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

[Ajay]

I doubt Apple will leave x86 anytime soon, if ever. I have a feeling its gonna go the other way.

I think that it's more likely that Intel will eventually be building Apple's ARM CPUs. Are you thinking that Apple will abandon it's own design team (something the company is very proud of and, more importantly, is precisely responsive to Apple's needs) to go with Intel's small core x86 designs at some point in the future?

Intel is going to need to get as many high end (high margin) ARM design wins as possible to keep their volume and margins up to remain the premiere fab company that they are. Otherwise, TSMC could eventually threaten Intel's lead, unless the stagnation in big core x86 processor demand is simply a transitory event. It seems to be the direction that Intel is moving in, though their success to date may be limited by ARM developers not needing anything better than N-1 right now (and, perhaps, a desire to work with a pure fab which has no contradictory product motive).

 

[ShintaiDK]

I think that it's more likely that Intel will eventually be building Apple's ARM CPUs. Are you thinking that Apple will abandon it's own design team (something the company is very proud of and, more importantly, is precisely responsive to Apple's needs) to go with Intel's small core x86 designs at some point in the future?

Intel is going to need to get as many high end (high margin) ARM design wins as possible to keep their volume and margins up to remain the premiere fab company that they are. Otherwise, TSMC could eventually threaten Intel's lead, unless the stagnation in big core x86 processor demand is simply a transitory event. It seems to be the direction that Intel is moving in, though their success to date may be limited by ARM developers not needing anything better than N-1 right now (and, perhaps, a desire to work with a pure fab which has no contradictory product motive).

Remember what many seems to forget. And the same applies to Samsung for that matter. They dont earn much money making chips. They earn a bunch selling devices. So the device always got the priority over the chips, should there be any conflict.

 

[Fjodor2001]

Perhaps. But it makes their life much easier if AMD remains a little bit alive. Since AMD's market share costs Intel almost nothing, and gives them cover from the assorted anti-trust agencies around the world.

Well I'm just surprised that AMD would enter into an x86 licensing deal where the existence of the company is completely dependent on the goodwill of another company. Sounds very risky to me.

 

[ShintaiDK]

Well I'm just surprised that AMD would enter into an x86 licensing deal where the existence of the company is completely dependent on the goodwill of another company. Sounds very risky to me.

What do you call the WSA? Its not exactly something new for the company.

 

[Ajay]

Remember what many seems to forget. And the same applies to Samsung for that matter. They dont earn much money making chips. They earn a bunch selling devices. So the device always got the priority over the chips, should there be any conflict.

Apple, effectively, makes no money CPUs; they do make money, in part, on their CPU providing a positive marketing differential. It's worked for Apple on ultra mobile devices - it may eventually work for laptops and desktops (AIOs).

One thing I do see happening, and that's Apple wanting standard CPU architecture across their product lines. I won't say it can't be x86, I just feel like ARM is garnering allot of attention right now that will erode the quality of Intel's brand over the next ~5 years.

With x86's incremental performance increases, ARM stands a decent chance asymptotically approaching x86 performance while keeping power fairly constant for dual core CPUs and getting even closer on a Quad core design (which Apple is eschewing for now as it's not needed for tablets and phones ATM).

If Apple approached Intel with a Quad core ARM design** that they wanted manufactured on Intel's premiere node of the time, do you think Intel would say no?


P.S. Also, as a company and as their revenues grow (they 'only' are drawing ~ $1B/yr in revenues right now) that they should learn from the best and start a co-branding arrangement with their partners.

** I am assuming that if ARM has any brains, they will eventually be adding SMT designs (when the power performance curve for such is right). No doubt they'll learn the lesson that failed AMD and stick to something tried and true.

 

[ShintaiDK]

I think you should ask yourself, have all the low hanging fruits been grapped at ARM yet? Then you are back to incrementals there too.

 

[MisterMac]

I think people should realize LP vs HP processes - and wait til they see an ARM quadcore above 3 ghz - let alone above 4 nearing 5.

And one that keeps it's internal logics fed enough - something AMD can attest to always being easy.
(Or Intel during prescott).


There's a harsh reality from many core, high clock performance design (and node process) - to a lowpower ultra mobile design that's not meant to crush the heavnyess of a desktop.

 

[Ajay]

I think you should ask yourself, have all the low hanging fruits been grapped at ARM yet? Then you are back to incrementals there too.

I doubt it. For example, SMT is relatively low hanging fruit. But some of the low hanging fruit just got picked up by the A7 (2x GPRs, 2x FPU, 64 bit).

