Let me rephrase my thinking here. Apple need to find a good partner to do the fabbing of these chips to power iPods and iPhones. In the case of the iPhone they need a partner willing to keep up with Intel in the smaller fabbing process. They can't afford to partner with someone who stops at 45nm while Intel keep pushing things smaller. This does worry me a little. Not many are able to keep up with Intel in the chip fab business. At this point Apple and Intel are now competitors because Atom chips at 35nm will likely start showing up in competing smart phones.
I wasn't denying that Apple would need a cooperative company with a fab facility. But being that there is one Apple, hundreds of other companies to buy intels chips, and Apple will be keeping their chips all to themselves, I don't see intel as denying Apple the fab. Intel will still make money from it, and Apple already is working with, and has cooperation from inside intel to produce other products. I think with all the existing business intel and Apple do together I don't think this will put a strain on either one of them or their relationship. It's a huge technology world, and intel is inside everything one way or another. That's their business. Being a Fab facility is also their business.
I wonder how a 45nm RISC/ARM PA Semi custom chip would compare to a 32nm Atom. I know its a hypothetical question but someone at Apple surely has looked at this. I'm guessing that the RISC/ARM blows it away or Apple wouldn't be going this route.
ARM does amazingly well with respect to power usage even at 65nm. I don't think Intel will ever be able to catch up with that as there is just a huge advantage with respect to the number of transistors ARM needs to implement. On top of that ARM has been in the embedded world for years and building SoC is well understood for the hardware.
I don't really expect any surprises here other than a highly integrated SoC that allows for a minimal of external chips.
Not in its current iteration. However when its fab moves to 32nm everything I've read suggests that Intel will push hard for its adoption in smart phones.
I'm not sure Intel will want to fab a chip that competes with one of their products even though it would only be used by Apple.
Intel can push all they want but to put ATOM into a cell phone would so disadvantage a manufacture as to make them totally uncompetitive. That is not a negative on ATOM though, as I see it as an excellent effort by Intel and an awesome chip in the right sort of devices.
The right sort of devices being hardware you would want desktop style OS'es on. Sorry but I simply don't want or need a desktop OS on my cell phone nor on any mini tablet type device.
As an interesting aside I could also see ATOM going into low cost desktop multi processor systems. Think a Mini with 8 of these cores. With Apples move to enhance OS/X for multiprocessing and system resource usage this could actually end up being a very nice machine for many people. That would put 8 real thread or 16 if you look at SMT as being viable. The big thing here is the machine wouldn't even draw the power the current Mini does for the most part.
I find it funny that people see Snow leopard as Apple trying to optimize for the Mac Pro class machine. With ATOM such an OS re-factoring will be leveraged on the lowest cost machines.
I wasn't denying that Apple would need a cooperative company with a fab facility. But being that there is one Apple, hundreds of other companies to buy intels chips, and Apple will be keeping their chips all to themselves, I don't see intel as denying Apple the fab. Intel will still make money from it, and Apple already is working with, and has cooperation from inside intel to produce other products. I think with all the existing business intel and Apple do together I don't think this will put a strain on either one of them or their relationship. It's a huge technology world, and intel is inside everything one way or another. That's their business. Being a Fab facility is also their business.
I may very well be wrong but I don't think Intel will want to fab ARM/RISC chips for Apple. Atom is an important product for Intel and its success in the market is far from assured. I suspect that if it does not gain a foot hold into the smart phone market it will not meet Intel's goals for the product. I think the smart phone, handheld, market is one that Intel covets and currently has no product for.
In two years(roughly,IIRC), Intel will have 32 nm Atom chips. These will go into devices(Android phones perhaps) that will be direct competitors to the iPhone. If Intel are fabbing Atoms at 32nm and ARM chips are at 45nm that may keep Atom competitive enough performance wise to give Atom chips a chance in the market place. If Intel are fabbing Apple's custom ARM/RISC chips at 32nm it likely will be no contest and the performance of iPhones will be so much better than Atom powered smart phones will be at a severe disadvantage.
It looks like Atom vs ARM/RISC is only a fair fight if Atom is one node ahead in the fab process. To me it looks like the smart phone battle is going to be Atom at 32nm vs ARM at 65 or 45 nm.
Making your own chip is one way to differentiate yourself from the competition, but it all comes down to who has the talent. If Intel keep vacuuming up the best talent then the only way Apple will differentiate themselves is in a negative way. Of course they must think this company they just bought has the potential to out-do Intel, in this one specialized field at least.
Oh come on now Intel has never been successful in the embedded world. It is not something the focus on.
ARM on the other hand is heavily into the embedded world, it is almost impossible to build a cell phone without at least one ARM chip in the thing. Even iPhone is rumored to have several ARM chips in it besides the main processor.
IBM first used someone else's CPU (Intel 8088) in its IBM PC. Having a common CPU architecture was a catalyst for the nacent PC market. Sony and Philips both used their own ICs in many of their systems. Ultimately both saw the wisdom in using openly available, commodity-priced chips.
In the cell phone space, Nokia and Ericsson both started out using their own proprietary IC designs. Once TI and others offered cell-phone chipset solutions, no company could afford to continue down that proprietary IC road.
If Apple goes too far down that road, it will not only open the door to Samsung and more traditional competitors but to upstarts like HTC, too. HTC's "diamond touch" is built using off-the-shelf ICs. HTC's contribution is functional choice, user interface issues, and style.
Intel can push all they want but to put ATOM into a cell phone would so disadvantage a manufacture as to make them totally uncompetitive. That is not a negative on ATOM though, as I see it as an excellent effort by Intel and an awesome chip in the right sort of devices.
