What are the chances of Apple getting in the x86 business?

[Exophase]

It's pretty ridiculous. A chip that's a tenth as fast as your laptop is going to replace it.

You say you think Apple should start making their own laptop chips, then you say how ridiculous it is for an ARM processor to be in laptops. You don't get it, either Apple can (and will) make a laptop class chip or they can't (or won't). It doesn't matter if it's x86 or ARM. Going x86 gives them virtually zero performance benefit over staying ARM. The only reason they'd want to make their own x86 processors is to maintain compatibility with MacOS X software. On the other hand, with ARM they'd maintain compatibility with iOS software, so it's a tradeoff.

 

[dastral]

By the time arm can scale to acceptable performance levels, efficiency is also out the window which is the entire point of why they'd even consider it in the first place. The ARM worship is funny at times. 30 watts in a laptop? How would that perform? Probably about as good as a 17 watt x86? And close to no compatibility? Where exactly is the benefit again?

Anand (or Intel) said that a uArch can usually be scaled to a factor of 10 (i'm too lazy to look for the post)
So a 2W ARM, could with time be scaled correctly to a 20W ARM.... The performance gain may not be 10x but you get the idea.

Now if Apple could design a 20W ARM Chip, performing 8 times better than a 2W A7...
Why would it waste money paying the Intel Tax and having zero control ?

We see today (more or less) a Tablet x86, performs like a Tablet Arm.
Since Intel was able to scale down, why can't Apple scale up ?

I don't see why in 2017-2020 a 20W Arm Chip shouldn't give the exact performance as a 20W x86 Chip.
And being Apple, which uArch will give you more control ? house designed A7+ or x86 ?

 

[Rakehellion]

You say you think Apple should start making their own laptop chips, then you say how ridiculous it is for an ARM processor to be in laptops. You don't get it, either Apple can (and will) make a laptop class chip or they can't (or won't). It doesn't matter if it's x86 or ARM. Going x86 gives them virtually zero performance benefit over staying ARM. The only reason they'd want to make their own x86 processors is to maintain compatibility with MacOS X software. On the other hand, with ARM they'd maintain compatibility with iOS software, so it's a tradeoff.

Macs run x86, iOS runs on ARM. There's no need to cross the two.

 

[Exophase]

Anand (or Intel) said that a uArch can usually be scaled to a factor of 10 (i'm too lazy to look for the post)
So a 2W ARM, could with time be scaled correctly to a 20W ARM.... The performance gain may not be 10x but you get the idea.

What Anand said (or was trying to say?) is that a uarch can have an order of magnitude dynamic range where it's a "good" fit and not overwhelmingly better served by another uarch. That would be a very very loose reflection of the product positioning done by CPU manufacturers.

And I'd say it needs more data points. Actually, I don't even really see anything so broad outside of Haswell (4 cores @ 8W to 4 cores at 80W). Maybe Silvermont. Certainly not seeing it with Jaguar (4 core definitely isn't a good fit at only 2.5W) or VIA's Nano (bottoms out at 6W with two cores, certainly doesn't go to 60W with that), or pretty much anything else I can think of.

The bottom line is we really don't know Cyclone is capable of clocking at. If it simply can't handle more than 1.5GHz then all the power budget in the world won't change this. There are limits to timings and you CAN design a processor around these limits - and there's a good indication that Apple has made at least some of these tradeoffs here. And even if it could cover a 10x range your numbers still assume viability starts at 2W, while it probably starts well below that.

But they could just as well design another processor targeting a different dynamic range. They have the resources. How close could they get to Intel doing this? Who knows, but so long as they're behind in manufacturing there's no way they'll match them. But this thread is nothing about that, instead it's about Apple switching to x86 for some reason, which does nothing for their perf or perf/W.

Macs run x86, iOS runs on ARM. There's no need to cross the two.