I think people should realize LP vs HP processes - and wait til they see an ARM quadcore above 3 ghz - let alone above 4 nearing 5.

And one that keeps it's internal logics fed enough - something AMD can attest to always being easy.
(Or Intel during prescott).


There's a harsh reality from many core, high clock performance design (and node process) - to a lowpower ultra mobile design that's not meant to crush the heavnyess of a desktop.

Well, ARM is already in some servers and I'm sure that they want to get into more. If so, I think we will see HP silicon in the future as opposed to LP only. And, of course it won't be easy, but thanks to Intel's success and AMD failure - there are a number of lessons that ARM doesn't have to learn the hard way.

If someone, like Apple, wanted a DTR CPU, they'd probably need to turn to Intel and be running on something more like an SHP (not that SHP will really exist in 5-10 years, heck, I'm not sure if it really exists even now).

 

[Rakehellion]

Numerically, the days of x86 have already passed. Considering the headroom still available to ARM (as demonstrated by Apple's A7), there is no reason for Apple to consider x86. With 64b ARMv8-A available to them, I'm guessing Apple has started compiling Mac OSX on the A7 waiting to see if it offers sufficient performance to potential unite the company's hardware under one architecture again. It may take ten years, but what does Apple have to lose?

The A7 is still 90% slower than a MacBook. These gains won't go on forever. Try transcoding video on an iPad.

 

[Rakehellion]

I think you should ask yourself, have all the low hanging fruits been grapped at ARM yet? Then you are back to incrementals there too.

Pretty much. The free performance gains are nearing an end.

 

[Exophase]

The A7 is still 90% slower than a MacBook. These gains won't go on forever. Try transcoding video on an iPad.

I'm kind of confused. You advocate Apple acquiring an x86 license, do you think that will suddenly open up new performance opportunities for them?

If Apple decided to start designing processors for the same dynamic range of power consumption Haswell is then they'd gain performance over A7 (most likely at the expense of worse perf/W). This isn't what A7 is trying to be.

 

[Ajay]

The A7 is still 90% slower than a MacBook. These gains won't go on forever. Try transcoding video on an iPad.

Pretty much. The free performance gains are nearing an end.

Faster transcoding requires special streaming instructions , FMA, faster/larger caches and faster DRAM (or by being offloaded on to a special function module). All of which are, relatively speaking, cheap to develop up to a point. So there is plenty of low hanging fruit.

As ARM usage grows, so will ARM Holding's design team and so will the size of the design teams for companies implemented leading edge ARM CPUs.

I think you are underestimating the potential ARM has by a fair margin. Please understand that I'm not expecting ARM to be capable of being used by something like an iMac for something like 10 years, maybe a little sooner. If you look at ARM, it's performance envelope is booming right now. Just look at how far thing have come from ARM7 (ARMv3 arch) which was still in use 5 years ago.

Lastly, so I don't get quoted out of context, this is speculation. Allot can and will change over the next decade - so if any of us turn out to be right, it will just be because of a lucky educated guess.

 

[Rakehellion]

I'm kind of confused. You advocate Apple acquiring an x86 license, do you think that will suddenly open up new performance opportunities for them?

If Apple decided to start designing processors for the same dynamic range of power consumption Haswell is then they'd gain performance over A7 (most likely at the expense of worse perf/W). This isn't what A7 is trying to be.

It would make more sense for them to develop x86 for their laptops, less so for their phones.

 

[Rakehellion]

I think you are underestimating the potential ARM has by a fair margin. Please understand that I'm not expecting ARM to be capable of being used by something like an iMac for something like 10 years, maybe a little sooner. If you look at ARM, it's performance envelope is booming right now. Just look at how far thing have come from ARM7 (ARMv3 arch) which was still in use 5 years ago.

Because no one cared about ARM until the iPhone, now everyone's suddenly pouring billions into it. We're pretty much filling in the gap of lost opportunity. In the next 10 years, x86 will get appreciably faster too.

Would a 30W ARM chip be on the level of a desktop processor?

 

[Blitzvogel]

ARM will have to pack on the muscle to catch x86 much like x86 has had to lose muscle to hit the same kind of efficiency and cost.

I think Apple switching to ARM in notebooks and would be less about ARM being able to match x86's performance than it would be about Apple being able to ditch Intel and the dues they pay them. Apple would be able to develop their own homogeneous processor and software ecosystem that works across all their devices. Once Apple develops a wide, 256 bit SIMD capable core that is readily available in dual, quad and beyond, Intel's gig is probably up.

 

[Ajay]

Because no one cared about ARM until the iPhone, now everyone's suddenly pouring billions into it. We're pretty much filling in the gap of lost opportunity. In the next 10 years, x86 will get appreciably faster too.