The right sort of devices being hardware you would want desktop style OS'es on. Sorry but I simply don't want or need a desktop OS on my cell phone nor on any mini tablet type device.
Intel have a different vision for Atom than you have. They have stated in the past that Moorestown(the next Atom chip) is the chip that the iPhone would want to have if it were available.
Those are all good points and I'm sure this was a major factor in Apple's decision to make their big switch to Intel X86 CPUs. If PowerPC couldn't create a sustainable advantage then it really became a liability for Apple. That shouldn't mean that any proprietary approach is wrong. Heck, OS X by itself is a proprietary approach toward the computer market. Apple must feel that they are able to create a sustainable advantage in handheld devices by going the route of proprietary chips. Going it alone in handheld hardware is not nearly the gamble that it was in computers since there really isn't any true standard for a chipset in handhelds anyway.
Interesting point. But unlike cell phones, smartphones makers (Apple in particular) are courting independent developers. It's like the PC market in its infancy. Software developers want to have the biggest possible market. Anything Apple does that diminishes that potential in the minds of developers puts it at a disadvantage. You are correct about OS X and Windows. Both are proprietary operating systems. Irrespective of their relative merits, Windows and OS X have far different installed bases. Developers, by and large, will put their efforts, first, into a Windows product. An OS X version is purely opportunistic. Users stick with Windows to preserve their software equity. If OS X was written to run on a "PC," it would only give Windows a run for its money only if file compatibilities were preserved.
What could Apple possibly do on it's own that the larger chip makers couldn't duplicate? My problem now is that my brain might be too weak to understand what Apple has in mind. Ultra-low power processors is one possibility. Attempting to fit whole iPhone and iPod circuits on a single chip? Apple really seems to be trying to leave the smartphone industry in the dust. RIM keeps saying they have nothing to worry about from the iPhone. RIM's people must have balls of steel.
Yes to attempting to fit a whole iPhone onto a single chip, this is exactly what they are talking about. What processor they would be using really doesn't matter, the smart money is on an ARM core though.
Some people seem to be thinking that this is a far fetched idea, that is a SoC smart cell phone. It really isn't if you look at what the embedded hardware manufactures have been up to the last few years. Now don't think that a SoC leaves the motherboard few of other chips, what it does do is provide for lots of room for other chips such as Flash memory. You still would have a bit of external circuitry for RF and such too.
The reality is this with an ARM core, you could put several onto a chip along with specialized processors and other electronics saving substantial board space and a vast amount of power. At a small enough process node you will likely be able to stuff RAM on the chip for the GPU and maybe even main memory. The more that goes on board the more you save in space and power. Of course each process size has limits on what is economical, but at 65 nm and smaller an amazing number of transistors can go on a chip, those transistors add up to a lot of function blocks.
Intel have a different vision for Atom than you have. They have stated in the past that Moorestown(the next Atom chip) is the chip that the iPhone would want to have if it were available.
It still comes down to this: what Intel wants make no difference at all, especially if the difference in power usage and the ability to do custom SoC remains the same. There is not need for OS compatibility on a cell phone or even for that matter some of the tablets I've imagined.
ARM's strengths are many and one of those is the ability to morph into all sorts of custom implementations. ARM cores are found all over the place and frankly I just haven't seen convincing evidence that Intel will be doing a bunch of custom implementations to meet different manufactures needs. They have talked about it from what I hear, but that is far from having the support in place to service the needs of a thousand different manufactures.
It still comes down to this: what Intel wants make no difference at all, especially if the difference in power usage and the ability to do custom SoC remains the same. There is not need for OS compatibility on a cell phone or even for that matter some of the tablets I've imagined.
ARM's strengths are many and one of those is the ability to morph into all sorts of custom implementations. ARM cores are found all over the place and frankly I just haven't seen convincing evidence that Intel will be doing a bunch of custom implementations to meet different manufactures needs. They have talked about it from what I hear, but that is far from having the support in place to service the needs of a thousand different manufactures.
Dave
Look I'm not going to argue whether Atom will be successful in the smart phone space but its clear that Intel are positioning it for that market. This therefore gives me some reason to doubt that Intel will fab chips that will be direct competitors for one of their products,
[...] custom circuitry to let the Apple gear do cool stuff that's prohibitively hard/costly to clone.
As for what that cool new stuff is, that's where the juicy speculation begins...
It would seem logical that they try to build some kind of hardware acceleration.
Think about it:
Apple controls the OS and the hardware it runs on. How about integrating them even further?
Beyond a general purpose CPU or a CPU-optimised OS?
They know exactly what functions their 'Core' library needs to support, how about implementing the most important ones directly in hardware?
I wouldn't be surprised if the recent announcement of OpenCL has something to do with it: build a chip that is neither CPU nor GPU - yet both at the same time.
Add to that hardware accelerated functions most often used by the core OS libraries and you get a really fast machine that can balance its focus on either graphic intensive or computing intensive tasks shifting freely between them, executing complex OS functions efficiently and fast all in hardware.
Interesting point. But unlike cell phones, smartphones makers (Apple in particular) are courting independent developers. It's like the PC market in its infancy. Software developers want to have the biggest possible market. Anything Apple does that diminishes that potential in the minds of developers puts it at a disadvantage. You are correct about OS X and Windows. Both are proprietary operating systems. Irrespective of their relative merits, Windows and OS X have far different installed bases. Developers, by and large, will put their efforts, first, into a Windows product. An OS X version is purely opportunistic. Users stick with Windows to preserve their software equity. If OS X was written to run on a "PC," it would only give Windows a run for its money only if file compatibilities were preserved.