Doesn't sound like much of a defense to me, more "that's how it's done because that's how it's done." Apple could see some value in a new class of laptop that has a touchscreen and runs iOS software. There could be a market for that. Retaining software selection on MacOS X is important, but I'd argue that a large amount of that importance resides in a relatively small number of programs controlled by Apple or big third parties that can manage proper ports. I'm not so sure I'd say the same thing about iOS's library, where there's much more quantity vs quality going on (and not necessarily in a bad way, but people buying a bunch of $1 apps for more specialized purposes than one big $60 app)

 

[Rakehellion]

Anand (or Intel) said that a uArch can usually be scaled to a factor of 10 (i'm too lazy to look for the post)
So a 2W ARM, could with time be scaled correctly to a 20W ARM.... The performance gain may not be 10x but you get the idea.

Now if Apple could design a 20W ARM Chip, performing 8 times better than a 2W A7...
Why would it waste money paying the Intel Tax and having zero control ?

We see today (more or less) a Tablet x86, performs like a Tablet Arm.
Since Intel was able to scale down, why can't Apple scale up ?

I don't see why in 2017-2020 a 20W Arm Chip shouldn't give the exact performance as a 20W x86 Chip.
And being Apple, which uArch will give you more control ? house designed A7+ or x86 ?

Are there any high performance ARM chips out there?

I'm talking about Apple creating a chip to replace the i5, not the Atom.

 

[Rakehellion]

But they could just as well design another processor targeting a different dynamic range. They have the resources. How close could they get to Intel doing this? Who knows, but so long as they're behind in manufacturing there's no way they'll match them. But this thread is nothing about that, instead it's about Apple switching to x86 for some reason, which does nothing for their perf or perf/W.

Instead of them building a MacBook around the parts Intel gives them, they can build the chip, the device, and the OS at the same time, netting them huge productivity gains.

Instead of them having to choose between a mobile chip or a desktop chip and building around that, they can make something completely different with each iteration of the device

 

[dastral]

Are there any high performance ARM chips out there?
I'm talking about Apple creating a chip to replace the i5, not the Atom.

So am I and as far as i know there is no Arm able to replace i5.

Now ask yourself, Intel was able to scale a modern x86 to 2W. With 100Billion in bank, don't you think that Apple can't scale up A7 ?
Of course you can't just join two to make a quad core, overclock and pump 20W to reach i3-3000M performance level.

But do you doubt that Apple could do it if deemed "necessary" ? And something MUCH more desirable than making x86 ?

 

[Exophase]

Are there any high performance ARM chips out there?

I'm talking about Apple creating a chip to replace the i5, not the Atom.

This is basically the kind of argument someone makes when they don't really understand the root causes of things. "No one's ever made a fast ARM chip so it can't be done." But that's not true.

Until recently no one ever made an x86 phone chip, and you had lots of people flailing their hands in the air saying it'll never happen because x86 is always going to be too power hungry. You know how that turned out.

Instead of them building a MacBook around the parts Intel gives them, they can build the chip, the device, and the OS at the same time, netting them huge productivity gains.

Instead of them having to choose between a mobile chip or a desktop chip and building around that, they can make something completely different with each iteration of the device

Yes, having their own CPU instead of Intel's is good. Except the real advantage is cost per CPU. Waiting on Intel to release the next x86 whatever doesn't stall their device development very much nor their OS development at all.

And if they want their own CPU that means using ARM (or another ISA) because there's no way Intel will ever give them an x86 license.

 

[ShintaiDK]

Apple would also always be behind on the process node. Intel only needs to make an almost as good product. Then create it on a leading process node and still win flat out.

But look at Silvermont as an example. The ARM hype is very quickly dying. ARM went nowhere, and x86 is now beating ARM in ARM land. Its not long ago the same ARM supporters said that x86 could never reach ARMs performance/watt. Guess what happend.

 

[Rakehellion]

So am I and as far as i know there is no Arm able to replace i5.

Now ask yourself, Intel was able to scale a modern x86 to 2W. With 100Billion in bank, don't you think that Apple can't scale up A7 ?
Of course you can't just join two to make a quad core, overclock and pump 20W to reach i3-3000M performance level.

But do you doubt that Apple could do it if deemed "necessary" ? And something MUCH more desirable than making x86 ?

