Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Reasons Behind Apple's Strategic Shift Into Chip Design
- Thread starter MacRumors
- Start date
- Sort by reaction score
They outsource the FAB. It is the mask they need to design.
Listen carefully to the key benefits described. Exclusivity, so no one element of the ecosystem that is open source opens the remainder of the system.
Low power. Have you noticed the primary human interaction "issue" with electronic devices is the need to fixate on constant awareness of recharging? Minimizing that would be a huge human interaction leap.
By customizing to a "crippled" ecosystem, they get smaller die sizes, lower power, and purpose built graphics and conversion silicon. A place for everything and everything in its place.
By being at the bleeding edge of size, power, specified capabilities and leading edge I/O, Apple will not only be more advanced and desired, but more useful to more people.
I would estimate Apple will have a worldwide handset within 5 years.
Rocketman
As I see it
Ok, here's my analysis.
You need 4 things to effectively branch out into chip design: expertise, money, time, and volume. The first two are self-evident and also the most important. This is good, because Apple obviously now has a lot of both.
Time is a bit less self-evident, so I'll explain. Back when IBM was making the PC, they were a juggernaut. They has piles of cash, and plenty of chip design expertise. In fact, they were developing the 801 at the same time as the IBM PC; a chip that ran circles around intel. SO why did they go with intel? Simply put, they had nothing on the market. They were rightly scared that every second they didn't have a product was another second that their competitors become more entrenched and powerful. In short, they had no time.
Apple's situation is totally different. They have THE phone and tons of mindshare. While this does not mean that they can rest on their laurels, it does mean that they are in a position of power, and thus don't have to rush anything.
Finally, there is volume. let me preface this by saying I have no idea how much it costs to develop a chip. What I do know is that a chips development is largely a one time cost: after it starts shipping, you might only do minor tweaks to it's design. So let's say it costs 200 million dollars to design a chip that manufactures for 10 dollars, and that chip sells only 10 million units. Yes, it only cost $10 to manufacture, but the real cost was $30. Now say that same chip sells 100 million. Now the real cost is $12. You need volume to recoup the cost of designing to chip in the first place.
That's why I'm predicting a slightly different strategy from Apple this time. I think that they will go for marketshare in a serious way. I don't think anyone here would put it past Apple selling 100m iphones/ipod touches a year in the near future (heck, given the 4th quarter boost, they might break 50m this year). If they do, volume concerns are a non-issue. 100m is more than AMD ships, more than Ti's OMAP ships. It would easily place them as a top ARM design house by volume, except with themselves as their only customer.
In short, Apple has the pieces they need to do this, and are probably the only ones that can say that. I can't think of anyone else with the combination of expertise, money, time, and volume needed to pull this off. But there is one more thing Apple needs: luck. This is a monstrously difficult thing that they are trying to do, and hats off to them for trying it.
Ok, here's my analysis.
You need 4 things to effectively branch out into chip design: expertise, money, time, and volume. The first two are self-evident and also the most important. This is good, because Apple obviously now has a lot of both.
Time is a bit less self-evident, so I'll explain. Back when IBM was making the PC, they were a juggernaut. They has piles of cash, and plenty of chip design expertise. In fact, they were developing the 801 at the same time as the IBM PC; a chip that ran circles around intel. SO why did they go with intel? Simply put, they had nothing on the market. They were rightly scared that every second they didn't have a product was another second that their competitors become more entrenched and powerful. In short, they had no time.
Apple's situation is totally different. They have THE phone and tons of mindshare. While this does not mean that they can rest on their laurels, it does mean that they are in a position of power, and thus don't have to rush anything.
Finally, there is volume. let me preface this by saying I have no idea how much it costs to develop a chip. What I do know is that a chips development is largely a one time cost: after it starts shipping, you might only do minor tweaks to it's design. So let's say it costs 200 million dollars to design a chip that manufactures for 10 dollars, and that chip sells only 10 million units. Yes, it only cost $10 to manufacture, but the real cost was $30. Now say that same chip sells 100 million. Now the real cost is $12. You need volume to recoup the cost of designing to chip in the first place.