And those are interesting points, too. What we don't know yet is how different the handheld software market is from the traditional computer market. Certainly we know that the cell phone market dwarfs about everything else. What portion of that market will ultimately become smart phones (and associated handheld "devices") and consequently a target for software developers is hard to say. Apart from potential size, the market itself will surely be different but, again, by how much? Obviously, there won't be anything like Adobe and Microsoft monolithic software that comes on DVDs. Those things take years to develop and so naturally will gravitate toward Windows to justify the development costs.
If Apple's handheld business becomes big enough and developing software for it becomes easy and quick enough (we know the software "size" as in megabytes will be tiny compared to big computer software) then it may be attractive to many developers even if there's no potential to port it to other handhelds. Impossible to know right now but it will be fun to watch.
You're right that I don't much about what I'm saying. But considering how complex the iphone is getting together with the big push towards a unified/cleaner platform, I just think that they are eying up an x86 derivative.
Quote:
Originally Posted by wizard69
*SNIP* ARM's strengths are many and one of those is the ability to morph into all sorts of custom implementations. ARM cores are found all over the place and frankly I just haven't seen convincing evidence that Intel will be doing a bunch of custom implementations to meet different manufactures needs. They have talked about it from what I hear, but that is far from having the support in place to service the needs of a thousand different manufactures. Dave
Quote:
Originally Posted by backtomac
I wonder how a 45nm RISC/ARM PA Semi custom chip would compare to a 32nm Atom. I know its a hypothetical question but someone at Apple surely has looked at this. I'm guessing that the RISC/ARM blows it away or Apple wouldn't be going this route.
After seeing the disappointing performance of current 1.8ghz Atom/Silverthorne, I'm even less convinced that Intel will be able to compete with ARM, even when the 45nm Morrestown SoC is ready. I also don't think it's necessarily a lack of engineering prowess, much of it probably has to do with the limitations of the x86 architecture in regards to small, lower power, highly-efficient embedded processors. ARM *already* has incredible chip cores ready to go, with their Cortext-A8 and out-of-order Cortex-A9 availble in multi-core implementations, running up to 1.0Ghz. And even these chips use an order of magnitude less power than Intel's Atom. ARM also has some great new technology for SIMD/vector acceleration called NEMO, in addition to the optional FPU.
I also think it's a great idea to have PA SEMI inhouse and integrating these fabulous () ARM cores into next-gen SoCs. I'm sure it will allow Apple to further customize them to their needs.
Quote:
Originally Posted by retroneo
Awesome now hopefully they PA Semi will be the first to release a Cortex-A9 quad core system on a chip. The Cortex-A9 gives today's desktop levels of performance in the 250mW class... Amazing.
ARM has finished this outstanding core a while ago and is waiting for licensees to put it in their systems-on-a-chip. Hopefully licensees like Apple.
Haha, I'm not sure if we'll see a quad-core Cortex-A9 iPhone anytime soon, but heres to hoping!
[QUOTE=Wiggin;1263890]And there's also the issue of make powerful but energy efficient graphics processors. If Apple is going to expand the "touch" line to larger tablets/slate handheld computers they are going to need more powerful graphics processsing to run the larger screens. Are the current chips used in the iPhone/iPod touch scalable to a screen 2-3 times larger?
the newer ARM cores (compared to whats in the iPhone) are available with powerful graphics processors built-in, in addition to 3rd party IP like the PowerVR from Imagination. I believe the current iPhone uses the PowerVR MBX for graphics. The newer PowerVR "SGX" is crazy fast, with full-on pixel, vertex, and geometry shaders. It's compatible with OpenGl 2.0 and Shader Model 4.1. It could easily handle a larger screen and higher resolutions.. Besides, if they had to, they could always call on Nvidia to get some of their next-gen mobile IP.
Quote:
Originally Posted by IQ78
Q. What would Apple lose if they licensed their OS to other PC makers? A. Hardware market share. However, what if Apple did the following *snip* So that OSX ran way fast on Apple specific hardware. Could Apple then license their OS to Dell, HP, etc.? Basically, get the OS into a lot more people's hands, and then let it be clear that OSX best hardware platform is Apple's via their proprietary enhanced chip sets that is fully supported in OSX? This way they could actually increase hardware share rather than lose it, and increase OS marketshare through the roof.
There are plenty of threads regarding this issue that contain better information than I can provide, however it seems to me that that concept would be a losing proposition. One of the primary advantages of purchasing a Mac computer is the stability/compatibility/performance, and that is a direct result of them producing both the hardware and the software.
Also, in regards to your idea of making OSX available to PCs while maintaining a performance edge on Apple hardware, you have to remember that *MOST* people are cheap, and they want to spend the absolute lowest amount of money possible. There is no doubt in mind that that would immediately halt much of the market growth of Apple's Mac hardware.
Quote:
Originally Posted by coolfactor
Proof that OS X will remain universal, rather than going Intel-only.
Quote:
Originally Posted by Hudson1
I'm quite the ignorant one in this subject (someone will step in to help me, I'm sure) but isn't PA Semi technology essentially PowerPC? There must be a motive for Apple doing these three things: *snip*
As others have covered, PA SEMI's team has expertise on many different architectures. As it is now, It's very unlikely POWER will play any part in Apple's future.
I may very well be wrong but I don't think Intel will want to fab ARM/RISC chips for Apple. Atom is an important product for Intel and its success in the market is far from assured. I suspect that if it does not gain a foot hold into the smart phone market it will not meet Intel's goals for the product. I think the smart phone, handheld, market is one that Intel covets and currently has no product for.
