Apple A9X the new mobile SoC king

[raghu78]

Seems its another case of the enemy of my enemy is my new best friend.

Shintai you will see Apple challenge Intel much better than AMD. They are a much more powerful company. In fact Apple could buy out Intel with their USD 202.8 billion cash pile. Intel's market cap is USD 136.3 billion. This competitor can outspend Intel in R&D and can even help foundries like TSMC to stay competitive with Intel on process nodes by sheer weight of their business on the bleeding edge process nodes.

http://www.forbes.com/sites/stevesc...t-massive-cash-pile-its-all-about-the-iphone/

http://finance.yahoo.com/q?s=INTC

 

[ShintaiDK]

The fact that Apple got such a big pile of money just shows their lack of competence. The return of that cash was ~1% last year. But that is another matter.

And Apple dont have 202.8B$ in cash:
http://www.google.com/finance?q=NASDAQ:AAPL&fstype=ii&ei=gnQCVsHPJMKcUsP5mZAG

Also you seem to forget the actual cost of those SoCs. Not to mention, what their real performance and target is. And are they even able to compete outside Apples closed system and sky high prices? Whenever its not a bench completely screwed towards ARM or their own ecosystem, their performance drop quite fast. Apple have now several times in a row doubled the memory bandwidth for example. If we compare 5775C vs 4790K. We all know you can easily get 20-30% without changing the core.

We can see on Samsung how fast a company can lose revenue and profit when the smartphone market shifts. From hero to zero in no time.

Its quite clear ARM, and specially Apple, is your new best friend after you gave up on your long loved AMD.

but ARM havent moved anywhere in terms of new markets. x86 however moved deep into tablets.

 

[raghu78]

The fact that Apple got such a big pile of money just shows their lack of competence. The return of that cash was ~1% last year. But that is another matter.

And Apple dont have 202.8B$ in cash:
http://www.google.com/finance?q=NASDAQ:AAPL&fstype=ii&ei=gnQCVsHPJMKcUsP5mZAG

Apple has already returned USD 112 billion to shareholders in the last 2.5 years.

https://www.apple.com/pr/library/2015/04/27Apple-Expands-Capital-Return-Program-to-200-Billion.html

"From the inception of its capital return program in August 2012 through March 2015, Apple has returned over $112 billion to shareholders, including $80 billion in share repurchases."

btw did you see the long term investments in their balance sheet. USD 168 billion. so what exactly are you arguing here.

Also you seem to forget the actual cost of those SoCs. Not to mention, what their real performance and target is. And are they even able to compete outside Apples closed system and sky high prices? Whenever its not a bench completely screwed towards ARM or their own ecosystem, their performance drop quite fast.

We can see on Samsung how fast a company can lose revenue and profit when the smartphone market shifts. From hero to zero in no time.

Its quite clear ARM, and specially Apple, is your new best friend after you gave up on your long loved AMD.

but ARM havent moved anywhere in terms of new markets. x86 however moved deep into tablets.

dude x86 has moved into tablets at the cost of contra revenue. Intel's cost structure sucks for mobile. btw ARM decimates Intel in mobile market overall share. Intel's share in phones which is a growing market is almost zero. In the shrinking tablet market they are losing money to get market share. Nice strategy :biggrin:

http://finance.yahoo.com/news/why-did-intel-start-recording-010537142.html

ARM's impact on servers will start in 2017. The software ecosystem is still not ready for ARM. More importantly FINFET based ARMv8 server products will not arrive in high volume until 2017. Anyway both of us are going to see who wins - ARM or x86.

 

[Nothingness]

Shame that A9 still sucks at 3DMark Physics

Has this been tested?

 

[NTMBK]

but ARM havent moved anywhere in terms of new markets. x86 however moved deep into tablets.

Intel burnt how many millions to win those x86 tablet sales? Meanwhile Apple is happily making money hand over fist.

ARM has already started moving into the laptop market with the Chromebooks, and the iPad Pro is competing with ultrabooks and 2-in-1 laptops.

 

[Nothingness]

but ARM havent moved anywhere in terms of new markets.

People think ARM only plays in the mobile market, but that's far from the truth and gets less true every year. See latest ARM investor day slides slides for instance.

 

[davygee]

Honestly though, how will the average user notice any difference from the 6 to 6S? Both are monsters but I don't see how it makes much difference in day to day use. Do you need that raw performance aside from e-peen status? Don't think so.

Which is precisely why I agree with raghu; going for this extreme performance must have a longer-term objective. And that objective is unseating Intel, otherwise it makes little sense.