That's why I'm predicting a slightly different strategy from Apple this time. I think that they will go for marketshare in a serious way. I don't think anyone here would put it past Apple selling 100m iphones/ipod touches a year in the near future (heck, given the 4th quarter boost, they might break 50m this year). If they do, volume concerns are a non-issue. 100m is more than AMD ships, more than Ti's OMAP ships. It would easily place them as a top ARM design house by volume, except with themselves as their only customer.
In short, Apple has the pieces they need to do this, and are probably the only ones that can say that. I can't think of anyone else with the combination of expertise, money, time, and volume needed to pull this off. But there is one more thing Apple needs: luck. This is a monstrously difficult thing that they are trying to do, and hats off to them for trying it.
People obsessed with quality/cost and other concerns.
Personally I don't think that quality even enters into the equation when considering custom components. At least not in the sense of a more reliable machine.
Rather what I see is potential for Apple to offer up a lot more performance for a given dollar. If you focus on products using ARM, the vast majority of those products are powered by custom implementations. All of those custom ARM systems are out there because it is the low cost way to implement required functionality. It is only recently that manufactures have tried to produce ARM SoC that meet broad needs of the handset business. So it isn't really a surprise that Apple would do a custom SoC for portable devices, they have to to remain competitive.
The big question is how far will they go custom. Will it be a more or less standard ARM at the core or will they enhance things even more. I can't see Apple doing it's own custom GPU core on a SoC simply because they are heavily invested in a company that does that. So I'd look at other possibilities such as enhancing the ARM instruction set to accellerate the execution of Objective C code or possibly a separate vector core. One possibility here would be reworking ARMs Java accelleration feature to accellerate Objective C. The alternative would be to expand the ARM instruction set but that is already pretty much used up as in few op codes free.
The real dark horse here is the possibility of a co-processor, maybe modelled on the vector cores in cell. But I'm not sure if this is realistic considering efforts put forth to make sure the coming embedded GPUs are OpenCL compatible. The way I see it if Apple is going to add additional features they will have to be features they can't get from standard components on a SoC.
Dave
Personally I don't think that quality even enters into the equation when considering custom components. At least not in the sense of a more reliable machine.
Rather what I see is potential for Apple to offer up a lot more performance for a given dollar. If you focus on products using ARM, the vast majority of those products are powered by custom implementations. All of those custom ARM systems are out there because it is the low cost way to implement required functionality. It is only recently that manufactures have tried to produce ARM SoC that meet broad needs of the handset business. So it isn't really a surprise that Apple would do a custom SoC for portable devices, they have to to remain competitive.
The big question is how far will they go custom. Will it be a more or less standard ARM at the core or will they enhance things even more. I can't see Apple doing it's own custom GPU core on a SoC simply because they are heavily invested in a company that does that. So I'd look at other possibilities such as enhancing the ARM instruction set to accellerate the execution of Objective C code or possibly a separate vector core. One possibility here would be reworking ARMs Java accelleration feature to accellerate Objective C. The alternative would be to expand the ARM instruction set but that is already pretty much used up as in few op codes free.
The real dark horse here is the possibility of a co-processor, maybe modelled on the vector cores in cell. But I'm not sure if this is realistic considering efforts put forth to make sure the coming embedded GPUs are OpenCL compatible. The way I see it if Apple is going to add additional features they will have to be features they can't get from standard components on a SoC.
Dave
Dave this is really the area to look at. We know Apple's not going to try to replace decades of ARM design or build their own GPU when they have Imagination to do that. It's the "extras" we need to be looking at. It's the gray area between the CPU and GPU that Apple's going to look to differentiate itself otherwise there's little point to building your own chip.
And infinitely proprietary. No more Hackintoshes if you can't get the chips at all.
Not to mention that Apple seems to really dislike giving out detailed specs. With their own chips, they can keep everyone in the dark.
Think different.
Apple is definitely not "most big electronic firms" and will continue to design more and more of their own components until they are shed of outsourcing all together.
Apple is definitely not "most big electronic firms" and will continue to design more and more of their own components until they are shed of outsourcing all together.
Agreed
I agree...
I do not think Apple will exactly fab their own chips - but design/optimise them inhouse before out sourcing them.
This would decrease the lead time for leaks and other manufacturers to copy (or just buy the same compenents and clone).
In the end its not that hard for the competition to buy an Apple product and reverse engineer it - although Apple would have the IP for it.
I agree...
I do not think Apple will exactly fab their own chips - but design/optimise them inhouse before out sourcing them.
