[AppleInsider] Apple may abandon Intel for its Macs starting with post-Broadwell

[sm625]

Please justify the $4 billion a year in profit calculation. The calculation in your OP doesn't make sense because you forgot to figure out how much it would cost Apple to design the new Ax chip for the Mac, and then how much it would cost Apple to get those chips packaged and tested. The reason semiconductor companies tend to have high gross profit margins is that R&D costs are very high and need to be paid for.

Also, I seriously doubt Apple is paying more than $150 for any MacBook chip.

$4 billion is obviously a very rough ballpark estimate. I figure they spend an average of $150 per cpu. Keep in mind that i7 mobile cpus that turbo up to 3.7GHz have tray prices of almost $400. So the average cpu price of $150 may be on the low side. It could easily be $200. But let's use $150.

$150 times 16 million units annually = $2.4 Billion.

By integrating the SSD controller they will save at least $10 per unit on NAND costs, as well as $5 per unit on the DDR that most SSDs use, as well as $10 for the controller itself. They will also save on other components, such as the interconnect between the SSD and the mainboard. At any rate I estimate $25 per unit saved by integrating the NAND controller.

$25 times 16 million units annually = $0.4 Billion.

Also, they would integrate the PCH to save another $25.

$25 times 16 million units annually = $0.4 Billion.

So we're already up to $3.2 billion. Assuming their sales growth holds steady, that $3.2 billion (according to 2013 sales numbers) becomes about $3.6-$4.2 billion for 2016, the first year it is even remotely possible for them to do this. So I think my original ballpark estimate of $4 billion was pretty accurate.

Sure it might cost them a few billion to get rolling, but there is no question that it would be an immensely profitable move. And they're going to do it sooner or later. It isnt just about the performance of the intel core, its about integration. Intel is just too far behind in that regard to be useful to apple for much longer. Hybrid memory is coming. Apple will jump all over that and intel / microsoft will be way behind the curve in implementing it. I been saying for almost 3 years that they should design a split NAND/DRAM bus so that when hybrid memory does become available, the architecture will already support it. But there is still no sign of that. Apple will not wait around for it. I predict that Apple will be the first company that puts all their memory and storage onto a single package. Regardless of the form it takes, it is all going to have to talk over the same bus.

 

[dahorns]

$4 billion is obviously a very rough ballpark estimate. I figure they spend an average of $150 per cpu. Keep in mind that i7 mobile cpus that turbo up to 3.7GHz have tray prices of almost $400. So the average cpu price of $150 may be on the low side. It could easily be $200. But let's use $150.

$150 times 16 million units annually = $2.4 Billion.

By integrating the SSD controller they will save at least $10 per unit on NAND costs, as well as $5 per unit on the DDR that most SSDs use, as well as $10 for the controller itself. They will also save on other components, such as the interconnect between the SSD and the mainboard. At any rate I estimate $25 per unit saved by integrating the NAND controller.

$25 times 16 million units annually = $0.4 Billion.

Also, they would integrate the PCH to save another $25.

$25 times 16 million units annually = $0.4 Billion.

So we're already up to $3.2 billion. Assuming their sales growth holds steady, that $3.2 billion (according to 2013 sales numbers) becomes about $3.6-$4.2 billion for 2016, the first year it is even remotely possible for them to do this. So I think my original ballpark estimate of $4 billion was pretty accurate.

I'm not sure you understand what "profit" means. Are you really suggesting that Apple would have no additional operating costs AND no additional sales costs (not to mention R&D/fixed costs)? Are they simply wishing these chips into existence?

 

[kimmel]

That is not an insurmountable challenge, but it's a challenge you don't need to overcome of you use ARM. With an ARM device you will have a fraction of the power draw and heat.

So all those overheating Snapdragon 810's don't have to worry about heat because... ARM?

