Opinion: Apple transition provides huge boost for Arm

Bob O'Donnell

Posts: 79   +1
Staff member
What just happened? You have to imagine that yesterday was a pretty good one for the folks at Arm—the little understood, but highly influential chip design company. Not only were they able to report that their designs power the world’s fastest supercomputer, there’s also that little detail about Apple choosing to switch from Intel-based CPUs to Apple designed custom silicon built on Arm’s core architecture for future generations of Macs.

A word on the supercomputer news first. Every year at the opening of the ISC high-performance computing conference, the organization running it releases the Top 500 performing supercomputers. As with most years, this year’s list was utterly dominated by Intel-based machines, but there was a surprise at the top.

For the first time ever, Arm-based chips (in this instance, built by Fujitsu) are the CPU brains being used in the number 1 ranked machine—the Fugaku supercomputer, which is operated by the RIKEN Center for Computational Science in Japan. In addition to the prestige, it’s a huge psychological win for Arm, which has been working to make an impact on the enterprise computing world with its Neoverse CPU architecture for the last several years.

"Arm notched an equally impressive victory with the official unveiling of the long-rumored Arm-powered chips for next generation Macs. Apple doesn’t have the largest market share in the PC market—it’s around 7% or so overall—but its impact, of course, greatly outstrips those numbers."

In the personal computing world, Arm notched an equally impressive victory with the official unveiling of the long-rumored Arm-powered chips for next generation Macs. Apple doesn’t have the largest market share in the PC market—it’s around 7% or so overall—but its impact, of course, greatly outstrips those numbers. As a result, by making the official announcement of custom Apple Silicon for the Mac, which was designed leveraging Apple’s architectural license of Arm’s chip IP designs (though Arm is never mentioned in the keynote or any of the press releases for the event), Arm scored a huge gain in credibility and awareness.

Of course, awareness doesn’t translate to success, and as exciting as the development may be, there are a great deal of questions, as well as previous history, to suggest that challenges await. First, while Apple talked about switching to this new design to both improve performance and reduce power consumption, it has yet to show any comparative benchmarks to existing Intel-based Macs for either of those metrics. Of course, that’s likely because the silicon isn’t done. Heck, Apple didn’t even announce the name of the new chips (the A12Z Bionic chip in the developer system, and currently in the iPad Pro, is only an interim solution). My guess is that we won’t get any of these details until the end of the year, when the first-generation Macs with these new chips are unveiled.

Apple’s primary stated reason for making the move away from Intel to custom silicon was to improve the experience, so these comparative details are going to be critically important. This is particularly true because of the generally disappointing performance of Arm-based Qualcomm and Microsoft chips in Windows on Arm PCs like the Surface Pro X. The key question will be if Apple is able to overcome some of the limitations and truly beat Intel-level performance, while simultaneously offering significantly better battery life. It’s an extremely challenging task but one that Apple clearly laid out as its goal.

There are also many unanswered questions about the ability to pair these new chips with external GPUs, such as the AMD Radeon parts Apple currently offers in certain Macs, or any other companion chips, such as 5G modems. While Apple currently uses Qualcomm modems for the iPhone and certain iPads, the company is known to be working on its own modems, and it’s not clear if those will be available in time for the launch of a 5G-equipped Macbook (should they choose to do so).

As for graphics, Apple only uses its own GPU designs for its other custom parts for iPhones and iPads, but some computing applications require more graphics horsepower than those devices do, so it will be interesting to see if Apple offers the option to pair its new Mac-specific SOCs with external GPUs.

Finally, of course, there is the question of software. To get the best possible performance on any platform, you need to have software developers write applications that are native to the instruction sets being used. Because that can take a while, you also have to have a means to run existing software (that is, designed for Intel-based Macs) on the new chips via emulation.

