Who here is scared of change? I'm just against change for the sake of change. Intel has brought down their power use to 4.5 watts. Also, I suppose for most people a Chromebook will do...
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.
Intel CEO Responds to Rumors of ARM-Based Macs, Says Relationship With Apple Is 'Strong'
- Thread starter MacRumors
- Start date
- Sort by reaction score
Who here is scared of change? I'm just against change for the sake of change. Intel has brought down their power use to 4.5 watts. Also, I suppose for most people a Chromebook will do...
You're not wrong, but Apple doesn't necessarily have to be the license holder. They could work a deal with Intel/AMD to produce a generation of Ax chips based on X86/x64. My point was mostly that with the Ax branding, Apple is pretty much free to do with as they see fit, they are not marketed as ARM or x86 or anything really but Apple's branding of a CPU, what is behind that could change as Apple sees fit. Apple doesn't want to see themselves completely reliant on any specific vendor after the PPC (mostly IBM) issues.
My point with the GPU's was mostly a reminder of the PPC days, When us Mac users almost always had to have a custom "Mac version" of Graphics cards, which almost always carried a price premium and were not always up with the latest graphics chipsets. We do see more GPU circuitry all in the same silicon, but the integrated graphics are still not on par with the discrete, many machines (especially the Pro lines of Macs) will need Nvidia / ATI(AMD) graphics and we don't want a return of how that was handled in the PPC days.
I think the opposite is true. On the high end there is no ARM architecture chip remotely close to x86. The 18-core Xeon E5-2687W v3 does about 788 gigaflops, I/O bandwidth is 76 gigabytes/sec. If anyone wants to compare this to their ARM-powered device, they can download the linpack benchmark from the iTunes store.
If any ARM CPU has ever been produced that's capable of a sizeable fraction of the E5-2687W v3's performance, I've never seen it. I don't see ARM moving on a trajectory that would allow them to competitively intersect with the high end.
What about power efficiency on the low end? The thin/light notebook range is a more reasonable market niche ARM might conceivably move up into. However here Intel is aggressively moving downward with improved power-efficient x86 designs. E.g, consider this new Broadwell-powered Asus UX505. It is thinner and lighter than a MacBook Air, has a better screen and probably equal or longer battery life (though that's yet to be measured).
http://www.engadget.com/2014/09/03/asus-zenbook-ux305/
The next MacBook Air will probably use that CPU be better all around, and current ones are already more refined and polished. The point *not* notebook vs notebook, but that improved, efficient x86 CPUs are aggressively moving into the low end -- whether the device is made by Apple or somebody else. There are already $200 x86-powered Windows notebooks. How much more could an ARM CPU (which is just one component) lower the price? I don't see that ARM has much room to maneuver outside the mobile segment.
Yes but PC marketshare aside, HP also purchases a lot of high-end processors for their servers.
----------
Yes you are right - this reminds me of the transition to Intel except that this situation is completely different in almost everyway, but other that that, its very similar.
Looking at the comments here, quite many. Several people shout: if Apple goes away from Intel, Ill quit Apple. Its a free country, but most users dont really give rats ass if their computing device has an ARM or an x86 inside. All they care is that it works smoothly, comes in a nice box, doesnt overheat and has a good battery life. Look at iPhones and iPads, they work quite brilliantly for most common users.
Currently, iPad Air 2 compares very well to 2014 Macbook Air. Macbook is only 30% faster on a single thread, slightly slower on multi thread (yes iPad has one more core). And thats with a *current* Apple designed SoC that is being manufactured by the millions and that is designed to be way cooler than the i5.
So, it looks like Apple could build a way leaner and almost as performant machine with their current ARM chips. But why bother, transition takes time and investment on the toolchain. (although way less than with PPC->Intel transition, as iOS toolchain is *already* ARM+x86)
Its all about control and money. Building their own chips allows Apple to miniaturise further, use the transistor budget of the SoC to integrate more functionality on one chip. They can customise their own chip way easier than a chip from Intel.
The cost of manufacturing an A8X is estimated to be between $20$30, Intels list price for the i5-4260U is $315. Sure, Apple gets a discount but you bet its not $300. Basically Apple gets 10 comparable chips for price of one Intel.
Chips come with Apple R&D costs too, but theyre already doing it for the iPhone and iPad chips. Additional investments to design of Macbook chips would likely be small compared to the rise in margin alone.
That isn't them being scared, it's them knowing their programs don't work on ARM.
Yup, looks like a nice machine.
