Arm Processors with Mac Pro Level Performance Possible Today

[leman]

Is it worth all this effort and loss of x86 compatibility with the rest of the world?

Since Apple can potentially transcompile x86 binary to ARM binary on the fly, I can imagine ARM Macs running native x86 code without any overhead. This would require hardware extensions to the ARM architecture to better support some more exotic x86 instructions (mostly SIMD), but it can be done.

Now, booting Windows on an ARM Mac... don't really see that happening.

 

[whooleytoo]

Honestly, I’d rather have Intel compatibility at this point. I don’t really need more performance out of my system. In fact, I’m happy with pretty much everything my iMac does. I don’t sit down and think “man, I wish my processor was 2x as powerful.” I sit down and enjoy my speedy iMac and I’m happy.

The only reason I’d upgrade is to be able to update to the latest macOS, as I get obsoleted.

Yeah. Losing virtualisation would be a massive loss for me, longer battery life doesn't mean much (I rarely use my laptop without a charger), and quieter fans while nice aren't a must-have.

I guess Apple could launch ARM Macs alongside Intel, starting at the lower end. But it'd be hard for me to see it as anything other than a step backwards.

 

[theluggage]

The article is slightly misleading. These CPUs are developed for the server market and their performance comes from a large amount of parallel cores.

Its quite true to say that the Ampere processor is designed for server workloads and wouldn't necessarily be the chip to replace the x86 in Macs. 80 cores is probably too much for any single-user workstation (...I know nobody will need more than 640k of memory, the world will only need about 6 large computers etc... so let's add an 'with current software' rider).

However, the Intel Mac Pro range is in a similar boat - the individual Xeon cores are slower than the i7/i9 cores in lesser Macs and as you go up the MP range, any extra performance is mainly dependent on having software that can exploit multiple cores and GPUs. Where applications can parallel process, doubling the cores can double the performance - far more significant than making individual cores 10% faster.

Then looking at the other Macs: at the MBA/possible 12" MB replacement/low-end MBP end, battery life and thermals are vitally important - they're already compromising performance for power consumption and an ARM chip could deliver similar performance at lower power. I'd take the claims that the iPad Pro was 'as fast as' a MacBook Pro as, at least, meaning that it was fast enough for a 12" MB/Air. Plus, there's the cost saving (to Apple at least) and performance benefits of having a single chip with CPU, GPU, T2 equivalent (SSD controller, security, image processing etc.) and USB4 controllers all together.

Then, for all of the Macs that currently use Intel integrated graphics, smaller/cooler ARM cores mean more space on the chip for GPUs and other acceleration technologies. Imagine a Mac Mini that wasn't knobbled by the weakest iGPU that Intel can make. Or, iMacs with "afterburner" technology on the same chip as the CPU...

This isn't just about whether ARM can beat Intel on single-core Geekbench.

 

[Enerccio]

And give up native x86 virtualisation? Count me out.

The move to ARM offers nothing drastically better, at best you can hope to achieve the same level of performance - so why switch architectures if it's more of the same?

But muh battery. Muh needs 21 battery without charging, etc

 

[canonical]

So, someone somewhere has a prototype offering Intel performance using ARM, for half the power. So what? (a) Intel also have their skunkworks projects in the pipeline, and by the time they get the ARM competitor to market, the goalposts will have moved. (b) Mac Pro performance is nothing special anyway: the Mac Pro mostly gets its speed benefits from multi-processing, not from single processor grunt; (c) In order to justify the pain of shifting, and the mess that it involves, and everyone having to re-write their code natively for optimal performance, and the whole thing not dying a premature death, or killing Apple's computer line in the process, it needs to be not half the power, nor twice as fast: it needs to be 10 times better to justify the pain. This article just confirms to me that ARM is not a serious contender in the desktop space, and has no hope of taking on Intel in the foreseeable future.

 

[Enerccio]

That’s the CPUs fault?

Not yet! It’s obvious that the transition from x86 to ARM architecture is not a matter of months.

The issue with iOS is that it is locked down. Sorry but I want my computer completely open with root access, not the crap they put on tablets.

 

[StrangeNoises]

Odds are macOS is already living a secret double life on ARM processors!

I would be shocked if it wasn't. I'd consider it terribly remiss of them if they haven't been doing so for years already, even if it was some skunksworks project to "hey let's see if we can bring up osx on this ipad pro"...

 

[robbietop]

I love this guy, always so insightful and principled. The board made two mistakes in its history: Push out Jobs, and Push out Gassée.

As soon as Jean-Louis left, Sculley started licensing the Mac and the whole company began its decline.

 

[xyz01]

The 12” Macbook was a really good test bed. It is not difficult to see a similar ARM based model in the near future.
[automerge]1583754926[/automerge]
The Apple of today is not the same Apple that transitioned from Motorola/Power PC/Intel CPU’s. If Apple makes another transition to in house ARM based CPU’s you can bet it will be done in a way where users will not notice any difference.

