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

[TreVader]

Don't underestimate the impact of losing a halo, and high-margin vendor. No doubt some still believe Intel still has a chance of winning Apple designs, not just fabbing for their chips. IMO, Apple abandoning Intel will kill ALL hopes, even foundry.

How much emotional impact it'll have on other vendors? That'll translate into financial impact sooner or later.

This is another good point. Even if Apple doesn't drop intel completely, and just makes their next macbook air with ARM, what does that say about Intel? It at least says that Apple no longer has confidence in their low power mobile processors.


I mentioned it before but without Apple, what is anyone's experience with Intel?If Apple drops x86 in favor of ARM then Intel and Microsoft are forever more hitched. Does that look good for Intel?


In my mind, Apple choosing intel legitimized their whole Core 2 Duo line. Before that, intel was just a poor competitor with Athlon FX. When Apple drops intel it is gonna hurt.


If I were intel, I would be concentrating on preventing ARM from getting ANY ground in the server market. Contra Revenue x 1000. That's what intel is good at anyways, business. Not engineering. ARM clearly can compete for Apple, what's next?

 

[tential]

Trevador, I think your argument gets mixed up because people are assuming you mean Apple will ditch Intel completely.

If you're going for saying that Apple may start to enter using it's models for a higher up Macbook Air Line (12 inch macbook I saw you state, which I had already theorized with earlier in the thread), I tend to agree that IF Apple were going to enter the market, they'd start with a Macbook Air, 12 inch convertible. Meaning it's an Ipad Air, that Docs into a keyboard to become a Macbook. Not sure what they'd call it, but I believe that'd be the best bet. It'd allow them to draw from Ipad's great eco system, but still give a mouse/keyboard. Perfect for Word/Excel, but still great for consumption.

If that device sells well, you can best believe Apple will use that in it's other Macbook Airs.

I don't see Apple using it in a Macbook Pro/Desktop line though as of yet. Raw performance still matters a lot there.

The largest issue for Apple is the ecosystem. That's why I say it HAS to be a convertible. People need to realize that they're still using a "tablet" because when all of their programs don't work they'll be very upset.

 

[tential]

Sorry, but LOL if you think the Core 2 Duo line was legitimized by Apple. Core 2 Duo was legitimized by it's INSANE performance. Every single person remembers reading Conroe's benchmarks that first time and going "Oh wow.... this just got real".

 

[TreVader]

Trevador, I think your argument gets mixed up because people are assuming you mean Apple will ditch Intel completely.

If you're going for saying that Apple may start to enter using it's models for a higher up Macbook Air Line (12 inch macbook I saw you state, which I had already theorized with earlier in the thread), I tend to agree that IF Apple were going to enter the market, they'd start with a Macbook Air, 12 inch convertible. Meaning it's an Ipad Air, that Docs into a keyboard to become a Macbook. Not sure what they'd call it, but I believe that'd be the best bet. It'd allow them to draw from Ipad's great eco system, but still give a mouse/keyboard. Perfect for Word/Excel, but still great for consumption.

If that device sells well, you can best believe Apple will use that in it's other Macbook Airs.

I don't see Apple using it in a Macbook Pro/Desktop line though as of yet. Raw performance still matters a lot there.

The largest issue for Apple is the ecosystem. That's why I say it HAS to be a convertible. People need to realize that they're still using a "tablet" because when all of their programs don't work they'll be very upset.

I think they will drop Intel for the 12" rMBA, and over the next few years phase Intel out of mobile products. Unless Intel comes out with something to shake things up.


But for now, it's my opinion that only the rMBA will be dropped.

 

[tential]

I think they will drop Intel for the 12" rMBA, and over the next few years phase Intel out of mobile products. Unless Intel comes out with something to shake things up.

But for now, it's my opinion that only the rMBA will be dropped.

While this isn't my "opinion" I believe this is a possibility but ONLY if it's a convertible/dockable tablet design.

If Apple does make significant strides in their ARM processor development then of COURSE they'd phase intel out. But that's far down the road and Intel isn't the type of company to just stand still and take abuse. If you push them, they'll show you their hand. Last time it was Core 2 Duo/Conroe. We'll see what they have in store for us this time around.

 

[RampantAndroid]

This is another good point. Even if Apple doesn't drop intel completely, and just makes their next macbook air with ARM, what does that say about Intel? It at least says that Apple no longer has confidence in their low power mobile processors.

Jeez, stop with the silly conclusions. *IF* they ditched Intel for another 3rd party CPU, you could say this. Dropping Intel for their own ARM CPU then this is NOT a sound conclusion - it could simply be them saying their CPU is "good enough" and offers a much lower cost to them, since they own the A series CPUs. It's the same thing they did with PA semi.