We are probably looking at a 2-part reason for placing monster chips in their mobile phones. The first being that, they are virtually guaranteed strong sales of products containing these chips, so they can mass produce, thus reducing the overall cost and help finance the R&D of the chips. So to push forward with producing top-tier, monster performance in general, they need an avenue to push these chips via mass-market and iPhone and iPad fit the bill as they wouldn't get the numbers with MacBook & iMac sales.

The second, which is essentially a backup to the first point is that they are intent on creating true top performing mobile devices (iPhone/iPad) which will have essentially longer and longer lifespans.

They have just released iOS 9, which works on the 4 year old 4S, we are seeing huge increases of performance in chipsets since the 4S, especially with the 5S, which should essentially work fine for another 3-4 years.

The only big change I see coming is with iOS 10 (iOS X). Which may be 64bit-ONLY and may see the start of a full-on linkage with OS X as well.

It won't be long before we see ARM based laptops and desktops in the Mac range especially if the scores from the A9 and A9X are anything to go by.

 

[rainy]

And Apple dont have 202.8B$ in cash:

He is right and you're wrong (that happened to you quite often): http://www.ibtimes.com/apples-absur...ster-deal-game-changer-tesla-analysts-2020220

 

[Arachnotronic]

Has this been tested?

https://twitter.com/MoonshineDesign/status/646539639126294529

GFX score of 39663, Physics score of 10764.

 

[Enigmoid]

https://twitter.com/MoonshineDesign/status/646539639126294529

GFX score of 39663, Physics score of 10764.

*sigh* When will they implement a proper memory controller with random IO optimizations.

This is one of the reasons why the A7/A8/A9 are benchmark monsters but fall behind in real world tasks (relative to where you would expect them to be). Also why silvermont does so well in PCmark and 3dmark physics.

 

[Arachnotronic]

*sigh* When will they implement a proper memory controller with random IO optimizations.

This is one of the reasons why the A7/A8/A9 are benchmark monsters but fall behind in real world tasks (relative to where you would expect them to be). Also why silvermont does so well in PCmark and 3dmark physics.

Probably in the A10

 

[Eug]

*sigh* When will they implement a proper memory controller with random IO optimizations.

This is one of the reasons why the A7/A8/A9 are benchmark monsters but fall behind in real world tasks (relative to where you would expect them to be). Also why silvermont does so well in PCmark and 3dmark physics.

What real world tasks are you referring to?

 

[imported_ats]

Anybody can license it and make their own ARM chips. It isn't Open Source', but it's certainly close enough. x86 is closed and proprietary. Surely you understand the distinction?

That isn't an accurate statement. It can be only licensed by those that ARM chooses to license. This is in fact exactly the same as x86. In fact, it could be argued that x86 is in fact more open with multiple vendors being able to extend x86 in any way they wish to something that isn't really available with ARM.

 

[Face2Face]

So what do we go by? Geekbench or 3DMark Physic scores... Also, why doesn't AnandTech use Geekbench? There's got to be a reason for that?

 

[imported_ats]

So what do we go by? Geekbench or 3DMark Physic scores... Also, why doesn't AnandTech use Geekbench? There's got to be a reason for that?

Because Geekbench does have a lot of fundamental issues. For example, the datasets used for mobile and desktop are actually different. Primate Labs claims this only results in a 1% difference, but...

 

[Dribble]

That isn't an accurate statement. It can be only licensed by those that ARM chooses to license. This is in fact exactly the same as x86. In fact, it could be argued that x86 is in fact more open with multiple vendors being able to extend x86 in any way they wish to something that isn't really available with ARM.

ARM chooses to licence to pretty well everyone. Intel chooses to restrict x86 licenses to them, and AMD. No one else builds x86. Hundreds of different companies build ARM chips for all sorts of purposes. With those ARM licenses you can make all sorts of different custom chips with your choice of graphics, networking, video decoder, sound, modem, whatever else at whatever size/power/budget you want. Intel sells just a few types of chip - if it doesn't match your requirements tough. Intel chips are very expensive, ARM chips cost much less.

Intel made their fortune by completely locking down the market (x86) and forcing everyone to use their chips at their prices, often forcing manufacturers to buy Intel mainboards as well. That control allowed them to run their fabs flat out and keep nice high profit margins. That only works as long as you control the market, they no longer control any market where ARM has a presence.

In addition Intel needs huge profits to survive - in the PC market they take the profits. It's the parts not the final device that has the big margins. In ARM devices the manufacturers of the phones/tablets/whatever broke free from that. The parts are now cheap, there is no profit in them, the profit comes in the final device. So it's now Apple/Samsung making all the profit each time a phone is sold, not whoever made the bits that make up that phone which will be some Chinese/Taiwanese company working at stupidly small profit margins.