Yes, there are cheap x86 machines. But the nice ones and the cheap ones are *not* the same.
Switching Macbook Air from Intel to an A8X would mean approximately $200 to Apple's margin. And no, they're not giving it to the customer unless they're forced to. Given they can build a nicer (smaller, quieter, slightly more battery life, less hot) machine with a more efficient chip, they don't need to.
Making a *nice* laptop is not the point. That is why I specifically stated it's not about laptop vs laptop. Rather it's assessing how significant the ARM price advantage would be on the low end. The thing that makes it *nice* is not the CPU (which the customer cannot perceive) but the overall package and finish. The low end is the only place ARM can compete. If an x86-powered notebook is already at $200, what else does ARM have to offer?
If an entire x86 laptop can already be marketed at a retail price of $200, how could switching from x86 to A8X save $200 on a *future* laptop? We are talking about a future Broadwell-powered laptop, vs whatever ARM could produce in the 2015-2016 timeframe. You're saying on a MacBook Air with a $900 *retail* price that going to ARM would save Apple $200? One component change?
The 200$ laptop doesn't have processor that the MBA does. The 4360U has a market price of 315$, for example. I don't expect that the bulk rate is much lower than half of that, so we can surmise that Apple is spending ~155$ at least per processor. So it isn't the 200$ savings at that rate, but a decent 100$.
The 4260U in the bottom-end $899 MBA 11" retails for $315 per chip. You are assuming Apple's price is 1/2 of that or about $157.50. However their manufacturing cost for the laptop is not $899 but much less. So if their cost to manufacture the 11" MBA was, say, $700, then a $157.50 CPU would be 22% of the total cost. That sounds extremely high. If it *were* that high I could see the pursuing an ARM replacement, even if it cost customer dissatisfaction due to no Windows compatibility, poorer performance, etc.
However -- in 2010 one analyst investigated the 11" MBA and the CPU then was $80 on a $1000 machine: http://www.computerworld.com/articl...e-than-other-apple-laptops--says-analyst.html
If the x86 CPU is closer to $80, you obviously won't save $200 by going to ARM.
A) That's an analyst, I see no one reasoning beyond a guess.
B) Different chip than his guessed 80$, likely more expensive. I doubt they're giving Apple more than 50% off. My guess has just as much logic as his.
Do most consumers care about the CPU architecture?- No.
Do consumers care about stuff working?- Yes.
If you just use OSX for safari, itunes, Office, spotify, etc. Chances are you will be fine by any switch since these high profile programs will likely see a port or at least be ok with any dynamic recompiler (like with Rosetta).
If however you use windows software, niche software, or code which is optimised for X86 CPUs e.g. games, chances are you will suffer to at least some degree.
The issue for Apple will be whether the 'x86 or bust' crowd is small enough that there is minimal loss from any users switching to non-Apple machines.
Intel's price list puts it at $315 in quantities of 1000, Apple gets it cheeper for sure. I assumed that Apple would get it at $215 hence about $200 impact to margins.
But sure Apple may get better reduction in price, but a 75% reduction as your $80 figure would suggest, no way.
And that is exactly the problem. In a speculation a year ago the price of CPU in (then) MacBook Air was estimated just at 22%. Nice.
http://soumyajitsworld.blogspot.com/2014/01/apples-upcoming-12-inch-ipad-pro-will.html
Also if you look at the results Intel is generating on the x86 chips and quesstimate how many percentage of that is coming form Apple, one might think those billions could be used better, especially when Apple has very good chip design expertise in-house and very good track record on iOS devices.
Yes, transition is problematic and maybe it's not yet the time, but either Intel's prices come down or Apple will go for ARM in a low end Mac product(s) too. They *love* margins. I give them 3 years.
----------
I totally agree. I bet Apple has numbers on bootcamp popularity across their machines and perhaps even numbers on Virtualization app launches.
Two things would indicate positive things for ARM transition:
1. iOS devices sell fine and people are loving them without using Windows software.
2. Small laptops are not the main category for virtualisation and e.g. games.
Bootcamp, probably since IIRC that downloads software when you go to use it. VMs, unlikely, not unless the makers send Apple the details.
In regards to:
1. How do you know said people would be happy if their macs couldn't? For example, I don't expect my iPad to run windows apps because it's not meant for that but I expect my Mac to.
2. I was speaking generally as opposed to specific models, but again, how do you know people aren't? My brother recently won an MBA and I'm sure he'll run software on it which would suffer from a transition e.g. games (albeit light ones).