 

[Lepton87]

Except a 5960x is generally slower in most workloads. The only workloads where a 5960x does better is in highly parallel workloads. Which makes sense because its not a desktop part and not designed to be a desktop part. Its an extremely cut down server part that still contains all the core functionality needed to make a 15-18 core part work, which is why it has significantly higher latency than a Devil's Canyon part and hence generally slower performance on workloads that matter on the desktop.

You are so misinformed, because it is slower it ends up on top in most benchmarks and mostly by a good margin. And it's so cut down that it only has 8 cores out of 8 cores the silicon has, so cut down!!

ps. I don't know if you are just trolling or ...but HW-E has 3 dice, 8 cores, 12 cores and 18 cores. Starting from ~355mm2 to ~680mm2

No you can't say the same thing. The design used for the -E parts aren't at all designed for the -E part. They are designed for the -EP and -EX parts. AKA, they are designed as 15-18 core parts (for the last 2 generation core counts), period. And, YES, HSW-E is a chopped and harvested server part. Every EE part since the initial release of EE parts has been a harvested server part.

Never mind, you are just trolling, you were already told that 5960X is NOT a harvested die. bye.... Also the only harvested EE parts were SB-E and IVY-E.

 

[krumme]

1B is basically an educated wag. Its unlikely you are doing a competitive laptop/desktop processor or less than a Billionish these days. You are likely looking at a 4-5 year development cycle from inception to ship. You'll probably looking at an average of 200+ people. You'll have various licensing costs as well. Etc. Its somewhere in the range of 500m-1b in NRE. A lot of its going to depend on fully loaded costs per person. Plus you'll have all the fab related NRE costs.

I dont get it and nobody is talking desktop or highend laptops here.

The zakate hyderabad team was about 80 people. And they even had to assemble the IP from cpu, gpu whatever.

Using a h2 2015 ipad a9x how hard can that be? Do you even need different binning or packaging? What licencing is relevant?

There is tons of cost else where, software ...and if they get x86 cheap then who cares. But i simply cant see where 1b goes for slapping a a9x in a slightly bigger form factor? Its just less throttling.

 

[CHADBOGA]

I think they will drop Intel for the 12" rMBA, and over the next few years phase Intel out of mobile products. Unless Intel comes out with something to shake things up.


But for now, it's my opinion that only the rMBA will be dropped.

Just to be clear here, you mean an Apple device that has as its OS, OSX 1*.*, rather than iOS *.*?

 

[TreVader]

Just to be clear here, you mean an Apple device that has as its OS, OSX 1*.*, rather than iOS *.*?

I'm not sure, but not iOS in its current form. Either OSX or some sort of iOS/OSX hybrid operating system.


It will have to replace a MacBook Air functionally.

 

[Maximilian]

In my mind, Apple choosing intel legitimized their whole Core 2 Duo line. Before that, intel was just a poor competitor with Athlon FX. When Apple drops intel it is gonna hurt.

I loled :awe:

 

[ninaholic37]

^^ The Pentium M (2004-2005) was pretty popular too, most of the laptops I've seen from this era used it, even ones in the $2000-3500 category. It's interesting that Apple was still in it's iBook/PowerBook phase at this time, it probably took Apple some time to switch over after they realized that Intel could actually make mobile chips that weren't just slowed down/gimped versions of their desktop chips, and after Speedstep got off the ground and it's bugs in the P4 versions were worked out. It seems that Apple originally wanted to switch to Pentium M also, but by then a new version was coming out named Yohan, which was later renamed "Enhanced Pentium M" then "Core Duo" using "Enhanced Pentium M architecture" for marketing purposes. I believe this is the era when Windows laptops were ruined with 16:9 screens and Vista. Pentium M FTW

 

[Owls]

I say good riddance, it would restore the differentiation apple had on the pc space when they sported ibm cpus. To buy a mac that has idebtical hardware to what you can get in a pc its kinf of absurd, speciallu for the Premium they ask for their products

So apple is going to stop doing the thing that started bringing them marketshare?

 

[jhu]

Um, that's because they are running different code sets as should be obvious. Given the same browser however, its a reasonable approximation of what the majority of phone applications actually do.



GCC
xalancbmk
astar
sjeng
gobmk
mcf
bzip2
perlbench
sphinx3

I'm glad I'm not the only one who uses GCC on his phone.

 

[imported_ats]

You are so misinformed, because it is slower it ends up on top in most benchmarks and mostly by a good margin. And it's so cut down that it only has 8 cores out of 8 cores the silicon has, so cut down!!

If ends up slower in the majority of benchmarks: games and applications. The benchmarks that it is faster in are bascially of no concern to the majority of people who actually buy an EE.