In two years(roughly,IIRC), Intel will have 32 nm Atom chips. These will go into devices(Android phones perhaps) that will be direct competitors to the iPhone. If Intel are fabbing Atoms at 32nm and ARM chips are at 45nm that may keep Atom competitive enough performance wise to give Atom chips a chance in the market place. If Intel are fabbing Apple's custom ARM/RISC chips at 32nm it likely will be no contest and the performance of iPhones will be so much better than Atom powered smart phones will be at a severe disadvantage.
It looks like Atom vs ARM/RISC is only a fair fight if Atom is one node ahead in the fab process. To me it looks like the smart phone battle is going to be Atom at 32nm vs ARM at 65 or 45 nm.
What makes you think it would be an ARM processor? And X86 has essentially been more of a RISC than a CISC processor for over ten years now. If your trying to say PPC or ARM I don't think Apple intends to design using ARM, or PPC. Apple sold ARM and just abandoned PPC. Just because this company was experimenting technologies on an inexpensive "fabless" ARM PPC, does in no way mean that Apple is going to implement those experimental findings on that same architecture. I just don't see them going back to either of those technologies that they just dusted.
If anything I could see Apple going to IBM with this to design a SOC - Cell processor implementing technologies PA Semi had success with and designing new processor, but in no way do I see one company, out of hundreds that are going to use intel, that is going to horde their design to themselves scaring intel. Intel doesn't need the iPhone, or Apple's touchscreen to be successful because hundreds are going to try to imitate Apple, or be better, and that is what will semiconductors for them. If they Fab for Apple that is just one more source of revenue.
What makes you think it would be an ARM processor? And X86 has essentially been more of a RISC than a CISC processor for over ten years now. If your trying to say PPC or ARM I don't think Apple intends to design using ARM, or PPC. Apple sold ARM and just abandoned PPC. Just because this company was experimenting technologies on an inexpensive "fabless" ARM PPC, does in no way mean that Apple is going to implement those experimental findings on that same architecture. I just don't see them going back to either of those technologies that they just dusted.
If anything I could see Apple going to IBM with this to design a SOC - Cell processor implementing technologies PA Semi had success with and designing new processor, but in no way do I see one company, out of hundreds that are going to use intel, that is going to horde their design to themselves scaring intel. Intel doesn't need the iPhone, or Apple's touchscreen to be successful because hundreds are going to try to imitate Apple, or be better, and that is what will semiconductors for them. If they Fab for Apple that is just one more source of revenue.
Well as I understand it, PA Semi's expertise is in RISC chip design. Whether that would extend to x86 chip design, I don't know.
I do think Intel covets iPhone chip business. They may not need it to stay in business but that doesn't mean they don't covet it. Remember Munster thinks Apple will sell 45 million iPhones next year. Even if Apple only do half that, as many analysts are predicting, that's still a lot of chip business. Especially considering that its an area where they currently are getting no business.
The smart phone market is a growing area and I bet Intel want a big piece of the action.
I don't understand what the debate is about. In the mobile market power consumption rules. Minimizing power consumption is done by system-on-chip designs which deliver precisely what the product needs, and advanced silicon design techniques. PA Semi has the expertise to do sophisticated low power designs. They will license components designs they need (ARM cores, 3G cell technology, GPS technology, 802.11n technology, accelerometers, and so forth), rework them as necessary to apply their magical expertise at power reduction. The result is a design they can take to a fab that has the process tech they need, and order 10+ million of them which is enough to get volume discounts and make the whole effort worthwhile.
I'm surprised nobody has pointed out the ArtX guys acquired years ago who had the kind of GPU experience that these days would integrate nicely into SoC designs for mobile chips...
BTW, while x86 has essentially been RISC internally for a decade that only works on high-end processors where this expensive front-end / back-end design gets good mileage. On a compact design for a mobile core it doesn't work too well. The ARM has a serious advantage in the mobile space and Atom can only compete if its on a better process, which it won't be for long.
This would be a good thing IMO. Software compatibility is necessary, hardware compatibility leads to the lowest common denominator usually.
I would ignore the statement that PA Semi is going to be designing just for the mobile products also. Apple has always used custom chips on the MB of its computers and has already announced plans in the area of using built in hardware accelerators and decoders like for the H.264 standard etc.and I would expect them to continue.
Differentiation and customisation is what it's about and it will allow Apple to always be faster than any clone machine out there.
Very smart post. I'd expect Snow Leopard to have some custom chips under the hood also as you mentioned.
1) What chip architecture will Apple use for the iPhone? Recent actions by Apple, the PA Semi purchase, and Job's comment that Apple will design the chips for iPods and iPhones in house have led some to believe that Apple will go with RISC/ARM chips. However, since the debut of Atom many have speculated that Apple will eventually move to an x86 chip in the iPhone. This was fueled by Intel's actions at their 2007 developers conference where they showed prototype devices running on the Atom chip that bear a striking resemblance to the iPhone. As I've said in other posts, I feel Intel covet the smart phone chip business in general and the iPhone in particular.
2) How will Atom compare to ARM/RISC? This sort of seems like the whole RISC vs. CISC battle on the desktop all over again but maybe it'll be different this time. It's not really a battle now because Intel haven't pushed Atom in the smart phone market but that will change with Mooretown and SoC chip designs. I think the battle gets interesting when Atom gets fabbed at 32nm. I expect that Atom will always be ahead when it comes to fab process. That's why I asked the hypothetical question 'how will a 45nm ARM compare to a 32nm Atom?' IMO that's whewre the battle really begins.