Intel can't switch to be like ARM because they are a $100 billion company which needs huge revenues with high margins to survive. ARM is a tiny company by comparison that can survive quite happily licensing their chips out for peanuts. There simply isn't the money selling cpu's any longer, you want money you've got to sell premium devices not the cpu's that go in them. That's basically why Intel are screwed outside of x86.

 

[paffinity]

With Intel's 99% share of x86 server market, they are doing pretty well with margins these days.

 

[scannall]

That isn't an accurate statement. It can be only licensed by those that ARM chooses to license. This is in fact exactly the same as x86. In fact, it could be argued that x86 is in fact more open with multiple vendors being able to extend x86 in any way they wish to something that isn't really available with ARM.

No, you can't make that argument. That you would even try is a little on the crazy side. Multiple vendors? Uh, AMD who they cross license with, and who is mostly irrelevant. Via makes embedded x86 parts for an even smaller market share.

 

[imported_ats]

ARM chooses to licence to pretty well everyone. Intel chooses to restrict x86 licenses to them, and AMD. No one else builds x86. Hundreds of different companies build ARM chips for all sorts of purposes. With those ARM licenses you can make all sorts of different custom chips with your choice of graphics, networking, video decoder, sound, modem, whatever else at whatever size/power/budget you want. Intel sells just a few types of chip - if it doesn't match your requirements tough. Intel chips are very expensive, ARM chips cost much less.

There are significantly more than 2 companies that can make x86 chips, fyi.

And the variety of ARM chips, at least outside of the low end embedded market isn't that great. In fact, the vast majority of them are almost completely the same, they are basically taking the ARM kits and fabing them.

Intel made their fortune by completely locking down the market (x86) and forcing everyone to use their chips at their prices, often forcing manufacturers to buy Intel mainboards as well. That control allowed them to run their fabs flat out and keep nice high profit margins. That only works as long as you control the market, they no longer control any market where ARM has a presence.

Intel controls the market largely because no one else is competitive in the various markets Intel competes in. And their pricing tends to be roughly equal or less than competitive solutions both now and historically.

In addition Intel needs huge profits to survive - in the PC market they take the profits. It's the parts not the final device that has the big margins. In ARM devices the manufacturers of the phones/tablets/whatever broke free from that. The parts are now cheap, there is no profit in them, the profit comes in the final device. So it's now Apple/Samsung making all the profit each time a phone is sold, not whoever made the bits that make up that phone which will be some Chinese/Taiwanese company working at stupidly small profit margins.

The only one making real profits in mobile phones/devices is Apple which makes roughly 90+% of ALL mobile device profits. There is almost no profit anywhere in the whole mobile ecosystem be it chips or complete devices. Even samsung operates their mobile devices division on basically zero profit. In fact, it can largely be argued that outside of Apple, the only companies making any real profit in mobile devices are ARM Inc and TSMC.

Intel can't switch to be like ARM because they are a $100 billion company which needs huge revenues with high margins to survive. ARM is a tiny company by comparison that can survive quite happily licensing their chips out for peanuts. There simply isn't the money selling cpu's any longer, you want money you've got to sell premium devices not the cpu's that go in them. That's basically why Intel are screwed outside of x86.

Intel can survive just fine on significantly less profits than they currently make. The primary reason their profits are so high is that they take the profits from both the fab side and the chip side. In fact, if you look at their profits in that regard they aren't really making much more than say Nvidia+TSMC.

And actually, Intel is much less screwed than the majority of players in the mobile device market simply because unlike those players, Intel is actually in a market where profits can actually be made.

 

[imported_ats]

No, you can't make that argument. That you would even try is a little on the crazy side. Multiple vendors? Uh, AMD who they cross license with, and who is mostly irrelevant. Via makes embedded x86 parts for an even smaller market share.

AMD, Via, IBM, and actually several others. The fact that their competitors couldn't compete and largely vacated the markets that Intel competes in is basically immaterial.

 

[pepone1234]

AMD, Via, IBM, and actually several others. The fact that their competitors couldn't compete and largely vacated the markets that Intel competes in is basically immaterial.

AMD, Via and no one else.

Others like nvidia tried to build an x86 cpu and intel almost kills them.

 

[simboss]

I wonder if you actually believe what you are saying or are trolling?

There are significantly more than 2 companies that can make x86 chips, fyi.

It is impossible for anyone today to license the x86 ISA and design a new CPU with it, not only for legal reasons, but also technical reasons. Basically it is too complex to do it from scratch without infringing Intel patents and burning an insane amount of money, as Intel or AMD will not help you in any way to get it right.

Intel has killed all the competition with a combination of successful engineering and dubious business practices (what weight you give to each is a very personal view), so that means that nowadays the situation is that if you need x86, you need Intel.