Without any research it is difficult for anyone to say whether the overall OSX market would be negatively affected by an ARM transition.
We are not going to go back to the days of transition (PowerPC to Intel, remember?) cause it is just bad.
Nobody would have their computer systems have different architectures and then worry about all the software compatibility and issues that will arise.
Can you imagine all the new threads about "Should I buy a new ARM MBA or stick to the Intel one?"
Nobody would have their computer systems have different architectures and then worry about all the software compatibility and issues that will arise.
Can you imagine all the new threads about "Should I buy a new ARM MBA or stick to the Intel one?"
1. At this moment iOS devices are bought mainly for content viewing, if they are used on a professional base, it's mostly as a digital clipboard (surveys, controls, content showing), but not as the work device. That is still a notebook/desktop.
This means that they don't have desktop software or an iOS (limited) version of desktop software, yet we won't have any problems with that because that is what we expect and we have our notebooks/desktops for the heavy work.
The majority of us will have problems if we have to start to virtualize software of Wintel office (MS Office) or Lintel (LibreOffice).
I'm not even talking about very specific software that now has a port to OSX/Intel or that can run because we can dual boot into Windows and Linux.
I'm talking about MS Office, which is the industry standard, and we can see that the iOS version is limited and LibreOffice, that doesn't even have an iOS version.
Please don't say that we all have to use iWork, because I'm also not seeing Apple bring out an iWork version for Wintel and Lintel.
If we remove all the Wintel and Lintel software that made its way to OSXintel, well it's going to become a barren wastland unless we start messing around with virtualizers. The majority of people aren't going to do that.
2. The MBA is a "small laptop" in size and power compared to MBPs, so what are they going to run if they can't run a virtualisation? Safari and iOS apps? An iPad can do that better and you can even buy keyboards for it.
Hypothetically if Apple was willing to split the MBA product line and accept that fragmentation, there are theoretical advantages to a laptop that can run iOS apps, or a iPad Pro that could run OS X.
E.g, everywhere I go I take my iPad and MBA. An iPad with a keyboard is pretty good but for some things a laptop is better. Both MBA and iPad are very thin and light but it would be great if somehow one device could do 80-90% of both.
Apple could definitely port all their software to OS X ARM, and could probably convince some other major developers to do it. How much or how many, I don't know. You wouldn't need all OS X apps, this would be a *new* category of device with sufficient advantages to justify it. They wouldn't be switching all MacBooks to ARM, it would be a new, additional category.
It would also require a fundamental change from Apple's "no touchscreen" policy for OS X. However that is doable and would not require re-writing all the desktop apps. With Windows 8 Microsoft showed how easy it is to run unmodified desktop apps on a touch screen. It's not great but it's workable.
Whatever Microsoft can do, Apple can do better. So hypothetically it would be like a smoother, better, more refined version of how Windows 8 runs desktop apps via a touch interface.
There is absolutely no way Apple can switch all Macs (or even all iMacs) to ARM -- the current and projected performance is just not there. However it is possible they could introduce a new ARM-powered device category that merges some laptop and tablet features. I tend to think they won't do this, but it's possible. It would also be a "shot across Intel's bow" to show they have other options in the OS X space.
Everybody is talking about PPC-to-Intel transition as something bad.
It was bad because it was forced.
It was forced because PPC was severely lagging behind Intel and could't be power efficient.
What if Intel was lagging and PPC was fast improving, power efficient and cost-effective? Many would support PPC architecture and it would Intel left in dust.
But we are witnessing the same for ARM. Mobile is future and ARM is best in mobile.
Ergo, mobile will be ARM, including some very thin, light, production oriented subnotebooks ala Chromebooks but faster and more efficient. Will it be OS X for ARM (why not? no one is forced to buy them, new market for developers), or Pro version of iOS, I don't know but writing is on the wall.
It was bad because it was forced.
It was forced because PPC was severely lagging behind Intel and could't be power efficient.
What if Intel was lagging and PPC was fast improving, power efficient and cost-effective? Many would support PPC architecture and it would Intel left in dust.
But we are witnessing the same for ARM. Mobile is future and ARM is best in mobile.
Ergo, mobile will be ARM, including some very thin, light, production oriented subnotebooks ala Chromebooks but faster and more efficient. Will it be OS X for ARM (why not? no one is forced to buy them, new market for developers), or Pro version of iOS, I don't know but writing is on the wall.