Comments
Good points.
Let me rephrase my thinking here. Apple need to find a good partner to do the fabbing of these chips to power iPods and iPhones. In the case of the iPhone they need a partner willing to keep up with Intel in the smaller fabbing process. They can't afford to partner with someone who stops at 45nm while Intel keep pushing things smaller. This does worry me a little. Not many are able to keep up with Intel in the chip fab business. At this point Apple and Intel are now competitors because Atom chips at 35nm will likely start showing up in competing smart phones.
I wasn't denying that Apple would need a cooperative company with a fab facility. But being that there is one Apple, hundreds of other companies to buy intels chips, and Apple will be keeping their chips all to themselves, I don't see intel as denying Apple the fab. Intel will still make money from it, and Apple already is working with, and has cooperation from inside intel to produce other products. I think with all the existing business intel and Apple do together I don't think this will put a strain on either one of them or their relationship. It's a huge technology world, and intel is inside everything one way or another. That's their business. Being a Fab facility is also their business.
I wonder how a 45nm RISC/ARM PA Semi custom chip would compare to a 32nm Atom. I know its a hypothetical question but someone at Apple surely has looked at this. I'm guessing that the RISC/ARM blows it away or Apple wouldn't be going this route.
ARM does amazingly well with respect to power usage even at 65nm. I don't think Intel will ever be able to catch up with that as there is just a huge advantage with respect to the number of transistors ARM needs to implement. On top of that ARM has been in the embedded world for years and building SoC is well understood for the hardware.
I don't really expect any surprises here other than a highly integrated SoC that allows for a minimal of external chips.
Dave
Not in its current iteration. However when its fab moves to 32nm everything I've read suggests that Intel will push hard for its adoption in smart phones.
I'm not sure Intel will want to fab a chip that competes with one of their products even though it would only be used by Apple.
Intel can push all they want but to put ATOM into a cell phone would so disadvantage a manufacture as to make them totally uncompetitive. That is not a negative on ATOM though, as I see it as an excellent effort by Intel and an awesome chip in the right sort of devices.
The right sort of devices being hardware you would want desktop style OS'es on. Sorry but I simply don't want or need a desktop OS on my cell phone nor on any mini tablet type device.
As an interesting aside I could also see ATOM going into low cost desktop multi processor systems. Think a Mini with 8 of these cores. With Apples move to enhance OS/X for multiprocessing and system resource usage this could actually end up being a very nice machine for many people. That would put 8 real thread or 16 if you look at SMT as being viable. The big thing here is the machine wouldn't even draw the power the current Mini does for the most part.
I find it funny that people see Snow leopard as Apple trying to optimize for the Mac Pro class machine. With ATOM such an OS re-factoring will be leveraged on the lowest cost machines.
Dave
I wasn't denying that Apple would need a cooperative company with a fab facility. But being that there is one Apple, hundreds of other companies to buy intels chips, and Apple will be keeping their chips all to themselves, I don't see intel as denying Apple the fab. Intel will still make money from it, and Apple already is working with, and has cooperation from inside intel to produce other products. I think with all the existing business intel and Apple do together I don't think this will put a strain on either one of them or their relationship. It's a huge technology world, and intel is inside everything one way or another. That's their business. Being a Fab facility is also their business.
I may very well be wrong but I don't think Intel will want to fab ARM/RISC chips for Apple. Atom is an important product for Intel and its success in the market is far from assured. I suspect that if it does not gain a foot hold into the smart phone market it will not meet Intel's goals for the product. I think the smart phone, handheld, market is one that Intel covets and currently has no product for.
In two years(roughly,IIRC), Intel will have 32 nm Atom chips. These will go into devices(Android phones perhaps) that will be direct competitors to the iPhone. If Intel are fabbing Atoms at 32nm and ARM chips are at 45nm that may keep Atom competitive enough performance wise to give Atom chips a chance in the market place. If Intel are fabbing Apple's custom ARM/RISC chips at 32nm it likely will be no contest and the performance of iPhones will be so much better than Atom powered smart phones will be at a severe disadvantage.
It looks like Atom vs ARM/RISC is only a fair fight if Atom is one node ahead in the fab process. To me it looks like the smart phone battle is going to be Atom at 32nm vs ARM at 65 or 45 nm.
HP seems to have the visible 'touch', yeah!
it reaches in, and grabs right hold of TouchSmart!
HP seems to have the visible 'touch', yeah!
it takes control and slowly tears you apart
HP is still going after touch computers in a big way.
Personally, I think this kind of computer could become very tiresome to interact with.
Making your own chip is one way to differentiate yourself from the competition, but it all comes down to who has the talent. If Intel keep vacuuming up the best talent then the only way Apple will differentiate themselves is in a negative way. Of course they must think this company they just bought has the potential to out-do Intel, in this one specialized field at least.
Oh come on now Intel has never been successful in the embedded world. It is not something the focus on.
ARM on the other hand is heavily into the embedded world, it is almost impossible to build a cell phone without at least one ARM chip in the thing. Even iPhone is rumored to have several ARM chips in it besides the main processor.
Dave
Please give a few examples.
IBM first used someone else's CPU (Intel 8088) in its IBM PC. Having a common CPU architecture was a catalyst for the nacent PC market. Sony and Philips both used their own ICs in many of their systems. Ultimately both saw the wisdom in using openly available, commodity-priced chips.
In the cell phone space, Nokia and Ericsson both started out using their own proprietary IC designs. Once TI and others offered cell-phone chipset solutions, no company could afford to continue down that proprietary IC road.