On the contrary. If they do transition, users will notice it a lot. When transitioning from PowerPC

• Intel had a massive performance advantage compared to what the Mac was using
• This enabled running apps with Rosetta, without users losing that much performance
• There was a small user base (compared to today)
• The system had already just transformed from OS9 to OSX, so there wasn't that much legacy around anymore
• Virtualisation capabilities were not lost.
• CPU compatibility with the market leader (e.g. for hand tuned part of libraries/engines) were not lost

Here, the CPUs will be slower, emulating x86-64 on the ARM would make these already slower CPUs run the SW even worse, gaming engines and the like would not work, running virtual machines (especially for Windows) would stop running, far more old software exists - causing more problems, and far more people use the system and would complain.

If the Apple went ARM only, I'd switch away. I think many others would too.

That doesn't mean you couldn't launch an entry level device, e.g. for those who would only use Apples software and some basic Appstore stuff. If all you need is a web browser, email, and some basic spreadsheet /word processing, it could handle that.

 

[chucker23n1]

I wouldn’t use John Gruber as a benchmark for iPad. He’s a Mac guy. And when Gruber really doesn’t like something he’s not shy about saying so. There are people like Federico Vittici, Myke Hurley, Jason Snell that do a lot of work on iPad. Yes they’d agree multitasking isn’t perfect but they’re not nearly as harsh about it as Gruber is.

Yes, absolutely.

Though, honestly, the contortions those people go through to make an iPad just a bit more like a poor man's Mac are… wild. Listening to last week's episode of Upgrade made me want to throw my headphones in the river. (Yes, I understand the benefits of detaching the keyboard and trackpad, and having a more lightweight device. But I went through all that with the Windows 8 era, and it wasn't great. And so far, it doesn't really look like Apple has that figured out much better than Microsoft did back then.)

 

[throAU]

Will Apple still insult their consumers with ridiculous prices?

Does an ursidae defecate in the forest?

 

[Enerccio]

Does an ursidae defecate in the forest?

No idea, but Pope sure will

 

[CarlJ]

Odds are macOS is already living a secret double life on ARM processors!

Given that macOS and iOS share much of the same code, and iOS obviously runs on ARM processors, I think it's pretty much a certainty. Not to mention that they had OS X running on Intel for a long time before they announced that shift.

 

[cmaier]

So, someone somewhere has a prototype offering Intel performance using ARM, for half the power. So what? (a) Intel also have their skunkworks projects in the pipeline, and by the time they get the ARM competitor to market, the goalposts will have moved. (b) Mac Pro performance is nothing special anyway: the Mac Pro mostly gets its speed benefits from multi-processing, not from single processor grunt; (c) In order to justify the pain of shifting, and the mess that it involves, and everyone having to re-write their code natively for optimal performance, and the whole thing not dying a premature death, or killing Apple's computer line in the process, it needs to be not half the power, nor twice as fast: it needs to be 10 times better to justify the pain. This article just confirms to me that ARM is not a serious contender in the desktop space, and has no hope of taking on Intel in the foreseeable future.

Intel has been stagnant for five years, but they have a secret skunkworks project that will change everything? Sure.

 

[kemal]

There is nothing magical about the ARM instruction set that will make their 6 Watt chip keep up with an intel 150 Watt monster.

 

[citysnaps]

Will Apple still insult their consumers with ridiculous prices?

Apple prices might feel ridiculous to you and others who cannot afford their products. For others, they're very well-priced.
[automerge]1583769111[/automerge]

I actually don't think Apple should have full control of their hardware in house as a consumer (although from a business perspective it may make sense). Apple often stifles their own innovation due to "Not Invented Here" syndrome where they rebuild something that won't make them any money and has already been done elsewhere. Let a chip maker (perhaps Ampere) build powerful chips for multiple players and focus on that to make the best chips possible. The Snapdragon 865 is a better proc than the A13 because a company that builds silicon for its bread and butter is building it (companies like Samsung don't use it well but that's a different story). Specialization of companies makes sense at this boundary in my mind.

Apple for sure will develop their own CPUs. Otherwise they'd be relegated to using essentially commodity parts available to their competitors, with similar performance, and with no ability to add their own proprietary secret sauce.

 

[newdeal]

Saw this article on Engadget this weekend

AMD CPUs for the past 9 years are vulnerable to data leak attacks

It's not just Intel chips that are vulnerable to hard-to-fix security flaws. Researchers at the Graz University of Technology have detailed a pair of side channel attacks under the "Take A Way" name that can leak data from AMD processors dating back to 2011, whether it's an old Athlon 64 X2, a...

www.engadget.com

The Intel processors have always worked great for Apple so I don't know why they would want to switch to ARM

Always is a bit of a stretch since they used to use PowerPC cpus and only switched to Intel in 2005.