This would decrease the lead time for leaks and other manufacturers to copy (or just buy the same compenents and clone).
In the end its not that hard for the competition to buy an Apple product and reverse engineer it - although Apple would have the IP for it.
Why doesn't Apple go to the other extreme and instead of being veiled in secrecy with regards to their future products, why doesn't Apple just float out numerous trial balloons and let the rumor mill websites run with it. Apple can CYA by simply stating this is what we are R&D'ing but with regards to final product, you'll just have to wait and see what we have when we announce.
I know Microsoft will go berserk trying to create their own copycat departments with their "me too" products.
With so many "rumors" floating out there, it will bring back the suspense to Apple's announcements and have people more tentatively scouring the rumor websites for the latest gossip.
Hey, it worked for D-Day when the Allies had feints here and there to confuse the Germans as to their true intent.
It's just a thought because the whole "secrecy" thing seems to bring about it's fair share of predicaments.
I'm not a chip designer, from my limited perspective I just like the powerPC better.
The powerPC :
1. Started with a blank slate with very little legacy compatibility issues
1. Bigger L1 cache (downside: less chip yield, more heat)
2. More general purpose registers to work with
3. RISC based
4. I loved my G5 PowerMac, it was fast at the current time
5. My g4 powerbook continues to run nice and cool, the new Intel Macbook is super super hot!
The powerPc just seems to be the way I would want to build the NeXT chip, from the ground up with little baggage.
However, Apple needed Intel to be as successful as they are today. I just miss my very quiet and cool PowerBook G4. My intel MacBook fan noise and heat continue to be a major bummer for me.
custom chips can save costs
You have missed the point - the trick to making electronics cheap is to cut the chip count - this save manufacturing costs, improves reliability and cuts power consumption. Also allows you to reduce the size which is a big issue for Steve. If by designing their own chips Apple can cut the chip count in the device the savings in chip count would probably more than offset the increased cost of the individual chip. If it makes the device thinner and gives it better battery life it makes it cooler
Makes perfect sense to me
Proprietary parts and design. Increased per unit costs. Sounds like a grea fit for Apple. Brilliant.
You have missed the point - the trick to making electronics cheap is to cut the chip count - this save manufacturing costs, improves reliability and cuts power consumption. Also allows you to reduce the size which is a big issue for Steve. If by designing their own chips Apple can cut the chip count in the device the savings in chip count would probably more than offset the increased cost of the individual chip. If it makes the device thinner and gives it better battery life it makes it cooler
Makes perfect sense to me
It's not like Intel spends billions on building a fab every 3 years, let alone billions more in RnD, not to mention a few more billion on equipment, then sure throw a couple more billion to actually manufacture the chips.
It will be interesting to see how this will affect the cost to the consumer.
Oh good Lord!
That a hot tamale! I have a 2 year old, 17" MBP that I use to encode video sometimes and that 2.33GHz Dual Core got as high as 190F before I panicked & started using a small external fan to bring temps below about 160F. Am I too protective of the processor in my Mac?...
Pinky and the Brain plotted to over the world.
Ren and STimpy sang 'happy happy joy joy'
/napolean GOSH!
ON TOPIC: Do leaks even matter if you literally control and are the only source and only customer of the only hardware that can run it? It's not like there would be enough leak to make emulators avaialble, and even if there were, no one is going to run an 'iphone emulator' on their ZunePhone or whatever.
chinese shanzhai (knock-off) phone industry
Over at Hacker News there was an interesting comment by someone with the username "9oliYQjP" which suggested that this move may be more about thwarting the knock-off industry that makes things like the HiPhone.
Over at Hacker News there was an interesting comment by someone with the username "9oliYQjP" which suggested that this move may be more about thwarting the knock-off industry that makes things like the HiPhone.
Concurrency. GPGPU. Apple seems to be lining up to have the benefits from 10.6 added into it's mobile devices. They're coming from an OS from the computing side, which benefits from more power in all the senses. By getting better performance per watt from going SoC, custom chips, they're improving battery life and performance. It's classic hand me down benefits - try something in one area, then transfer the knowledge and benefits to others. If they can get GPGPU, Grand Central Dispatch, OpenCL etc working - what's to stop them going for custom designs to add this into their products?