If Apple goes too far down that road, it will not only open the door to Samsung and more traditional competitors but to upstarts like HTC, too. HTC's "diamond touch" is built using off-the-shelf ICs. HTC's contribution is functional choice, user interface issues, and style.
Intel can push all they want but to put ATOM into a cell phone would so disadvantage a manufacture as to make them totally uncompetitive. That is not a negative on ATOM though, as I see it as an excellent effort by Intel and an awesome chip in the right sort of devices.
The right sort of devices being hardware you would want desktop style OS'es on. Sorry but I simply don't want or need a desktop OS on my cell phone nor on any mini tablet type device.
Intel have a different vision for Atom than you have. They have stated in the past that Moorestown(the next Atom chip) is the chip that the iPhone would want to have if it were available.
They even showed pictures of smart phones with Moorestown chips in them that bear a striking resemblance to the iPhone. I have no doubt they will try to position Atom for the smart phone market. Whether they are successful or not is another story.
Those are all good points and I'm sure this was a major factor in Apple's decision to make their big switch to Intel X86 CPUs. If PowerPC couldn't create a sustainable advantage then it really became a liability for Apple. That shouldn't mean that any proprietary approach is wrong. Heck, OS X by itself is a proprietary approach toward the computer market. Apple must feel that they are able to create a sustainable advantage in handheld devices by going the route of proprietary chips. Going it alone in handheld hardware is not nearly the gamble that it was in computers since there really isn't any true standard for a chipset in handhelds anyway.
Interesting point. But unlike cell phones, smartphones makers (Apple in particular) are courting independent developers. It's like the PC market in its infancy. Software developers want to have the biggest possible market. Anything Apple does that diminishes that potential in the minds of developers puts it at a disadvantage. You are correct about OS X and Windows. Both are proprietary operating systems. Irrespective of their relative merits, Windows and OS X have far different installed bases. Developers, by and large, will put their efforts, first, into a Windows product. An OS X version is purely opportunistic. Users stick with Windows to preserve their software equity. If OS X was written to run on a "PC," it would only give Windows a run for its money only if file compatibilities were preserved.
What could Apple possibly do on it's own that the larger chip makers couldn't duplicate? My problem now is that my brain might be too weak to understand what Apple has in mind. Ultra-low power processors is one possibility. Attempting to fit whole iPhone and iPod circuits on a single chip? Apple really seems to be trying to leave the smartphone industry in the dust. RIM keeps saying they have nothing to worry about from the iPhone. RIM's people must have balls of steel.
Yes to attempting to fit a whole iPhone onto a single chip, this is exactly what they are talking about. What processor they would be using really doesn't matter, the smart money is on an ARM core though.
Some people seem to be thinking that this is a far fetched idea, that is a SoC smart cell phone. It really isn't if you look at what the embedded hardware manufactures have been up to the last few years. Now don't think that a SoC leaves the motherboard few of other chips, what it does do is provide for lots of room for other chips such as Flash memory. You still would have a bit of external circuitry for RF and such too.
The reality is this with an ARM core, you could put several onto a chip along with specialized processors and other electronics saving substantial board space and a vast amount of power. At a small enough process node you will likely be able to stuff RAM on the chip for the GPU and maybe even main memory. The more that goes on board the more you save in space and power. Of course each process size has limits on what is economical, but at 65 nm and smaller an amazing number of transistors can go on a chip, those transistors add up to a lot of function blocks.
Dave
Intel have a different vision for Atom than you have. They have stated in the past that Moorestown(the next Atom chip) is the chip that the iPhone would want to have if it were available.
They even showed pictures of smart phones with Moorestown chips in them that bear a striking resemblance to the iPhone. I have no doubt they will try to position Atom for the smart phone market. Whether they are successful or not is another story.
It still comes down to this: what Intel wants make no difference at all, especially if the difference in power usage and the ability to do custom SoC remains the same. There is not need for OS compatibility on a cell phone or even for that matter some of the tablets I've imagined.
ARM's strengths are many and one of those is the ability to morph into all sorts of custom implementations. ARM cores are found all over the place and frankly I just haven't seen convincing evidence that Intel will be doing a bunch of custom implementations to meet different manufactures needs. They have talked about it from what I hear, but that is far from having the support in place to service the needs of a thousand different manufactures.
Dave
It still comes down to this: what Intel wants make no difference at all, especially if the difference in power usage and the ability to do custom SoC remains the same. There is not need for OS compatibility on a cell phone or even for that matter some of the tablets I've imagined.
ARM's strengths are many and one of those is the ability to morph into all sorts of custom implementations. ARM cores are found all over the place and frankly I just haven't seen convincing evidence that Intel will be doing a bunch of custom implementations to meet different manufactures needs. They have talked about it from what I hear, but that is far from having the support in place to service the needs of a thousand different manufactures.
Dave
Look I'm not going to argue whether Atom will be successful in the smart phone space but its clear that Intel are positioning it for that market. This therefore gives me some reason to doubt that Intel will fab chips that will be direct competitors for one of their products,
[...] custom circuitry to let the Apple gear do cool stuff that's prohibitively hard/costly to clone.
As for what that cool new stuff is, that's where the juicy speculation begins...
It would seem logical that they try to build some kind of hardware acceleration.
Think about it:
Apple controls the OS and the hardware it runs on. How about integrating them even further?
Beyond a general purpose CPU or a CPU-optimised OS?
They know exactly what functions their 'Core' library needs to support, how about implementing the most important ones directly in hardware?
I wouldn't be surprised if the recent announcement of OpenCL has something to do with it: build a chip that is neither CPU nor GPU - yet both at the same time.