Would a 30W ARM chip be on the level of a desktop processor?

I don't think x86 is likely to become significant significantly faster overall. The low hanging has been picked on it. New instructions, for specific workloads, and an ever more powerful iGPU are de rigueur. You are right about everyone pouring billions into billions into ARM, but not because of lost opportunity, but because of present opportunity.

If Apple decides to create a market opportunity (something it can chose to do in it's own ecosystem, if conditions are right), then a higher power CPU could well be the way to meet there goals - 30 watts could be that power number; There is no way to know right now.

 

[2is]

I doubt it. For example, SMT is relatively low hanging fruit. But some of the low hanging fruit just got picked up by the A7 (2x GPRs, 2x FPU, 64 bit).




Well, ARM is already in some servers and I'm sure that they want to get into more. If so, I think we will see HP silicon in the future as opposed to LP only. And, of course it won't be easy, but thanks to Intel's success and AMD failure - there are a number of lessons that ARM doesn't have to learn the hard way.

If Intels success and AMDs failures are going to help ARM designing a chip, why isn't it helping AMD design a chip? If that's all you needed for a good chip design, there would be a whole lot more players doing it.

I think x86 is a better chance of competing with ARM than the other way around personally. Intel has made far more progress showing it has the ability to at least compete with ARM in its own game than ARM competing in the high performance segment.

 

[shady28]

If Intels success and AMDs failures are going to help ARM designing a chip, why isn't it helping AMD design a chip? If that's all you needed for a good chip design, there would be a whole lot more players doing it.

I think x86 is a better chance of competing with ARM than the other way around personally. Intel has made far more progress showing it has the ability to at least compete with ARM in its own game than ARM competing in the high performance segment.

Not sure where you're getting that thought. Arm powered tablets are largely credited with the decline of the desktop PC, which is to say Intel's 'turf'. Intel's last true bastion of strength is the server market.

 

[2is]

Not sure where you're getting that thought. Arm powered tablets are largely credited with the decline of the desktop PC, which is to say Intel's 'turf'. Intel's last true bastion of strength is the server market.

We aren't talking about the same thing. Im not talking about the declining PC market. I'm talking about low power/high efficiency processors which Intel has made deeper strides in than ARM has made into high power processors that Intel has already cornered. Considering Intel is still the largest CPU manufacturer, I'm not really sure where you're doom and gloom idea is coming from.

 

[SammichPG]

I doubt Apple will leave x86 anytime soon, if ever. I have a feeling its gonna go the other way.

Given your past Nostradamus moments I for one welcome our new ARM overlords.

 

[greenhawk]

Intel could give up a license to produce something they aren't able to.

they have no desire to. They are in the game to make money, giving away a license to something that could compeat against their actual main market is a very bad business practice.


As to Apple changing to ARM, they only just got to x86 after 10+ years of speculation of changing from RISC to the x86 instruction set.

They might change in the future, and if they are smart they might have some coding done in to the direction needed, but it will not see the light of day for, at a guess, 10 years.

 

[greenhawk]

So the licenses are not perpetual? I.e. Intel can also refuse to renew AMD's x86 license when it expires and effectively put AMD out of business?

I think they tried that back in the late 90's or there abouts. IIRC there was a court case and part of the ruling in AMD's favour was that if Intel went ahead, they would have a monopoly on the license and that was considered a big no no in the US courts.

 

[dastral]

I doubt Apple will leave x86 anytime soon, if ever. I have a feeling its gonna go the other way.

Well if i remember well Anand said (or had intel tell him) than an architecture can scale to a ratio 10.

If this is indeed true, with some work Apple could get a 20W A7 thus making the A7 series sufficient for every single portable product.
How about Desktops or Workstations with double triple quadruple 20W A7 ?
(With consoles having several low powered cores, Devs will slowly be forced into multicore too)

The chances of Apple getting into the x86 are LOWER THAN ZERO, it would make absolutely no sense.

Now as for "Apple Leaving x86"... well if someone can provide the product you want, why bother R&D to make it yourself ?
I mean, Apple doesn't make LCD or Flash so why did it end up making CPU ?
Maybe because no one had anything good enough for them ?

If Jobs was still here, Apple being Apple, i have no doubt they would leave x86 as soon as they could make 20W A7+.

2W for Phones, 4W for Tablets, 20W for Laptops, 3x20W for Workstations.... All on A7+

 

[2is]

No, saying chances are lower than zero is what doesn't make sense. 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?

 

[Rakehellion]

The ARM worship is funny at times.

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