Ironically, Apple has chosen to use the exact same playbook to transition away from Intel processors that it used to transition into Intel processors. In fact, it’s even using the same names (with the addition of a version 2) for the core technologies: Universal Binaries 2 are combined applications that run on both Intel CPUs and the new Apple custom silicon chips and Rosetta 2 is the software used to emulate Intel instructions. This time around Apple also added some virtualization capabilities and demoed the ability to run Linux in a virtualized container. However, interestingly, there was no discussion of Windows running on the new Mac. Presumably all the work that Microsoft and its partners have done to bring Windows to Arm-based CPUs should port over fairly easily to Apple designs as well, but the details on this are not clear just yet.

To the company’s credit, Apple did an impressive job when it created this playbook to move from PowerPC-based chips to Intel, so here’s hoping the same strategy works the other way around. While Apple made it seem like it was a fairly trivial task to shift from x86-based instructions to Arm, if you use its Xcode development environment, history strongly suggests that the transition can be a bit daunting for some developers. Apple did show functioning demos of critical Microsoft Office, Adobe Creative Cloud, and Apple professional apps running natively in the new environment. One concern Apple didn’t address at all was hardware device drivers. That was a key challenge for early Arm on Windows devices, so it will be interesting to see how Apple does with this.

One nice advantage that Apple and its developers gain by moving over to the same Arm-based architectures that it uses for the iPhone and iPad is that iOS and iPadOS applications should easily run on these new Macs—a point Apple was eager to make. As exciting as that first sounds, however, there is that detail of a lack of a touch screen on any existing Mac. Imagine trying to use a mouse with your iPhone, and you can see how initial enthusiasm for this capability may dampen, unless Apple chooses to finally allow touchscreens on Macs. We shall see.

"Apple ultimately chose to move to Arm to gain complete control over the Mac experience"

The last point to make regarding all of these developments is that Apple ultimately chose to move to Arm to gain complete control over the Mac experience. As good as Intel’s processors have been, Apple has shown with its other devices that it likes to own the complete vertical technology stack, and the only way to do that was to design the CPU as well. It’s the last critical piece of the puzzle for Apple’s strategy to control its own destiny.

Regardless of that reasoning, however, it’s clear that both Apple’s decision and the supercomputing win mentioned earlier provide a great deal of credence to Arm’s efforts. At the same time, it arguably puts even more pressure on Arm to continue its pace of innovations. For a company that so few people really appreciate and understand, it’s great to see how far and how wide Arm has pushed the boundaries of computing. Now let’s see how they continue to evolve.

Bob O’Donnell is the founder and chief analyst of TECHnalysis Research, LLC a technology consulting and market research firm. You can follow him on Twitter . This article was originally published on Tech.pinions.

Permalink to story.

 
Apple is playing the game very smart - and will continue to be rewarded by its stock investors (such as myself).

They got their parts from the best manufacturers they could which fit their needs and turned a profit, and now they want to focus on capturing more profit and making even more efficient chips.

Can't hurt!
 
Just FYI, as far as I know Rosetta 2 isn't emulation in a traditional sense, its translation. It doesn't convert instructions from x86(64) to arm at RUN time, it does it at build/install/maybe first run time.

Should result in much better performance assuming you can get it done reliably. They did show it running Maya and Tomb Raider which was very cool.
 
All of this was set in motion years back. I blame Intel for sitting on it's butt for far too long. It slit the achilles tendon the x86 for 10 years so it makes sense that alternatives would look extremely appealing. No one wants to place bets on processors that have gone nowhere for that long. Doing nothing cost Intel the mobile market and it costing them their primary marketshare right now.
 