 

[BGPL]

We would see tremendous innovation in desktops if Apple starting having control over their ARM processor development. I don't think the biggest improvement would be CPU improvement, but actually GPU improvement - something we desperately need in MacOS. The GPUs in iOS devices are quite impressive. My kids' iPad Air 2 can run fortnite extremely well, better than a few of my PCs.

Nonetheless, it would put a fire under Intel, just like the iPhone pushed every tech leader to focus on personal communications. Or more recently, how Tesla caused EV innovation across the board.

 

[Zdigital2015]

So, someone somewhere has a prototype offering Intel performance using ARM, for half the power. So what? (a) Intel also have their skunkworks projects in the pipeline, and by the time they get the ARM competitor to market, the goalposts will have moved. (b) Mac Pro performance is nothing special anyway: the Mac Pro mostly gets its speed benefits from multi-processing, not from single processor grunt; (c) In order to justify the pain of shifting, and the mess that it involves, and everyone having to re-write their code natively for optimal performance, and the whole thing not dying a premature death, or killing Apple's computer line in the process, it needs to be not half the power, nor twice as fast: it needs to be 10 times better to justify the pain. This article just confirms to me that ARM is not a serious contender in the desktop space, and has no hope of taking on Intel in the foreseeable future.

That secret skunkworks project has been trying to figure out why yields on their 10nm process have been so low for so long without telling anyone that it is essentially untenable and that they are too proud too admit they failed. Then marketing takes over to try and spin the past three 14nm+++++++++ refreshes as the latest and greatest.

Intel boxed themselves in and still hasn’t figured out how to get out.

Tiger Lake holds a tiny glimmer of hope, but until the silicon is rolling off the line, it’s a unicorn’s pipedream drenched in dolphin tears.

 

[Freeangel1]

arm or intel
you can only shrink the dye of a computer chip so much. 5mm currently. maybe 2.5mm then what??
Dead end for chips?
parallel processing ?
Quantum computing?

 

[RogerWilco]

I believe Apple will use these advanced chips, but not in Macs. They will power Apple’s shift to cloud computing where Arm-based clients (the rumored “new” Macs) will access virtual machines running on remote servers. The service will be subscription based and will allow full use of existing Mac OS, Windows and Linux apps.

 

[TheShadowKnows!]

Still got around $700 for my Macbook Pro (Retina) after 6 years... which was $2300 when i bought it. So it ended up not being that expensive after all.

Nozuka. No doubt on/about the excellent used marketplace return on the pre-2016 MBP retinas.

Their resale cusp was reached-and-locked by the 2015 MBP.
But things have changed since 2015. Not for the better.

Walk through the resale values of MBPs from 2016 thru 2019. For example, use swappa.com as a guideline and check on their speed of closing (some posts many linger for many weeks, if too greedy.)

Apple and Apple MBP buyers at large disagree on the value of traditional ports, on the keyboard, on the magsafe, and on the value of physical function keys. And even on the size of the keypad. To add more disassociation, on the value of macOS Catalina.

i own two pre-2016 15-in MBPs, running Mojave, and although I would normally have considered upgrading I will not until my MBPs are declared EOL.

So, you are not wrong. And neither is the current state of the post 2015 MBP resale.
At least that is my personal view.

 

[robbieduncan]

I believe Apple will use these advanced chips, but not in Macs. They will power Apple’s shift to cloud computing where Arm-based clients (the rumored “new” Macs) will access virtual machines running on remote servers. The service will be subscription based and will allow full use of existing Mac OS, Windows and Linux apps.

I think this is very unlikely

 

[cmaier]

There is nothing magical about the ARM instruction set that will make their 6 Watt chip keep up with an intel 150 Watt monster.

You realize that it won’t be a 6W chip, right?

You also realize that scaling UP in power is a lot easier than scaling down, right?
[automerge]1583772708[/automerge]

arm or intel
you can only shrink the dye of a computer chip so much. 5mm currently. maybe 2.5mm then what??
Dead end for chips?
parallel processing ?
Quantum computing?

I started designing CPUs in 1992. Back then we were around 300nm (we used microns instead of nm, but I’ll use nm to keep it simple). Many people thought we were nearing the end of the road, and there was a lot of research about the next thing (unfortunately, I was involved in some of that - we were doing bipolar circuits in GaAs, for higher Ft, and because it was thought that vertical dimensions could shrink a lot more than lateral ones).

By 2000, we were around 130nm, and we were thinking about things like SoI, adding strain with SiGe, and even diagonal wires, because we kept being told that the end of the roadmap was nigh.

By 2005 (around 65nm), we had to start doing DFM techniques to “trick” the lithography into creating the design we had on paper.

Of course, the march of progress kept going, the nodes kept shrinking, transistor gates became three-dimensional, etc.

I think there’s a long way to go before we have to worry about not shrinking anymore.

 

[jerwin]

There's also this article from 18 months ago

https://www.anandtech.com/show/12694/assessing-cavium-thunderx2-arm-server-reality