A better question is should they? And alienate their Mac audience with another architecture change.

 

[Rakehellion]

This is basically the kind of argument someone makes when they don't really understand the root causes of things. "No one's ever made a fast ARM chip so it can't be done." But that's not true.

I didn't say it can't be done, but if anyone wanted them, it would have been done by now.

Yes, having their own CPU instead of Intel's is good. Except the real advantage is cost per CPU. Waiting on Intel to release the next x86 whatever doesn't stall their device development very much nor their OS development at all.

It isn't about waiting, it's having to build another Intel computer to Intel specifications instead of an Apple computer.

 

[2is]

Anand (or Intel) said that a uArch can usually be scaled to a factor of 10 (i'm too lazy to look for the post)
So a 2W ARM, could with time be scaled correctly to a 20W ARM.... The performance gain may not be 10x but you get the idea.

Now if Apple could design a 20W ARM Chip, performing 8 times better than a 2W A7...
Why would it waste money paying the Intel Tax and having zero control ?

We see today (more or less) a Tablet x86, performs like a Tablet Arm.
Since Intel was able to scale down, why can't Apple scale up ?

I don't see why in 2017-2020 a 20W Arm Chip shouldn't give the exact performance as a 20W x86 Chip.
And being Apple, which uArch will give you more control ? house designed A7+ or x86 ?

What idea? That it will consume more power, perform worse and not have the compatibility? I'll ask again, where is the benefit?

 

[Exophase]

A better question is should they? And alienate their Mac audience with another architecture change.

Wow, this is really amazing. You go from one post where you argue the merits of Apple making their own processors, to another post responding to the only viable means of them making their own processors with "why should they?"

It's like you don't realize you're arguing with yourself.

What idea? That it will consume more power, perform worse and not have the compatibility? I'll ask again, where is the benefit?

The benefits are so obvious they shouldn't even need to be said.

1) Will cost less per CPU for Apple, because the margin on the x86 CPUs Apple buys for ultrabooks are not low, even for them. The margins at fabs like TSMC don't have luxury costs for specialty items like ULV Haswells.
2) Will give Apple more control in every aspect of the CPU design to better tailor for what's relevant to them - for example, giving exactly as much attention to GPU as they do or don't think is necessary, or integrating other peripherals that are specific to their devices. And not having to play the product segmentation game Intel subjects them to, where it's pretty much impossible to get certain combinations of features even though they're all on the die. Another example would be if they wanted to pair a lower perf/higher perf/W core with a higher perf/lower perf/W one (ala big.LITTLE) - normally this would be strange for them but if they do a new laptop chip they'd have Cyclone around to pair as an option.
3) Allows Apple to develop in a way that naturally aligns well with their product schedule instead of having to align with Intel's (or use older processors)


That doesn't mean that those are good enough reasons, but they're still obvious reasons.

 

[Rakehellion]

What idea? That it will consume more power, perform worse and not have the compatibility? I'll ask again, where is the benefit?

That sounds like the Power Mac G5.

 

[Rakehellion]

Wow, this is really amazing. You go from one post where you argue the merits of Apple making their own processors, to another post responding to the only viable means of them making their own processors with "why should they?"

It's like you don't realize you're arguing with yourself.

I'm pretty sure most people understand that Macs currently run x86 and not ARM.

 

[Ajay]

Apple would also always be behind on the process node. Intel only needs to make an almost as good product. Then create it on a leading process node and still win flat out.

But look at Silvermont as an example. The ARM hype is very quickly dying. ARM went nowhere, and x86 is now beating ARM in ARM land. Its not long ago the same ARM supporters said that x86 could never reach ARMs performance/watt. Guess what happend.

Not if Intel fabs Apple CPUs, which I suggested in an earlier post. Everything really depends on whether Intel can get a hold in the ultra-mobile market - particularly higher end tablets and phones where they can get margins sufficient to help sustain their premiere fabs. The problem is that the "Golden Handcuffs" of software are already locked in for iOS and Android under ARM CPUs. It's going to be a tough go for Intel. Maybe they'll make it, but they don't have the advantage of having a boatload of applications tied to their architecture like they did under the Intel-MS duopoly.