Then you've got better resources, and better utilisation of resources, with Apple controlling a lot of it -And it can have benefits - Just look at the reverse, how Microsoft first moaned it was a silly thing to have such control on software and hardware and their integration, then they went running to HTC to get 6.5 integrated better, after stifling Android at Barcelona.
Apple's got a decent lead on Symbian, RIM in terms of OS. Palm's coming from a leaner lighter OS point of view, that has it's own pros and cons.
But it's not just mobile level devices - the custom silica is getting powerful enough to start being twinned with other chips to become development relevant to Apple TV, Mac mini etc.
Apple's had it's finger in the Chip design pie to an extent I'd imagine - it would be interesting if anyone can link to a decent sources on the history of what Apple's been up to. You can count Apple's relationship with Intel in this way too. (Anyone want to bet against Apple getting early access to Clarksfield/Arrandale, or having a special early access to something for the MacBook Air etc?)
Then you've got better resources, and better utilisation of resources, with Apple controlling a lot of it -And it can have benefits - Just look at the reverse, how Microsoft first moaned it was a silly thing to have such control on software and hardware and their integration, then they went running to HTC to get 6.5 integrated better, after stifling Android at Barcelona.
Apple's got a decent lead on Symbian, RIM in terms of OS. Palm's coming from a leaner lighter OS point of view, that has it's own pros and cons.
But it's not just mobile level devices - the custom silica is getting powerful enough to start being twinned with other chips to become development relevant to Apple TV, Mac mini etc.
Apple's had it's finger in the Chip design pie to an extent I'd imagine - it would be interesting if anyone can link to a decent sources on the history of what Apple's been up to. You can count Apple's relationship with Intel in this way too. (Anyone want to bet against Apple getting early access to Clarksfield/Arrandale, or having a special early access to something for the MacBook Air etc?)
I think it's more a thing of boosting up P.A. Semi's design staff. I really doubt they're looking to become AMD version 2.0 with these guys... I can't see them trying to out-do Intel with processors (for Macs). I think it's more a thing of getting some guys in that know and understand processors, be it Intel's or AMD's, in order to help P.A. Semi design other chips that work more effective alongside processors.
Just because these guys come from AMD doesn't mean the only thing they know and understand are CPUs. They could very well be there to help P.A. streamline whatever it is they're working on.
On the other hand, Apple's not using Intel processors for iPods and iPhones, right? Maybe they're using these guys to experiment with some designs that don't compete directly with Intel (or remove Intel from a product line). Apple's got too good a relationship with Intel to be pissing them off.
And, I'm sure Intel would love to see Apple pick up some major market share. Sure, they will sell just as many processors to whoever is selling units, but with Windows being such a debacle in the general public (with the exception of the few smart and knowledgeable users that can keep their Windows boxes ticking along virus and crash free), I'm sure they'd like to get away from people having trouble with their PC looking at the CPU box on each forced reboot seeing the "Intel Inside" sticker and (subconsciously perhaps) blaming them instead of M$.
Just because these guys come from AMD doesn't mean the only thing they know and understand are CPUs. They could very well be there to help P.A. streamline whatever it is they're working on.
On the other hand, Apple's not using Intel processors for iPods and iPhones, right? Maybe they're using these guys to experiment with some designs that don't compete directly with Intel (or remove Intel from a product line). Apple's got too good a relationship with Intel to be pissing them off.
And, I'm sure Intel would love to see Apple pick up some major market share. Sure, they will sell just as many processors to whoever is selling units, but with Windows being such a debacle in the general public (with the exception of the few smart and knowledgeable users that can keep their Windows boxes ticking along virus and crash free), I'm sure they'd like to get away from people having trouble with their PC looking at the CPU box on each forced reboot seeing the "Intel Inside" sticker and (subconsciously perhaps) blaming them instead of M$.
Great points. I would still argue that production quality is an industry problem. But add, as you point out, that what this means is the result of an interplay between customers' willingness to pay (and understanding of quality) and computer makers.
As you state, computer makers will only invest to improve quality if the incentives are right.
I did not mean to equate off the shelf parts with good quality. I intended to suggest that off the shelf parts are of sufficient quality for the industry and market. The quality does vary, but as long as it is within a tolerance range the industry and market will accept it.
The Apple uses predominately standardised components its products. However, the distinctive design does require modifications to some of these components.