 

[Roland00Address]

sm625 you do know that Intel's average selling price for their laptop chips is about $88 right? (Once you average out all the celerons with the i7s)

http://www.wsj.com/articles/SB10001424052702303417104579541791542046598

Yes apple buys better cpus than the median but they also have market share to force intel to give them better prices. I doubt they are paying $150 for the i5s, and most of the laptops intel sells are not the i7s but instead the i5s. I just bring this up for it shows that apple is not really losing billions with buying intel chips.

Do not forget all the transition costs and possibly lost revenue when the arm chips may have software problems. You are now interefering with "it just works" and are tanking your brand image and mindshare is just as important for apple. Mindshare is what convinces people to buy an apple just because without even considering other options, once they start considering apple has already lost.

 

[ninaholic37]

Let's assume the next MBA will be Broadwell based. Then what's held Apple back from releasing it already?

Some theories:

1. Apple was not satisfied with the Broadwell Y Core M 5Y70 performance, so they are waiting for Intel to provide a better SKU that satisfies their requirements. E.g. higher clocks at same TDP, maybe the 5Y71 model or even better.

2. Apple is expecting the new MBA to sell in very high volumes. So they have to wait for Intel to build up a sufficiently large stock of 5Y71 chips before launch. Normally Intel has already built up sufficient stock of CPUs when they launch a new CPU. But given the 14 nm yield issues, this may have taken longer than expected. So maybe they released small quantities of Core M in late 2014 (not sufficient volumes for Apple though) to still be able to claim a 2014 Broadwell release, and then make the "actual" launch in 2015?

What do you guys think is most likely, and are there any alternative theories?

Don't they always use the U series for the Air (U for Ultrabooks)? I don't see why they'd change that now, so my guess is the next one will either be Broadwell-U or Skylake-U.

People don't seem to understand Core M. Core M is name without any content. It's like saying "my laptop has i7!!1".

Core M is the brand for all fanless Core-Y SKUs from Broadwell on. Skylake-Y will also be called Core M and that's why you should call it BDW-Y or SKL-Y, not Core M.

Doesn't Core M have a much smaller die size than all the other Broadwells? I thought I read that in the spec sheet. It seems like the architecture might be different/squashed down enough that it warrants the new name.

 

[elemein]

On a related/semi-related note, I have a question about ARM CPU architecture.

I was doing some research today, and, as someone who is used to reading block diagrams from places like RWT, I noticed that the block diagrams for the Apple and Qualcomm architectures lack instruction fetch units.

Case in point:

http://www.phonearena.com/news/Sams...s-6-Snapdragon-805-performance-review_id63789

These are the only block diagrams I can find of the Apple archs, and while there are other diagrams of the A9 ( http://www.arm.com/files/pdf/armcortexa-9processors.pdf ), it's very, very jarring to go from diagrams such as: http://www.realworldtech.com/wp-content/uploads/2012/10/haswell-5.png?71da3d

to something completely missing out on a vital part of the CPU.

Can anyone explain what's going on here? Does Apple and Qualcomm have the fetch units built into the decoders or was info not released about them or is it just how the person who put the graph together does things or... I'm very confused as this is my best method of understanding architectures; reading the block diagrams.

 

[tential]

So all those overheating Snapdragon 810's don't have to worry about heat because... ARM?

Arm processors physically can't heat up. I thought you'd know that...

 

[CHADBOGA]

I am already staking my name on this forum. I'm not sure, I don't gamble with money as a principle but other stuff I'm fine with.

I expect to be roundly ridiculed if this thing comes out with Intel, but if it turns out to be ARM I would hope to gain some modicum of respect.

Nothing will change.

No one takes you seriously now, so you won't have lost anything when they continue to not take you seriously.

 

[Eug]

Don't they always use the U series for the Air (U for Ultrabooks)? I don't see why they'd change that now, so my guess is the next one will either be Broadwell-U or Skylake-U.

The rumour has the new MacBook being much lighter and possibly even fanless, hence the Y-series assumptions.