And yes, it is cut down, the design is designed around 15-18 core designs. There is significant overhead in the 8 core chop because its a chop of a 15-18 core design. If the design was actually centered on a high end desktop design the various latencies would be significantly less.

ps. I don't know if you are just trolling or ...but HW-E has 3 dice, 8 cores, 12 cores and 18 cores. Starting from ~355mm2 to ~680mm2

I'd wager that I know more about the actual design than anyone on this board...

Never mind, you are just trolling, you were already told that 5960X is NOT a harvested die. bye.... Also the only harvested EE parts were SB-E and IVY-E.

It very much is a harvested die, its a chop of a larger design. IVY-E btw was also a chop. As a chop, there is no optimization in the design for the lower core count parts. The actual design centers around the EP/EX functionality. The E dies are really a byproduct of the cheapest method to get a product in that given space. Hence the reason the E dies reuse the same package set from the EP/EX even though a significant number of the pads are basically blanks.

 

[imported_ats]

I dont get it and nobody is talking desktop or highend laptops here.

If they are going to do a switch over, they are going to have to do it in the majority of the Mac product line or its going to fail. Even the MacBook Air requires absolute performance levels significantly higher than the iPAD.

The zakate hyderabad team was about 80 people. And they even had to assemble the IP from cpu, gpu whatever.

The Zakate design is merely an exercise in integration at best. That required 80 people to make and didn't include anything new. The amount of work required was basically less than required for a raw shrink. If you think that's at all comparable to what is required for an entirely new design, you are underestimating the amount of work.

Using a h2 2015 ipad a9x how hard can that be? Do you even need different binning or packaging? What licencing is relevant?

It is highly unlikely that the a9x has performance goals/targets appropriate even for the low end of the laptop domain.

There is tons of cost else where, software ...and if they get x86 cheap then who cares. But i simply cant see where 1b goes for slapping a a9x in a slightly bigger form factor? Its just less throttling.

You are assuming a rather significant performance downgrade is acceptable to the user base. I would argue that it isn't. The user base would be up in arms unless the new product delivers better performance than the older product. Esp when you factor in a complete software transition.

 

[imported_ats]

Has iOS Safari been ported to other OS? That'd be a prerequisite to compare Apple CPU against other since as far as I know Apple doesn't allow JIT on its OS, so other browsers are limited in what they can do.

Depends on what you consider iOS Safari to be. IIRC, it is largely based on webkit, but has significant optimizations for the platform. It is likely that iOS safari isn't comparable to OSX Safari even.

GCC -> compiler
xalancbmk -> XML
astar -> path finding
sjeng -> Chess
gobmk -> Go
mcf -> Simplex
bzip2 -> compression
perlbench -> PERL scripts
sphinx3 -> voice recognition

Odd choice of benchmarks for sure: at least two of astar, sjeng and gobmk should go away or you are over-representing games. Simplex, well no comment... PERL, gcc, almost no one use those unlkess they are developers.

The 3 gaming related benchmarks are actually fairly different and represent code paths an algorithms that are also used outside of games.

gcc has been found to actually represent the types of code flow found in a significant amount of actual software. PERL is an interpretive scripting language which are actually somewhat common even in the mobile space.

I'm surprised you didn't include h264 which is likely more representative than mcf, perl or gcc, even though few people do software encoding.

The problem is that no one is going to do software based video enc/dec on a mobile platform. As such, it is rather unrepresentative.

BTW most of these SPEC benchmarks put *very* little pressure on instruction caches which is an issue. As an example mcf is less than 3K lines of code. And none of them does runtime code generation which has become very important (though that'd be partly covered by browser benchmarks).

runtime code generation is where SPEC has a significant hole. GCC has significant istream pressure.

Benchmark selection is a real pain :biggrin:

Yes it is. But I'd wager that benchmarks I pointed out would get a better overview of performance than something like geekbench which is even less representative of actual workloads and has some serious fundamental flaws.

 

[Lepton87]

It very much is a harvested die, its a chop of a larger design. IVY-E btw was also a chop. As a chop, there is no optimization in the design for the lower core count parts. The actual design centers around the EP/EX functionality. The E dies are really a byproduct of the cheapest method to get a product in that given space. Hence the reason the E dies reuse the same package set from the EP/EX even though a significant number of the pads are basically blanks.

blah, semantics. The other argument would be that EP/EX are just regular dice made with copy and paste technique not true server parts like the IBM Power 8, both equally valid.

As for the performance, almost nobody who cares about it and buys the Extreme CPU leaves that CPU at stock-after overclocking clock-speed advantage of the small HW die is almost entirely gone or entirely - silicon lottery. At stock clocks the 4790K is indeed a good CPU but it almost has no frequency headroom left. If I didn't OC I would have taken the 4790K over the 5820K not to mention the 5960X which is a horrible value by comparison.
I'm not going to argue that HW-E parts are better than the regular HW parts and that 8/6 cores is better than 4.