All of this was set in motion years back. I blame Intel for sitting on it's butt for far too long. It slit the achilles tendon the x86 for 10 years so it makes sense that alternatives would look extremely appealing. No one wants to place bets on processors that have gone nowhere for that long. Doing nothing cost Intel the mobile market and it costing them their primary marketshare right now.
It has done no good for Intel to lag behind the world in chip production. Intel is stuck on 10nm chips. TSMC and Samsung are at 5nm. Smaller processes mean lower voltages, faster circuits, less heat to dissipate. In part, this explains AMD's recent surge with Ryzen and its server chips, manufactured by TSMC. Apple uses and will doubtless continue to use TSMC as a its chip fab for the ARM processors and the ones that also power the iPhone and iPad world. You can expect ARM-based MacBooks that are both fast and power-efficient, with long battery life.
 
It certainly is true that historically, the Macintosh had a huge impact. Microsoft would not have considered trying to bring the technology of the Xerox experimental window and mouse interface to the home market without Apple proving it was possible.
And the iPhone with its innovations sparked the whole Android phone ecosystem.
To say that the Macintosh today, with its limited market share, still has a strong impact, though, is, I fear, questionable. It is clearly dying, with makers of applications for its last remaining strength, the graphics arts field, steadily deserting it.
Apple's behavior shows that it isn't interested in turning that around, it prefers to milk any computer users still, for whatever reason, stuck in the Macintosh ecosystem with inflated legacy pricing. It isn't going to revitalize it with innovations, it's going to kill it with neglect.
Going to OS X saved them the trouble of maintaining their own operating system, and now going to ARM is saving them the trouble of supporting a platform different from the iPhone and iPad.
Maybe an iPad with a keyboard and a more powerful CPU is an exciting new product category that will gain some traction, but while Intel is in a little trouble right now, so even a few lost sales could be a problem, I wouldn't worry much.
 
Having recently installed and tested Armbian Linux on my Old Amlogic s912 2Gb TV Box I can say for certain is as good as any Windows PC for "desktop OS" Web browsing and Video Watching experience.
Now Waiting for the release of Amlogic S908X or Rockchip RK3588 for my upcoming energy saving PC/Movies browsing build running Armbiam or Debian Linux.
Hopefully we we start seeing more direct comparisons of ARM Vs AMD Vs Intel CPU's from Popular online reviewers.
 
This is a great move for Apple. While I have not been a huge fan of their phones, their computers have always been top of the line.

But IMO I would like to see them move away from Intel than going full ARM.

I would really like to see all Pro line models move to AMD Ryzen and all non pro devices move to ARM. While the Apple ARM chips are top of the line, they rely heavily on hardware acceleration for tasks. Sure you can run a couple very specific benchmark and have it show the apple cpu's being very competitive. Or do some encoding of video, but that is not a apples to apples comparison. Most of the time it is showing iPad Pro video encoding with hardware acceleration vs CPU encoding on Intel. When if you going for fast encoding times we all know a Nvidia GPU is going to do it faster than either option.

Apple's chips will be great for non pro tasks, but I don't see it replacing high end Ryzen chips any time soon. Ryzen mobile already offers better performance/watt than intel, and that will only get better for AMD.

I feel like Apple will need a Rosetta upgrade at a later point in time allowing ipad/iphone apps to run on x86 Pro based macs.

But if apple wants to blow tons of money on low volume chipsets. They can. Apple will never make money on a high core count chip for a Mac Pro when AMD has 64+ core units ready to go. People are already disappointed Apple didnt go with AMD for the new Mac Pro. Intel is a joke when it comes to top of the line high core count chips.


So to recap.

This is great for their normal device lineups. The Avg Joe will not see a difference and will only benefit from being able to run ipad apps. At some point in the future we will probably be able to boot the ARM version of windows 10, but I wouldn't expect that for a year or two.

This is not so great for Pro users, atleast not in the next 5 years. AMD Ryzen is the clear path if they want to treat their pro users right. Being able to do things like running windows in a VM, or being able to boot directly into Windows is a big deal.
 
This is a great move for Apple. While I have not been a huge fan of their phones, their computers have always been top of the line.

But IMO I would like to see them move away from Intel than going full ARM.