 

[Sheep221]

Why would they design own CPUs? If they can buy them for dirt cheap from Intel.

 

[shady28]

Wow, this is really amazing. You go from one post where you argue the merits of Apple making their own processors, to another post responding to the only viable means of them making their own processors with "why should they?"

It's like you don't realize you're arguing with yourself.



The benefits are so obvious they shouldn't even need to be said.

1) Will cost less per CPU for Apple, because the margin on the x86 CPUs Apple buys for ultrabooks are not low, even for them. The margins at fabs like TSMC don't have luxury costs for specialty items like ULV Haswells.
2) Will give Apple more control in every aspect of the CPU design to better tailor for what's relevant to them - for example, giving exactly as much attention to GPU as they do or don't think is necessary, or integrating other peripherals that are specific to their devices. And not having to play the product segmentation game Intel subjects them to, where it's pretty much impossible to get certain combinations of features even though they're all on the die. Another example would be if they wanted to pair a lower perf/higher perf/W core with a higher perf/lower perf/W one (ala big.LITTLE) - normally this would be strange for them but if they do a new laptop chip they'd have Cyclone around to pair as an option.
3) Allows Apple to develop in a way that naturally aligns well with their product schedule instead of having to align with Intel's (or use older processors)


That doesn't mean that those are good enough reasons, but they're still obvious reasons.

Exactly. One important item missed though, maybe a subset of #3.

OS X and iOS are fundamentally based on Darwin. Take a look at OS X and you see moves to make it more like iOS.

From a support and development standpoint, it would allow consolidation of development efforts to a single platform.

 

[2is]

Wow, this is really amazing. You go from one post where you argue the merits of Apple making their own processors, to another post responding to the only viable means of them making their own processors with "why should they?"

It's like you don't realize you're arguing with yourself.



The benefits are so obvious they shouldn't even need to be said.

1) Will cost less per CPU for Apple, because the margin on the x86 CPUs Apple buys for ultrabooks are not low, even for them. The margins at fabs like TSMC don't have luxury costs for specialty items like ULV Haswells.
2) Will give Apple more control in every aspect of the CPU design to better tailor for what's relevant to them - for example, giving exactly as much attention to GPU as they do or don't think is necessary, or integrating other peripherals that are specific to their devices. And not having to play the product segmentation game Intel subjects them to, where it's pretty much impossible to get certain combinations of features even though they're all on the die. Another example would be if they wanted to pair a lower perf/higher perf/W core with a higher perf/lower perf/W one (ala big.LITTLE) - normally this would be strange for them but if they do a new laptop chip they'd have Cyclone around to pair as an option.
3) Allows Apple to develop in a way that naturally aligns well with their product schedule instead of having to align with Intel's (or use older processors)


That doesn't mean that those are good enough reasons, but they're still obvious reasons.

Giving apple more control over processors vs getting them from Intel isn't a benefit. Apple does pretty well, but there's no way their design teams where are in their infancy is going to rival the experience Intel has.

I'm not even sure what you mean by #3.

As far as I see it, you've listed one benefit, cost. But at the expense of performance, power consumption and backward compatibility.

In essence, a cheaper product that isn't as good. Where's the surprise there?

 

[Khato]

Why would they design own CPUs? If they can buy them for dirt cheap from Intel.

This is a point that seems to often be forgotten. Apple's resources are far better spent developing the 'next big thing' that makes huge amounts of money rather than a custom part for current markets that could be serviced just as well by a commodity item. But that's part of why their stock is only worth about 2/3 of what it was a year ago - iterating to try and maintain the status quo instead of innovating and creating another new market for themselves.

 

[Exophase]

Giving apple more control over processors vs getting them from Intel isn't a benefit. Apple does pretty well, but there's no way their design teams where are in their infancy is going to rival the experience Intel has.