LOL, 18 cores? 76GB/s IO? What a joke! Here's a 48 core ARM server chips with 100's of GB/s of IO:
http://www.cavium.com/ThunderX_ARM_Processors.html (Press Release)
Just because you can't see it doesn't mean it doesn't exist. Open your eyes.
Of course intel can't do anything like this, because it would be huge, power hungry and cost $10,000+.
Intel's Xeon is also available in multi-socket configurations, I just didn't mention that. Also Xeon Phi has 60 cores per chip: http://ark.intel.com/products/71992/Intel-Xeon-Phi-Coprocessor-5110P-8GB-1_053-GHz-60-core
There's a very simple way to see whether ARM can compete on the high end: look at www.tpc.org. You won't see any ARM chips there. They cannot currently compete on the high end in raw performance, price performance, or energy performance.
This is why you don't see ARM in big compute servers or any of the top supercomputers: http://www.top500.org/lists/2014/11/
The upcoming Summit supercomputer will be by far the world's fastest and uses IBM Power9 and nVidia GPUs, not ARM. The current most powerful supercomputer on earth uses Intel Xeon. https://www.olcf.ornl.gov/summit/
Maybe that will change someday, and it's good ARM is trying to compete.
I had a 2010 MBA 11. On some benchmarks, the iPad Air 2 is faster.
The $200 notebooks do not compete with the MBA. They have very different types of customers. However, Apple has tons of customer data on product price sensitivity, and thus knows how much more MBA class product they would sell if the price were just 10% or 20% lower (but still at Apple's sky high margins). They probably also know the percentage of customers for that class of product that use VMs or bootcamp who they might potentially lose if they don't use Intel. Crunch the numbers, and ARM may lead to an revenue increase that more than offsets the loss.
As well as providing better battery life and more Apple custom system integration features ready on Apple's product release time-table.
----------
The ISA/architecture used for both the Power9 CPUs and Kepler GPU is closer to arm64 than to x86-64. So if someone were to actually design an arm64 chip for a mongo water-cooled machine...
...was designed for a big heat-sink, that the reason there's no comparison (yet?).
I just ran GeekBench 3.3 on my iPad Air vs 2013 MacBook Air -- the MacBook is about 2.5 times faster on multicore speed (6397 vs 2661). Of course it's pretty good a tablet can even produce 41% of a laptop performance.
The latest iPad Air 2 is significantly faster at about 4500, so it's about 70% as fast as an i7 MBA in that one test. That is impressive. Based on that an iPad Air 2 is about as fast CPU-wise as a 2012 i5-powered MBA.
On the low end ARM is getting faster but so is Intel. It will be interesting to revisit this later this year when Broadwell-powered MBAs are available, then again when Skylake-powered MBAs ship, probably in 2016.
However it appears on the lower end of the MBA lineup, ARM already has the necessary performance. So the issues of any hypothetical switch would be things like business case, software compatibility, CPU price/availability advantage, etc.
Yes, that is why I said (with emphasis) it wasn't about laptop vs laptop but a predictive exploration of what is possible on the lower end. If it's possible to sell for a profit an x86-powered $200 Windows laptop now, obviously there's no $200 CPU inside. The entry Mac Mini is just $499, so it's unlikely there's a $200 Intel chip inside there either.
If the Mac Mini can produce decent performance and sell for a profit at $499, how much money could an ARM-powered version save? Maybe it's worthwhile, but I don't see the basis for Apple somehow saving $200 per unit on the low end by switching to ARM.
Nobody knows anything about the Power9 CPU design yet. The Power8 is a very complex, power-hungry, out-of-order superscalar chip with a likely TDP in the 190 watt range. Beneath the covers Power8 is more like x86 than different. Like Power8, internally x86 is a wide-issue aggressively out-of-order multicore RISC machine.
Power8 is apparently modestly faster than the highest-end Xeon E5-2699 v3 on some benchmarks, but burns a lot more power and is more expensive. OTOH I don't see any published results from Power8 systems on www.tpc.org. If they could compete on performance or price/performance, it would be there.
The point of all this re ARM is both Power8 (and soon to be Power9), and high end Xeon (and next year's Skylake Xeon) are vastly beyond ARM in performance. This makes no difference on a lower-end Mac Mini or MBA, but for those who posted ARM is catching up in performance across the board, there is no indication of that on the high end, especially in the compute-intensive space. If ARM ever does achieve that, the results will be published and self evident. There is talk about using ARM blade servers in datacenters for applications that don't require a lot of compute performance. They might be able to carve out a niche there.