Add to that hardware accelerated functions most often used by the core OS libraries and you get a really fast machine that can balance its focus on either graphic intensive or computing intensive tasks shifting freely between them, executing complex OS functions efficiently and fast all in hardware.
Interesting point. But unlike cell phones, smartphones makers (Apple in particular) are courting independent developers. It's like the PC market in its infancy. Software developers want to have the biggest possible market. Anything Apple does that diminishes that potential in the minds of developers puts it at a disadvantage. You are correct about OS X and Windows. Both are proprietary operating systems. Irrespective of their relative merits, Windows and OS X have far different installed bases. Developers, by and large, will put their efforts, first, into a Windows product. An OS X version is purely opportunistic. Users stick with Windows to preserve their software equity. If OS X was written to run on a "PC," it would only give Windows a run for its money only if file compatibilities were preserved.
And those are interesting points, too. What we don't know yet is how different the handheld software market is from the traditional computer market. Certainly we know that the cell phone market dwarfs about everything else. What portion of that market will ultimately become smart phones (and associated handheld "devices") and consequently a target for software developers is hard to say. Apart from potential size, the market itself will surely be different but, again, by how much? Obviously, there won't be anything like Adobe and Microsoft monolithic software that comes on DVDs. Those things take years to develop and so naturally will gravitate toward Windows to justify the development costs.
If Apple's handheld business becomes big enough and developing software for it becomes easy and quick enough (we know the software "size" as in megabytes will be tiny compared to big computer software) then it may be attractive to many developers even if there's no potential to port it to other handhelds. Impossible to know right now but it will be fun to watch.
You're right that I don't much about what I'm saying. But considering how complex the iphone is getting together with the big push towards a unified/cleaner platform, I just think that they are eying up an x86 derivative.
*SNIP* ARM's strengths are many and one of those is the ability to morph into all sorts of custom implementations. ARM cores are found all over the place and frankly I just haven't seen convincing evidence that Intel will be doing a bunch of custom implementations to meet different manufactures needs. They have talked about it from what I hear, but that is far from having the support in place to service the needs of a thousand different manufactures. Dave
I wonder how a 45nm RISC/ARM PA Semi custom chip would compare to a 32nm Atom. I know its a hypothetical question but someone at Apple surely has looked at this. I'm guessing that the RISC/ARM blows it away or Apple wouldn't be going this route.
After seeing the disappointing performance of current 1.8ghz Atom/Silverthorne, I'm even less convinced that Intel will be able to compete with ARM, even when the 45nm Morrestown SoC is ready. I also don't think it's necessarily a lack of engineering prowess, much of it probably has to do with the limitations of the x86 architecture in regards to small, lower power, highly-efficient embedded processors. ARM *already* has incredible chip cores ready to go, with their Cortext-A8 and out-of-order Cortex-A9 availble in multi-core implementations, running up to 1.0Ghz. And even these chips use an order of magnitude less power than Intel's Atom. ARM also has some great new technology for SIMD/vector acceleration called NEMO, in addition to the optional FPU.
I also think it's a great idea to have PA SEMI inhouse and integrating these fabulous (
Awesome now hopefully they PA Semi will be the first to release a Cortex-A9 quad core system on a chip. The Cortex-A9 gives today's desktop levels of performance in the 250mW class... Amazing.
ARM has finished this outstanding core a while ago and is waiting for licensees to put it in their systems-on-a-chip. Hopefully licensees like Apple.
Haha, I'm not sure if we'll see a quad-core Cortex-A9 iPhone anytime soon, but heres to hoping!
[QUOTE=Wiggin;1263890]And there's also the issue of make powerful but energy efficient graphics processors. If Apple is going to expand the "touch" line to larger tablets/slate handheld computers they are going to need more powerful graphics processsing to run the larger screens. Are the current chips used in the iPhone/iPod touch scalable to a screen 2-3 times larger?
the newer ARM cores (compared to whats in the iPhone) are available with powerful graphics processors built-in, in addition to 3rd party IP like the PowerVR from Imagination. I believe the current iPhone uses the PowerVR MBX for graphics. The newer PowerVR "SGX" is crazy fast, with full-on pixel, vertex, and geometry shaders. It's compatible with OpenGl 2.0 and Shader Model 4.1. It could easily handle a larger screen and higher resolutions.. Besides, if they had to, they could always call on Nvidia to get some of their next-gen mobile IP.
Q. What would Apple lose if they licensed their OS to other PC makers? A. Hardware market share. However, what if Apple did the following *snip* So that OSX ran way fast on Apple specific hardware. Could Apple then license their OS to Dell, HP, etc.? Basically, get the OS into a lot more people's hands, and then let it be clear that OSX best hardware platform is Apple's via their proprietary enhanced chip sets that is fully supported in OSX? This way they could actually increase hardware share rather than lose it, and increase OS marketshare through the roof.
There are plenty of threads regarding this issue that contain better information than I can provide, however it seems to me that that concept would be a losing proposition. One of the primary advantages of purchasing a Mac computer is the stability/compatibility/performance, and that is a direct result of them producing both the hardware and the software.
Also, in regards to your idea of making OSX available to PCs while maintaining a performance edge on Apple hardware, you have to remember that *MOST* people are cheap, and they want to spend the absolute lowest amount of money possible. There is no doubt in mind that that would immediately halt much of the market growth of Apple's Mac hardware.
Proof that OS X will remain universal, rather than going Intel-only.
