Jobs might be underestimating the difficulty of designing chips and the experience of some of the companies that currently help design and produce the chips in their devices. Apple may have some of the world's most brilliant engineers, but Intel, Samsung, and some of the other makers have a great deal more experience in these areas.
Apple will get it done, but it'll be rockier than he expects and could even affect quality.
A good friend of mine shared his philosophy on fixing bicycles once. Try to fix it yourself, and if you mess it up then just bring it in to a mechanic to fix it.
Apple can easily attempt to design their own chip in house, and if things get too complicated, they can switch to outsourcing.
I doubt very much that Jobs is underestimating the effort or the risk. This is more likely the Apple response to why an iPhone?
Look, if the COTS chips within the iPhone are available to everyone, (they are), then anyone can implement the same features (they will). The COTS chip makers know this, and in fact often they put only the features that have mass appeal in their SOCs creating a lowest common denominator solution, which maximizes their market.
Now consider what happens if Apple can design a chip that they, and only they use. Then the features within the iPhone will have a full lifecycle advantage over the competition; the SOC vendors will of course copy what makes the Apple feature "cool" but they will always be a generation behind...
You're making the assumption that such a chip would reach market at the same time and in the same condition. The existing chip design companies are, in fact, really good at what they do, and they are rushing these things to market about as fast as can be done. Hardware product cycles are much longer than those for software, especially if you are trying to "innovate" like you posit. You can't just rev a new chipset for ever device release.
Apple doesn't need a chip advantage. They are able to beat the competition just fine without such a thing.
There's no hardware in an iPod that couldn't be duplicated by any other company, easily. Their competitors think "Hey, we can make one of those things far cheaper, with more features." Then they produce a clunky ugly awful piece of junk which quickly fades from view. Yet another "competitor" vanquished.
In the PC realm, Macs have recently become MORE like PCs, not less. Again, a hardware advantage wouldn't help at all.
I think this is a horrible idea. There's virtually no hardware functionality that can't be attained cheaply in the OEM chip market. My guess is that Apple wants to introduce embedded DRM/locks at the hardware level that cannot be circumvented so that they can make sure that their iTunes based media, iPhone apps, and whatever else they have in the works cannot be knocked off.
This sounds like "plan b". There was lots of talk of "intel partnership", but if Intel won't build what Apple wants, perhaps because they think it is stupid, then enter plan b.
Apple has a long tradition of using custom chips where x86 uses a northbridge through the 68k and PPC eras. I don't know if they hired out the work or did it with an in house team, but the distinction is largely irrelevant now that they brought a chip design team in house.
I believe the rationale was to keep the part count down by integrating as many functions as possible. This is especially important in small consumer devices.
Consider as an example, not Apple but similar issues or how the big chip companies can miss on specific requirements... I need to run a server in a remote location off of solar power and batteries. I am tired of wrestling with the OpenWRT and wrtsl54gs that I use now, but it runs on a couple of watts and that is good. Intel releases their fabulous Atom processor that takes only 2 watts which is great because I can use a standard mini-itx motherboard, regular old debian x86 and be happy... except the northbridge chip on the motherboard takes 15 watts because Intel apparently didn't see a need for a lower power northbridge. I can't afford another 360 watthours/day just for the northbridge. Obviously my volume is wrong for custom silicon, but a system asic that supports just the peripherals I need and memory I need designed with an eye to power dissipation instead of top notch performance would be a big win.
I thinks its closed source, also what was the reason Apple came to intel on their computers? - and why would they reverse that on their phones? .... just seems apple is getting more evil - although probably most companies are
- IBM PPC manufacturer reliability was killing them, there was a certain scandal with the G5's getting past the 2.0Ghz barrier on a scale of 90 nano meters I believe.
- Boot camp, with an intel processor they could influence people on the fence to get a Mac, or people that wanted to (go forbid) run windows on a Mac.
Remember, OS X is a micro kernel architecture and can easily be ported to multiple processors.
I don't think that this is all that bad an idea, even though Apple's not much of a silicon design shop. Microsoft isn't either, and yet they were able to REDUCE costs by getting a custom-designed processor for the XBox360 (yes, I know that had its share of flaws).
Intel doesn't do custom designs, that isn't their business model. There are some companies that specialize in that sort of thing, like IBM. There are a number of fabless semiconductor developers that have been quite successful this way, including nVidia. Even SUN doesn't do its own semiconductor manufacturing, for years they relied on TI for that part.
