It doesn't. It's not really a certainty that Apple can improve significantly on power per watt on high performance computing. Current A-CPUs doesn't reach 5Ghz on 16 cores. It's a different game than making iPads.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Apple Hires ARM's Lead CPU Architect Amid Rumors of ARM-Based Macs as Early as 2020
- Thread starter MacRumors
- Start date
- Sort by reaction score
It doesn't. It's not really a certainty that Apple can improve significantly on power per watt on high performance computing. Current A-CPUs doesn't reach 5Ghz on 16 cores. It's a different game than making iPads.
Dunno, going from 2+4 to 16 core doesn't sound like much of an issue.
Well and how much computing could be gotten out these dies if the were in package suitable for desktop size cooling and power delivery?
Probably more, but people here tend to believe that ARM is somehow magical and that it can provide significantly lower power usage than x86-64 for high performance computing just because it's good in tablets.
The sheer amount of resources and manpower of Apple means that they might be able to contribute to moving CPU technology forward, but I don't see that this has anything to do with whatever architecture they use. They use ARM because that's simply what they have a license for. Not because ARM is magically better than x86-64.
The real reason for Apple to use ARM is that it allows them to choose their own pace at which their CPU technology moves forward, instead of being stuck with companies like Motorola and Intel that often derails progress. Whether this is worth it right now is questionable, since AMD is about to move x86-64 by leaps and bounds.
Last edited:
Those pushing ARM for Macs say it will be 'much faster', and in the same breadth that it is suited for bottom of the range MacBook Airs (not MacBook Pros or Mac Pros). The argument and proposition is self-evidently internally inconsistent, ridiculous and absurd.
Im a scientist, I run stuff which is often on old machines and operating systems from across the decades. Apple was a Swiss army knife where I could run anything. Now its becoming a turkey baster making a single dish look shinny and pretty.
Couldn't agree more. An ARM based notebook line-up would kill the whole idea of the Mac. ARM is not a good general purpose architecture. It's better than it was, If Apple seriously wants to get motivate intel, the best thing they could do is throw some business towards AMD. Competition between the two would do more to spark innovation than simply going with ARM.
This is a solid post. I think Apple's investment in their own ARM SOC is focused on their mobile computing and other platforms (AppleTV, etc). They have been doing a great job in this space with their own silicon and this hire will help keep their momentum going.
While technically possible - I don't think an ARM MacOS is the way to go. Desktop/Laptop OS really require the x86 architecture and MacOS has done well in this space too. It also doesn't make sense to introduce AMD as that complicated Apple's supply chain which could add cost.
You're talking about an OS [iOS] that deals with a fraction of the threads running in preemptive relative to XNU kernel in x86 with several hundred to even a thousand or more threads preemptively.
Geekbench methodology: https://www.geekbench.com/doc/geekbench4-cpu-workloads.pdf
Nothing in this practically reflects a real world workflow. You're talking about a lock down staged testing set against a narrowly specific set of operations that 99.99999999% of the world never uses.
There is no launching of major applications and running automated sessions within them across various devices and/or platforms.
It's the #1 reason benchmarks are utterly useless.
iOS has a sandboxed platform, through and through with a kernel that runs a fraction of the processes macOS runs just to boot up, never mind continously. Those processes eat at CPU time continuously.
If ARM CPUs were world beaters they'd own the server space. They are toys in that space. They'd own the scientific workstation space. I can go on and on. They own nothing but the mobile space because your phone/ipad/apple watch, etc are locked down with dozens of tasks running and then preemptively halted between tasks, with very tight memory management to keep the system from lagging.
When you have a Mac Pro with 1.5TB of RAM hitting the market, do some deep thinking on how much is going on in that machine.
You clearly have no idea where ARM performance is going. Look up the Fujitsu A64FX implementation of the ARM architecture: it is going to be used in the Post-K supercomputer, and also look up features like the SVE. There is no reason an ARM implementation cannot compete in performance with intel high end CPUs, it is only a matter of implementation. in fact, the architecture leads itself better to high performance, and because of this, nothing prevents it from matching intel's performance at lower power consumption. The only reason these chips haven't been built in the past is that the focus was on embedded. But Apple's implementations of the ARM architecture are already good for laptop and desktop applications (look at the benchmarks for the iPhone X CPUs).
AMD doesn't complicate the supply chain. AMD uses TSMC right along with Apple for all their CPUs.