 

[krumme]

If they are going to do a switch over, they are going to have to do it in the majority of the Mac product line or its going to fail. Even the MacBook Air requires absolute performance levels significantly higher than the iPAD.



The Zakate design is merely an exercise in integration at best. That required 80 people to make and didn't include anything new. The amount of work required was basically less than required for a raw shrink. If you think that's at all comparable to what is required for an entirely new design, you are underestimating the amount of work.



It is highly unlikely that the a9x has performance goals/targets appropriate even for the low end of the laptop domain.



You are assuming a rather significant performance downgrade is acceptable to the user base. I would argue that it isn't. The user base would be up in arms unless the new product delivers better performance than the older product. Esp when you factor in a complete software transition.

Looking at bw performance i have no doubt a a9x will walk all over it in sub 5w tdp territory.
But what absolute performance matters anyway for the future market? The batterylife and an ultra slim factor or 30% better cpu perf?
I can easily see a market for passive ultra slim mb air laptops with tons of battery life - thats very difficult using bw. On the contrary bw brings absolutely nothing new to the market.

 

[crashtech]

It very much is a harvested die, its a chop of a larger design. IVY-E btw was also a chop. As a chop, there is no optimization in the design for the lower core count parts. The actual design centers around the EP/EX functionality. The E dies are really a byproduct of the cheapest method to get a product in that given space. Hence the reason the E dies reuse the same package set from the EP/EX even though a significant number of the pads are basically blanks.

Usually I see the term "harvested die" in the context of a die with one or more substandard/malfunctioning cores fused off and sold as a lower core count part. If the die is based on an unoptimized "chop" of a higher core count design, I would think that there should be a better term to differentiate between that and an actual harvested die, which is indicative of something that happens during the binning process.

 

[Lepton87]

Usually I see the term "harvested die" in the context of a die with one or more substandard/malfunctioning cores fused off and sold as a lower core count part. If the die is based on an unoptimized "chop" of a higher core count design, I would think that there should be a better term to differentiate between that and an actual harvested die, which is indicative of something that happens during the binning process.

Usually? Have you seen it used in any other context? Because I haven't until I read this thread, maybe he forgot that HW-E actually has as much as 3 separe dice and now he is just saying he meant something else.
BTW. It's not true that there is no optimisation for the lower core count dice, the uncores of the 18 cores and the 8 cores dice are completly different, so this is BS. The memory controller is even different.

 

[Ajay]

This rumor is probably out there to give Apple better leverage when negotiating with Intel. It may even be for the purpose of negotiating a better fab deal with Intel for Apple's ARM SoCs. The last time around Apple didn't like Intel's price, but with their foundry partners having some problems, Apple may want to pin it's had on Intel foundries. Apple would gain access to the only 'A' list fab and get a competitive advantage in perf/watt - they just want more favorable terms. As an added bonus, Intel's process advantage may just give Apple the clock headroom to start moving Macs over to ARM (but, even if true, this would be something like 4-5 years out after signing a foundry deal).

 

[Arachnotronic]

Usually? Have you seen it used in any other context? Because I haven't until I read this thread, maybe he forgot that HW-E actually has as much as 3 separe dice and now he is just saying he meant something else.
BTW. It's not true that there is no optimisation for the lower core count dice, the uncores of the 18 cores and the 8 cores dice are completly different, so this is BS. The memory controller is even different.

They're difference dice, but as imported_ats said, they're "chops" of the main die. This means that the "uncore" of the chip is designed for the "big" 18 core configuration but then that main 18-core die is chopped down into smaller dice.

The point he's trying to make I think is that if Intel were to do a chip with an uncore focused on delivering the best performance for an 8 core processor, this would be a "better" solution than designing an 18 core beast that can be "chopped" into smaller bits.

 

[kimmel]

blah, semantics. The other argument would be that EP/EX are just regular dice made with copy and paste technique

Heh, copy paste. I love internet people commenting on things that they know nothing about. Designing a chip with 5 billion transistors is like copy paste. So easy any idiot could do it!

The point he's trying to make I think is that if Intel were to do a chip with an uncore focused on delivering the best performance for an 8 core processor, this would be a "better" solution than designing an 18 core beast that can be "chopped" into smaller bits.

he'd most likely be wrong.

 

[imported_ats]

blah, semantics. The other argument would be that EP/EX are just regular dice made with copy and paste technique not true server parts like the IBM Power 8, both equally valid.

No they aren't equally valid. Anyone who thinks those are equally valid is completely ignorant of computer architecture and the design of the IVB/HSW server parts.