And the variety of ARM chips, at least outside of the low end embedded market isn't that great. In fact, the vast majority of them are almost completely the same, they are basically taking the ARM kits and fabing them.

The increasing number of ARM licensees, both architectural and for ARM cores, says otherwise.
you can't seriously say x86 is more "open" than ARM. although I agree saying ARM is "open" may not be accurate either.

The fact that ARM is the only one defining the ISA is not even true, the ISA itself (actually the architecture, it is more than just an ISA) is a product they sell, they take input from their customers about it.
They also sell tools to make it possible to design a CPU from scratch and validate it.

It is true that you can't design an ARM core and add extensions to the ISA, so in that respect the x86 ISA is more "open", but I think the correct word for it is fragmented.
And although this has happened in the past, in the current situation only extensions done by Intel will ever get some adoption, meaning it is the de-facto standard, not much openness about it.

 

[Samwell]

It is impossible for anyone today to license the x86 ISA and design a new CPU with it, not only for legal reasons, but also technical reasons. Basically it is too complex to do it from scratch without infringing Intel patents and burning an insane amount of money, as Intel or AMD will not help you in any way to get it right.

Technical Reasons aren't such a problem, many companys have cross licensing agreements or there is a silent agreement not to sue each other. But that all doesn't help if Intel doesn't allow anyone to get X86. Nvidias case showed it to everyone, that you can forget it.
But looking at Intels market position i didn't understand why no one ever tried to sue intel to make x86 FRAND.

 

[imported_ats]

Technical Reasons aren't such a problem, many companys have cross licensing agreements or there is a silent agreement not to sue each other. But that all doesn't help if Intel doesn't allow anyone to get X86. Nvidias case showed it to everyone, that you can forget it.
But looking at Intels market position i didn't understand why no one ever tried to sue intel to make x86 FRAND.

The reality is that we are unlikely to ever know what the issues were wrt Nvidia/Intel concerning Nvidia wanting to build an x86 chip. What we know is that Nvidia had a bus license and claimed that the terms of that license allowed them to do x86 and Intel claimed it did not. The settlement such as it was ended up with Nvidia agreeing not to make an x86 processor, relinquishing its bus license and licensing Intel all of Nvidia patents for an unspecified period in exchange for some money.

As far as x86 and FRAND, a large portion of x86 is already unencumbered. Pretty much anything from P6 back should be in the public domain at this point as far as patents. The issues would largely concern additions to the x86 ISA from the post P6 era. The parts of x86 that aren't free at this point are primarily the SIMD, 64b support, and the various SMI/security features.

And suing to make things licensed under FRAND isn't really viable unless they have previous been offered up under a framework supporting FRAND aka open published standards.

 

[imported_ats]

It is impossible for anyone today to license the x86 ISA and design a new CPU with it, not only for legal reasons, but also technical reasons. Basically it is too complex to do it from scratch without infringing Intel patents and burning an insane amount of money, as Intel or AMD will not help you in any way to get it right.

Neither will ARM help you get it right if you take out an architectural license. And it is not impossible for anyone today to license the x86 ISA. Nor is it too complex, there are numerous examples of companies creating totally independent x86 designs.

The increasing number of ARM licensees, both architectural and for ARM cores, says otherwise.
you can't seriously say x86 is more "open" than ARM. although I agree saying ARM is "open" may not be accurate either.

Fine, I'll agree that neither is open.

The fact that ARM is the only one defining the ISA is not even true, the ISA itself (actually the architecture, it is more than just an ISA) is a product they sell, they take input from their customers about it.
They also sell tools to make it possible to design a CPU from scratch and validate it.

The tools they sell won't get you that far. There are free tools that are basically the same thing for x86 as well, fyi.

It is true that you can't design an ARM core and add extensions to the ISA, so in that respect the x86 ISA is more "open", but I think the correct word for it is fragmented.
And although this has happened in the past, in the current situation only extensions done by Intel will ever get some adoption, meaning it is the de-facto standard, not much openness about it.

AFAIK, there has only been 1 company that has received a full architectural license from ARM which was DEC. DEC's license allowed them to add to/modify the ISA, IIRC. After DEC's license it was over a decade before another architectural license was granted and AFAIK, it was not as extensive as the license that DEC originally got. I don't know if the license was modified when StrongARM was transferred to Intel. Nor do I know if Intel kept the license intact after selling StrongARM to Marvell.

Also, its important not to confuse Open with you can do whatever you want. There are lots of examples of open standards with de-facto standards/restrictions. SPARC is technically Open but good luck adding instructions and getting adoption/support. Same for RISC-V.