USA.
TSMC already fabricates tens of millions of A-series processors per quarter for iOS devices, and Apple is far more dependent upon the iPhone than the Mac in terms of revenue. Therefore, your argument lacks some logic. While maintaining Intel as a processor supplier would improve supplier diversity, the Mac and iOS processor worlds are currently isolated. Furthermore, the transition of Macs to A-series processors would only add about 10-20% to A-series SoC demand, depending on the average number of processors per Mac.
Ummm… this is all interesting except one small problem. If you read Wired’s article on the A Series procs, you would remember that the A series hasn’t followed ARM in the last 5 or so generations atleast, so there is absolutely no telling what dev has been done to Apple’s processor line. Unless you are in the Apple Labs, its just conjecture.
Since about the third or fourth generation of the A-series processors Apple has been licensing the ARM instruction set and a few other odds and ends of the ARM processor IP. Apple has not been licensing the ARM hardware designs for many years.
This is not to say that the A-series has not followed ARM. They have. When ARM changes their instruction set Apple does too.
Apple’s implementation is different in its specific hardware implementation. They’ve taken lots of elements from things like the old Powerficient stuff and judiciously folded that in.
Loss of Windows compatibility will be the death of the Mac computer.