Personally I'd be happy with a MacBook Air size and weight, ie. U-series, but with Retina, especially if it meant it'd keep more ports than that new Retina MacBook rumour and a proper clickable trackpad.

But like I said, I'm likely waiting for Skylake next year, for various reasons I've already mentioned in the other thread (and here).

 

[Roland00Address]

Lets be honest, the macbook airs does not need to be retina

What it needs is get rid of much of the bezels. Put an ips screen on there with good brightness and low glare and make the resolution 1440x900 on the 11" and 1920x1200 on the 13" if you keep the screens the same size or 1920x1200 if you do anything 12.5" or bigger, remove the fan and keep everything else the same. Then the next thing you need to do is reduce the price while still allowing apple to make money

Sure I want retina if it can be done with such a little opportunity cost, but I do not want retina on the cheap model if its means a massive loss of battery life or a massive performance hit. The whole point of ultrabooks is not the perfect screens (we have desktops for that) but for portability while still being a full fledge computer and not a tablet. The current macbook airs are near perfect if you like OS X the biggest problem is the viewing angles and the wasted bezels.

So if they took the 11” MacBook air chassis and put a 13.3” screen 16x10 screen you would have a screen with a width of 11.28” and a 11.8” chassis giving you 0.26” of an inch on each side
The height of the 13.3” screen would be 7.05” and the height of the chassis would be 7.56” giving you 0.255” on each side

---

If they took the 13” MacBook air chassis and put a 14.5” 16x10 screen in the 13” chasis you would have a screen with width 12.3” and a chassis of 12.8 inches leaving a bezel of .25”. The height of the screen will be 7.68” the height of the chassis will be 8.94 inches leaving a bezel of .63” inches

Please just do this and give us a great ultrabook for a great price. Leave the retina upgrade for next year when we will have more gpu power to actually run those screens (plus it will be a good reason to upgrade).

 

[RampantAndroid]

Lets be honest, the macbook airs does not need to be retina

Which is why everyone wanted retina air.

 

[tynopik]

With an ARM device you will have a fraction of the power draw and heat.

And a fraction of the performance.

 

[bullzz]

MS just reported record revenue on Surface. There are rumors that say MS is ditching windows on ARM. If they go all out x86, I think they will eat a chunk out of ipad/mac sales

 

[Nothingness]

On a related/semi-related note, I have a question about ARM CPU architecture.

I was doing some research today, and, as someone who is used to reading block diagrams from places like RWT, I noticed that the block diagrams for the Apple and Qualcomm architectures lack instruction fetch units.

Case in point:

http://www.phonearena.com/news/Sams...s-6-Snapdragon-805-performance-review_id63789

These are the only block diagrams I can find of the Apple archs, and while there are other diagrams of the A9 ( http://www.arm.com/files/pdf/armcortexa-9processors.pdf ), it's very, very jarring to go from diagrams such as: http://www.realworldtech.com/wp-content/uploads/2012/10/haswell-5.png?71da3d

to something completely missing out on a vital part of the CPU.

Can anyone explain what's going on here? Does Apple and Qualcomm have the fetch units built into the decoders or was info not released about them or is it just how the person who put the graph together does things or... I'm very confused as this is my best method of understanding architectures; reading the block diagrams.

All CPUs have instruction fetch unit or they couldn't load instructions from memory to feed the decoders :biggrin:

Joke aside these diagrams are just missing a lot of information, all of the CPU's mentioned have an instruction cache, a branch prediction unit and most likely a form of hardware prefetch.

 

[Nothingness]

There are rumors that say MS is ditching windows on ARM. If they go all out x86, I think they will eat a chunk out of ipad/mac sales

And why going full x86 would change their market share against ipad/mac given that RT tablets were a total failure and given that x86 Surface already is a success?

 

[bullzz]

@Nothingness - i think u answered your own question. RT was a failure and if surface mini is x86, I expect people to really buy it. if a company like asus can make a decent product like Transformer T100, I expect MS to make an amazing one. with windows 10 and skylake around the corner, it can only get better

 

[OBLAMA2009]

man, macs without intel would be a total joke, that would totally ruin their products

 

[TreVader]

Nothing will change.