The bulk of Apple customers don't need high performance computing. I wouldn't sleep on their abilities, it's not like I am suggesting throwing an A series into a desktop. This would be a chip designed for laptops and desktops.
No, you need performance. You want to run native. And you argue with memories (I don’t have) of long obsolete hardware.
How do you arrive at that conclusion?
They might be decent at benchmarks, but can they handle real world laptop/desktop paradigms where you have hundreds of process threads running concurrently?
If background tasks consume more than a tiny fraction of CPU time you either run a very badly designed OS or are infected by malware (or both).
Plenty of small ARM devices that run Linux on a scale similar to what an x86 device with the same constrains would do.
Problem is noone has bothered making a desktop class ARM CPU/system for decades, hence there is nothing that could compete with a Xeon based MPro (or even an i3/5/7 based MacBook, MacMini or iMac).
Oh come off it, from back then I didnt see too many sour grapes from professional G5 users over the Intel Mac Pro, it's mostly been high end big spender prosumers who didn't really need a G5 tower who felt burned, professionals moved to x86 by and large long before the security updates stopped coming. x86 was a big step forward at the time and most people were very happy for it. And the new Mac Pro is even more targeted to professionals than the G5 or original Mac Pros ever were
So speaking of the new Mac Pro: anyone buying the new Mac Pro is either buying it for a business need *now* (in which case they're probably not worried about that particular piece of hardware more than 5 years down the line since the machine will be completely deprecated in value by then and be replaced) or very high end enthusiasts (and if you've got that kind of money to burn on a tower without a business need you're part of a very small minority of consumers). Recommending someone not buy a Mac Pro *now* because there's rumors that years from now the Mac Pro *might* move to ARM is silly. The Mac Pro is a capital expense for almost anyone buying it, and if it doesn't more than pay for itself by year *3* they either never needed it or are doing something wrong.
Just for reference I use a nearly fully loaded 15" MacBook Pro for work, not a cheap machine either, and our refresh cycle is 2-3 years (I'll be updating this year, I'm actually able to request a new machine now but I'm waiting to see what Apple might announce in the fall) - and that refresh cycle has been true almost everywhere I've ever worked.
Last edited:
Have you not seen the Photoshop demo on the iPad Pro? Opening a large 3GB 12,000x12,000 pixel image with 157 layers and they can manipulate/edit it just fine.
People can make up excuses all day long about running hundreds of threads, but if a heavyweight App like Photoshop runs and I can get work done with it then that thread discussion becomes irrelevant.
The only thing that matters is if I can run software I’m used to and it operates the same. This is all users will care about.
There's nothing inherently problematic about using ARM or other non-x86 procs in scientific computing/HPC, and only 2 of the top 5 on the top500 list of HPC systems right now use x86 processors. The bulk of the top500 are on x86 because it's cheap and code is already written for general purpose use on the clusters, but particularly as more and more code is optimized for GPU workloads, as been happening for the past 10 years or so, the actual CPU gets a less and less important. In the general server end of things x86 has momentum, but I think that will falter if true high performing ARM chips become more common, especially when you realize the top 2 languages used right now for are both interpreted anyway (Python and JS) and the other top languages are mostly pretty easy to recompile cross arch.
It's not like there isn't 3 just now Intel, GPU & T2? Can't see how they couldn't keep Intel about to boot windows & Arm to boost the performance much like Rosetta did but in reverse, & obv still have the GPU, T2 & whatever else they stick in it. Point is, it would be seamless to use - Apple would deal with backend stuff / translation to boost performance & apps wouldn't have to be re-written as to how they are just now. MacOS for those that use ARM only is still there, Boot Camp software as an add on would deal with windows on Intel, but with the intro of iPadOS & Marzipan apps, it's not like it couldn't be done, never say never!
What PC games are you thinking of that were written in assembly language? Those ARM based games are just a compile away.You know of any real PC games that are A12 based?
I did not refer refer to benchmarks. So you openly admit you did not check out real word performance. Thank you so kindly for proving my point
No, Apple is apparently doing some clever undervolting in macOS to get the best performance per watt.
[doublepost=1561649328][/doublepost]
Apple is already doing it with the T2 chip.
VMware Fusion and Parallels wouldnt be so popular if people werent running VM's internal to the MAC OS and there surely wouldnt be bootcamp if Apple itself didnt recognize the need for alternative OS on the MAC