I would really like to see all Pro line models move to AMD Ryzen and all non pro devices move to ARM. While the Apple ARM chips are top of the line, they rely heavily on hardware acceleration for tasks. Sure you can run a couple very specific benchmark and have it show the apple cpu's being very competitive. Or do some encoding of video, but that is not a apples to apples comparison. Most of the time it is showing iPad Pro video encoding with hardware acceleration vs CPU encoding on Intel. When if you going for fast encoding times we all know a Nvidia GPU is going to do it faster than either option.

Apple's chips will be great for non pro tasks, but I don't see it replacing high end Ryzen chips any time soon. Ryzen mobile already offers better performance/watt than intel, and that will only get better for AMD.

I feel like Apple will need a Rosetta upgrade at a later point in time allowing ipad/iphone apps to run on x86 Pro based macs.

But if apple wants to blow tons of money on low volume chipsets. They can. Apple will never make money on a high core count chip for a Mac Pro when AMD has 64+ core units ready to go. People are already disappointed Apple didnt go with AMD for the new Mac Pro. Intel is a joke when it comes to top of the line high core count chips.


So to recap.

This is great for their normal device lineups. The Avg Joe will not see a difference and will only benefit from being able to run ipad apps. At some point in the future we will probably be able to boot the ARM version of windows 10, but I wouldn't expect that for a year or two.

This is not so great for Pro users, atleast not in the next 5 years. AMD Ryzen is the clear path if they want to treat their pro users right. Being able to do things like running windows in a VM, or being able to boot directly into Windows is a big deal.
While all fair comments, you did read that the world's fastest supercomputer is now ARM? I think your point of downplaying ARM performance is a bit overstretched.

Your point of handling windows translation is fair. I don't personally need it, as I suspect many people don't, but their are many who do use Windows under bootcamp and virtualization, and there appear to be no current abilities to handle this, is Apple going to keep certain pro devices, find a translation, or abandon this customer segment?
 
While all fair comments, you did read that the world's fastest supercomputer is now ARM? I think your point of downplaying ARM performance is a bit overstretched.

Your point of handling windows translation is fair. I don't personally need it, as I suspect many people don't, but their are many who do use Windows under bootcamp and virtualization, and there appear to be no current abilities to handle this, is Apple going to keep certain pro devices, find a translation, or abandon this customer segment?

Well that ARM supercomputer also uses 2.8x more electricity than the second on the list. It is mostly brute force. As the performance per watt isn't any better than existing configs. A Zen 2 and the Future Zen 3 arch w/ current/new Nvidia GPU's Will have better Performance/Watt if scaled up to that level. The biggest bottleneck is interconnect performance.

The most notable achievement with that new supercomputer is the interconnects. That is the one area that showed massive improvement compared to older supercomputers. This is an area that makes scaling up hard, and Fujitsu hit a home run in that regard.

So its not really a Win for ARM, but a Win for Interconnect tech. As ARM isn't yet a clear win in terms of Performance/Watt in the supercomputer area just yet.

I'm not down playing ARM, Apple just isn't going to be able to compete in the high end. Pro users want powerful systems that are highly flexible. Not just in Processing power but also in IO. We have yet to see how apple is going to solve the IO problem.

I'd totally buy a ARM based Macbook BTW, especially if I could dual boot Window 10 ARM. I use my desktop or work laptop for any Pro task. I think they will make great devices for normal users.
 
Well it’s funny to see AMD fanboys cheering at this...
By the way, I’m very curious to see if Apple will develop also high end CPUs for iMac. If they really want , they could deliver. A12Z, developed to work in a tiny iPad, in some test is competitive with Ryzen or Intel CPUs, and it is running at 3.8W TDP! We don’t know how it would scale on 28, 45 or even 65W computers like an iMac, with more cores.
Do not forget that Apple is bigger than Intel or AMD and they have a very tight relationship with TSMC for production.
 
Back