No one takes you seriously now, so you won't have lost anything when they continue to not take you seriously.

Lol says the person with no ideas about anything whatsoever. Are you upset that nobody takes you seriously chad?

Try to read what other people post and then copy it, ok? That way instead of reporting your personal attacks and directing you on how to properly post on a forum (something anybody here for so long should know by now) I can instead actually have conversations with people who have their own ideas. But I get it, you are a slow learner. So I will be patient with you and help you understand why Intel is such a terrible processor company in 2015, ok?


Personal insults are not allowed
Markfw900


Ok

 

[TreVader]

And a fraction of the performance.

If by "fraction" you mean 9/10ths the performance at 1/10th the peak power draw then yeah it's a "fraction" of the performance.

I wonder what apple will use to replace its iMac processors? It's easy to replace Broadwell-U because A8X is already faster than Broadwell in multicore and close in single core at a fraction of the power draw.

It will be much more difficult to replace the quad core desktop stuff.

 

[TreVader]

man, macs without intel would be a total joke, that would totally ruin their products

This is the general view of people who shouldn't be buying macs.


If you are buying macs and running windows on them, you are doing it wrong. Macs running windows is a design FLAW, not a feature. Apple would gladly lose every customer like you to have a few like me.


Good riddance!

 

[coercitiv]

It's easy to replace Broadwell-U because A8X is already faster than Broadwell in multicore and close in single core at a fraction of the power draw.

Quoted as memorabilia.

 

[Sweepr]

If by "fraction" you mean 9/10ths the performance at 1/10th the peak power draw then yeah it's a "fraction" of the performance.

I wonder what apple will use to replace its iMac processors? It's easy to replace Broadwell-U because A8X is already faster than Broadwell in multicore and close in single core at a fraction of the power draw.

It will be much more difficult to replace the quad core desktop stuff.

You do realize you can't simply take Geekbench scores and start making conclusions about performance on different architectures right? Otherwise Samsung's new Exynos 7420 smartphone SoC would be faster than a 15W Core i5 5200U (Broadwell-U) in MT tasks, and it most certainly isn't.

Also where does the 1/10 peak power draw statement comes from? Please enlighten us, I'm curious.

 

[VirtualLarry]

This is the general view of people who shouldn't be buying macs.


If you are buying macs and running windows on them, you are doing it wrong. Macs running windows is a design FLAW, not a feature. Apple would gladly lose every customer like you to have a few like me.


Good riddance!

Why are you so against consumer choice? From the same person that considers an SD card slot on a phone a "flaw", because it "ruins the esthetics of the perfect bezel". You're a flaming Apple fanboy, we get it.

 

[TreVader]

Why are you so against consumer choice? From the same person that considers an SD card slot on a phone a "flaw", because it "ruins the esthetics of the perfect bezel". You're a flaming Apple fanboy, we get it.

What good is consumer choice, if the choice is between a great company and a terrible one? There will always be a subset of ill-informed people who will choose the terrible company, and never stop supporting the terrible company.

That's what wintel is, it's a duo of terrible companies. They produce awful products that nobody enjoys using, and people always say "well without windows I couldn't game" or "without windows I couldn't use my favorite software"..... but there is no real love for windows itself. I almost never see people say Windows is a "better" OS than OSX, they just like that their favorite software runs on it.


And I like Apple, so what. I don't come in here calling you and others here "flaming intel fanboys". But It's OK, because I know you are just scared. Scared that your favorite awful company is about to go under! IT'S OK! There is life after Wintel()

 

[witeken]

Doesn't Core M have a much smaller die size than all the other Broadwells? I thought I read that in the spec sheet. It seems like the architecture might be different/squashed down enough that it warrants the new name.

BDW-Y is the same die as the GT2 BDW-Us, as far as I know.