I'm quite the ignorant one in this subject (someone will step in to help me, I'm sure) but isn't PA Semi technology essentially PowerPC? There must be a motive for Apple doing these three things: *snip*
As others have covered, PA SEMI's team has expertise on many different architectures. As it is now, It's very unlikely POWER will play any part in Apple's future.
I may very well be wrong but I don't think Intel will want to fab ARM/RISC chips for Apple. Atom is an important product for Intel and its success in the market is far from assured. I suspect that if it does not gain a foot hold into the smart phone market it will not meet Intel's goals for the product. I think the smart phone, handheld, market is one that Intel covets and currently has no product for.
In two years(roughly,IIRC), Intel will have 32 nm Atom chips. These will go into devices(Android phones perhaps) that will be direct competitors to the iPhone. If Intel are fabbing Atoms at 32nm and ARM chips are at 45nm that may keep Atom competitive enough performance wise to give Atom chips a chance in the market place. If Intel are fabbing Apple's custom ARM/RISC chips at 32nm it likely will be no contest and the performance of iPhones will be so much better than Atom powered smart phones will be at a severe disadvantage.
It looks like Atom vs ARM/RISC is only a fair fight if Atom is one node ahead in the fab process. To me it looks like the smart phone battle is going to be Atom at 32nm vs ARM at 65 or 45 nm.
What makes you think it would be an ARM processor? And X86 has essentially been more of a RISC than a CISC processor for over ten years now. If your trying to say PPC or ARM I don't think Apple intends to design using ARM, or PPC. Apple sold ARM and just abandoned PPC. Just because this company was experimenting technologies on an inexpensive "fabless" ARM PPC, does in no way mean that Apple is going to implement those experimental findings on that same architecture. I just don't see them going back to either of those technologies that they just dusted.
If anything I could see Apple going to IBM with this to design a SOC - Cell processor implementing technologies PA Semi had success with and designing new processor, but in no way do I see one company, out of hundreds that are going to use intel, that is going to horde their design to themselves scaring intel. Intel doesn't need the iPhone, or Apple's touchscreen to be successful because hundreds are going to try to imitate Apple, or be better, and that is what will semiconductors for them. If they Fab for Apple that is just one more source of revenue.
What makes you think it would be an ARM processor? And X86 has essentially been more of a RISC than a CISC processor for over ten years now. If your trying to say PPC or ARM I don't think Apple intends to design using ARM, or PPC. Apple sold ARM and just abandoned PPC. Just because this company was experimenting technologies on an inexpensive "fabless" ARM PPC, does in no way mean that Apple is going to implement those experimental findings on that same architecture. I just don't see them going back to either of those technologies that they just dusted.
If anything I could see Apple going to IBM with this to design a SOC - Cell processor implementing technologies PA Semi had success with and designing new processor, but in no way do I see one company, out of hundreds that are going to use intel, that is going to horde their design to themselves scaring intel. Intel doesn't need the iPhone, or Apple's touchscreen to be successful because hundreds are going to try to imitate Apple, or be better, and that is what will semiconductors for them. If they Fab for Apple that is just one more source of revenue.
Well as I understand it, PA Semi's expertise is in RISC chip design. Whether that would extend to x86 chip design, I don't know.
I do think Intel covets iPhone chip business. They may not need it to stay in business but that doesn't mean they don't covet it. Remember Munster thinks Apple will sell 45 million iPhones next year. Even if Apple only do half that, as many analysts are predicting, that's still a lot of chip business. Especially considering that its an area where they currently are getting no business.
The smart phone market is a growing area and I bet Intel want a big piece of the action.
I'm surprised nobody has pointed out the ArtX guys acquired years ago who had the kind of GPU experience that these days would integrate nicely into SoC designs for mobile chips...
BTW, while x86 has essentially been RISC internally for a decade that only works on high-end processors where this expensive front-end / back-end design gets good mileage. On a compact design for a mobile core it doesn't work too well. The ARM has a serious advantage in the mobile space and Atom can only compete if its on a better process, which it won't be for long.
This would be a good thing IMO. Software compatibility is necessary, hardware compatibility leads to the lowest common denominator usually.
I would ignore the statement that PA Semi is going to be designing just for the mobile products also. Apple has always used custom chips on the MB of its computers and has already announced plans in the area of using built in hardware accelerators and decoders like for the H.264 standard etc.and I would expect them to continue.
Differentiation and customisation is what it's about and it will allow Apple to always be faster than any clone machine out there.
Very smart post. I'd expect Snow Leopard to have some custom chips under the hood also as you mentioned.
I don't understand what the debate is about.
Let me explain the debate as I see it.
1) What chip architecture will Apple use for the iPhone? Recent actions by Apple, the PA Semi purchase, and Job's comment that Apple will design the chips for iPods and iPhones in house have led some to believe that Apple will go with RISC/ARM chips. However, since the debut of Atom many have speculated that Apple will eventually move to an x86 chip in the iPhone. This was fueled by Intel's actions at their 2007 developers conference where they showed prototype devices running on the Atom chip that bear a striking resemblance to the iPhone. As I've said in other posts, I feel Intel covet the smart phone chip business in general and the iPhone in particular.
2) How will Atom compare to ARM/RISC? This sort of seems like the whole RISC vs. CISC battle on the desktop all over again but maybe it'll be different this time. It's not really a battle now because Intel haven't pushed Atom in the smart phone market but that will change with Mooretown and SoC chip designs. I think the battle gets interesting when Atom gets fabbed at 32nm. I expect that Atom will always be ahead when it comes to fab process. That's why I asked the hypothetical question 'how will a 45nm ARM compare to a 32nm Atom?' IMO that's whewre the battle really begins.