Also, Apple acquired the company PA Semiconductor, which includes several folks with a lot of experience in designing processors (some of their products were among the best in the business, like the Alpha series), so it's not like Apple would be starting from ground zero here.
I'm not sure that it's necessary, but if Apple wants custom functionality for upcoming mobile products, it might be worth it in the long run. If they just wanted inexpensive, high-performance, general-purpose processors, the best route for the foreseeable future is probably to stick with x86. I don't think it's likely however that x86 will be able to compete with the ARM in the ARM's niche anytime soon though.
14 comments
[ 2.7 ms ] story [ 79.2 ms ] threadApple will get it done, but it'll be rockier than he expects and could even affect quality.
Apple can easily attempt to design their own chip in house, and if things get too complicated, they can switch to outsourcing.
Look, if the COTS chips within the iPhone are available to everyone, (they are), then anyone can implement the same features (they will). The COTS chip makers know this, and in fact often they put only the features that have mass appeal in their SOCs creating a lowest common denominator solution, which maximizes their market.
Now consider what happens if Apple can design a chip that they, and only they use. Then the features within the iPhone will have a full lifecycle advantage over the competition; the SOC vendors will of course copy what makes the Apple feature "cool" but they will always be a generation behind...
There's no hardware in an iPod that couldn't be duplicated by any other company, easily. Their competitors think "Hey, we can make one of those things far cheaper, with more features." Then they produce a clunky ugly awful piece of junk which quickly fades from view. Yet another "competitor" vanquished.
In the PC realm, Macs have recently become MORE like PCs, not less. Again, a hardware advantage wouldn't help at all.
Apple has already acquired PA Semi, a chip design company ( http://news.ycombinator.com/item?id=174301 ). Robert X Cringely implied that Apple could use the acquisition to reduce component costs by US$278 million or more just with the threat of switching to unproven developments ( http://www.pbs.org/cringely/pulpit/2008/pulpit_20080425_0047... ).
Apple has a long tradition of using custom chips where x86 uses a northbridge through the 68k and PPC eras. I don't know if they hired out the work or did it with an in house team, but the distinction is largely irrelevant now that they brought a chip design team in house.
I believe the rationale was to keep the part count down by integrating as many functions as possible. This is especially important in small consumer devices.
Consider as an example, not Apple but similar issues or how the big chip companies can miss on specific requirements... I need to run a server in a remote location off of solar power and batteries. I am tired of wrestling with the OpenWRT and wrtsl54gs that I use now, but it runs on a couple of watts and that is good. Intel releases their fabulous Atom processor that takes only 2 watts which is great because I can use a standard mini-itx motherboard, regular old debian x86 and be happy... except the northbridge chip on the motherboard takes 15 watts because Intel apparently didn't see a need for a lower power northbridge. I can't afford another 360 watthours/day just for the northbridge. Obviously my volume is wrong for custom silicon, but a system asic that supports just the peripherals I need and memory I need designed with an eye to power dissipation instead of top notch performance would be a big win.
- IBM PPC manufacturer reliability was killing them, there was a certain scandal with the G5's getting past the 2.0Ghz barrier on a scale of 90 nano meters I believe.
- Boot camp, with an intel processor they could influence people on the fence to get a Mac, or people that wanted to (go forbid) run windows on a Mac.
Remember, OS X is a micro kernel architecture and can easily be ported to multiple processors.
Intel doesn't do custom designs, that isn't their business model. There are some companies that specialize in that sort of thing, like IBM. There are a number of fabless semiconductor developers that have been quite successful this way, including nVidia. Even SUN doesn't do its own semiconductor manufacturing, for years they relied on TI for that part.
Also, Apple acquired the company PA Semiconductor, which includes several folks with a lot of experience in designing processors (some of their products were among the best in the business, like the Alpha series), so it's not like Apple would be starting from ground zero here.
I'm not sure that it's necessary, but if Apple wants custom functionality for upcoming mobile products, it might be worth it in the long run. If they just wanted inexpensive, high-performance, general-purpose processors, the best route for the foreseeable future is probably to stick with x86. I don't think it's likely however that x86 will be able to compete with the ARM in the ARM's niche anytime soon though.