I gave a lot of specific examples of how it could be. Just because Intel has the best process and very good CPU design capability doesn't mean that their decisions of what to put on the SoC are going to be optimal for every possible device.

This is already pretty clear in the phone space, where for instance A7 has more GPU power than BayTrail-T, and I doubt the SoC peripherals all line up nicely. Not to mention, is less likely to have interfaces for things an Apple device won't ever use. Every little bit helps them make something more cost efficient and more effectively integrated with the rest of their hardware and software.

I'm not even sure what you mean by #3.

I mean that Apple's development will follow a natural schedule based on how long they need to finish making things. If they're also developing the processor the broader schedule of the devices it's used in will have some influence on how it's developed (and vice-versa). But the development of Intel processors won't fit their schedule at all. There are going to be points where the rest of their hardware is ready but they're waiting for the release of a new Intel processor, or have to use an old one.

As far as I see it, you've listed one benefit, cost. But at the expense of performance, power consumption and backward compatibility.

In essence, a cheaper product that isn't as good. Where's the surprise there?

Also iOS app compatibility.

 

[StrangerGuy]

I gave a lot of specific examples of how it could be. Just because Intel has the best process and very good CPU design capability doesn't mean that their decisions of what to put on the SoC are going to be optimal for every possible device.

This is already pretty clear in the phone space, where for instance A7 has more GPU power than BayTrail-T, and I doubt the SoC peripherals all line up nicely. Not to mention, is less likely to have interfaces for things an Apple device won't ever use. Every little bit helps them make something more cost efficient and more effectively integrated with the rest of their hardware and software.

Really, about time people here stop looking everything through "best process + performance" glasses. Apple obviously places much higher priority to controlling their destiny and be as unchained as possible to any outside entity.

 

[dastral]

Apple does pretty well, but there's no way their design teams where are in their infancy is going to rival the experience Intel has.
As far as I see it, you've listed one benefit, cost. But at the expense of performance, power consumption and backward compatibility.

Their design teams may be in their infancy, yet they have the best mobile product despite these drawbacks.
Sure A7 may not be as fast as a S800 or Little.Big but it's so custom tailored it may be called "perfect" for the job required.

As for "Performance/Power Consumption" with their stockpile of cash, do you think it's a problem ? A couple of years, A couple of Billions and that's it.

As for backward compatibility, we know they don't care, they already gave up PowerPC,
Not to mention once you go with x86 you are stuck with Intel's decisions/prices/bargaining.
(Anand even hinted that Intel's GPU focus is because of Apple)

Now if you just need a Fab, you have Samsung TMSC even GF or Intel...
Apple has always been about "Control" and we've seen this in reviews how they change LCD, SSD, etc.
Having your CPU destiny in someone else's hands is not desirable, specially since a CPU change is a little bit harder than going from LG to Samsung for LCD.

The day 20W Arm exist and perform almost like 20W x86, Apple will switch to ARM.

The iMac are at what 20% of their total product sales ?
http://www.businessinsider.com/chart-of-the-day-apples-gross-profit-per-product-2012-8

Imagine the synergies of going Full Arm & iOS9 in 5 years.

Now someone said earlier he sees Apple going full x86, and if you want the best performance, that's probably Intel...
But Mass Market Users don't care about performance (dual, quad, sli) they care about "Looks" and "The Experience" and Apple is a control freak.
Full "InHouse" ARM is the only solution that makes sense if you want Control.

 

[2is]

But we aren't talking about mobile. You're suggesting they're going to use their mobile parts to use in their non mobile products. The other part of what you said illustrates perfectly with why its nonsense. Why invest billions of dollars to move their non mobile products to ARM when they make up a mere 10% of their revenue? Especially considering by the time they've spent their billions, Intel will likely have strengthened their process advantage. There's literally no scenario you've painted where it makes more sense than it does nonsense.

 

[lamedude]

The day 20W Arm exist and perform almost like 20W x86, Apple will switch to ARM.

How much abuse can macolytes take? The PPC601 and Core2 were faster than the 68K and G5 they replaced so emulation wasn't painful. Will the 4% defect when their Steam games are not playable?