Hacker Newsnew | past | comments | ask | show | jobs | submitlogin
Apple iPad biggest news is Apple A4 chip. Now they fully control supply chain (gizmodo.com)
94 points by nexneo on Jan 27, 2010 | hide | past | favorite | 80 comments


As far as I know, they aren't making screens, batteries, memory, and lots of other components, so they are yet to fully control the supply chain, but they are one step closer.

However, I wonder how the business scale of building their own processors is going to work for the. Will the A4 move into the iPhone products? If the iPad isn't a huge success, that processor is going to cost them more than having bought from a chip supplier.

What tech does the new processor really bring?


10 hours of battery life in a device half an inch thick.


The iPod/iPhone lines already do that. Now that Apple has the flexible battery technology it's no real surprise--the battery is probably nearly as big as the iPad.


10 hours of battery life playing video continuously. My iPhone (1st gen) gets maybe 2 hrs of continuous YouTube watching.


Was that actually stated? I thought they were just saying that use would last 10 hours and then made the separate statement that you could watch movies for an entire flight.


Yep. Also specifically stated: a standby mode that lasts months.


And Apple says my iPhone has 12 days of standby time. Ha. I'll wait for the reviews to believe anything about the battery life (as I do for all CE devices).


Yes. The entire flight he was referring to was from San Francisco to Tokyo. http://www.engadget.com/2010/01/27/live-from-the-apple-table... search the page for "hours"


Technically Apple says playing video, surfing the web or listening to music. It sounds like 10 hours of use to me, probably some of each. It has a 25 watt-hour battery. Nothing huge, for comparison the MacBook Air has a 40 watt-hour battery. They don't break out what the iPhone/iPod Touch have.

http://www.apple.com/ipad/specs/


In the iPad specs page you linked, footnote #4 states:

"Testing conducted by Apple in January 2010 using preproduction iPad units and software. Testing consisted of full battery discharge while performing each of the following tasks: video playback, audio playback, and Internet browsing using Wi-Fi. Video content was a repeated 2-hour 23-minute movie purchased from the iTunes Store. Audio content was a playlist of 358 unique songs, consisting of a combination of songs imported from CDs using iTunes (128-Kbps AAC encoding) and songs purchased from the iTunes Store (256-Kbps AAC encoding). Internet over Wi-Fi tests were conducted using a closed network and dedicated web and mail servers, browsing snapshot versions of 20 popular web pages, and receiving mail once an hour. All settings were default except: Wi-Fi was associated with a network; the Wi-Fi feature Ask to Join Networks and Auto-Brightness were turned off. Battery life depends on device settings, usage, and many other factors. Battery tests are conducted using specific iPad units; actual results may vary."


Processor is always a harder to produce part and there is only handful of suppliers. Not true in case of screens, batteries, may be also memory.

Why they choose their own product instead anyother? May be its better, at just 1Ghz 10" interface is super fluid. 10 hours of video on single charge. And owning supply chain doesn't only mean more profit. Now they control features and tweak chips to their own needs much more easily.


"Processor is always a harder to produce part and there is only handful of suppliers."

That's how ARM started. The designers of the Archimedes computer looked round for a powerful but cheap CPU and couldn't find one, so they decided to build their own. They had a very small team, so by necessity it was a simple design - but that ended up being its strength.


I'm not sure this is true with an arm processor. A lot of semiconductors companies are working on an arm version , and optimizations for low power.

And the arm platform is quite an open platform. I bet any successful company can work with big semi companies to tweak chips.

The biggest advantage is secrecy.If apple has some unique feature - it gives them the time to secretly do research on processor architectures for it.

But not a lot of value for features everybody are looking for. in this case open r&d is better.


There is such an overhead when dealing with an external company. Integration projects are rarely fun. I postulate that having your hardware designer report to the same guy as the chip designer makes the process much more streamlined. No waiting for contracts / bizdev, being scheduled with competing projects, et cetera. The powervr chip in A9 could be configured to support OpenGL (non-es), but the TI Omap series has it set up for OpenGL ES instead. Now, if you are apple, you could probably get TI to change that for you... but then you're talking about contracts & process management and integration once again. Wouldn't it be nice to just set directions and have it done?


ARM, MIPS, Sparc, PPC and x86/x86-64 compatible CPUs are all available from multiple suppliers.

There's still reasons, though, such as piecing together a systsem on a chip to meet specific power and feature requirements.


Apple designs and builds their own batteries now.


If they have the volume, which they probably will, more profit. This will allow them to do more R+D and eventually not give a portion of their profits to Intel for the big chips.

Chips are plenty powerful enough to run apps (for the standard consumer). Now the race is to be the most efficient chip, and since that's a fairly new race Apple should have a good chance at it.


Efficiency isn't a new race. What's happening is that the desktop market segment is learning efficiency to meet the demands of the developing performance mobile segment, as traditional mobile is learning performance to also take that segment. That's a race that has been developing in the semico designers for around eight years, give or take, but product has only been available on the store shelf for a year or so.


More accounting profit, not economic profit.

You have to earn a return on all the additional assets you have by going vertical. The Street will punish you if you just buy companies to make more "profit" but your return on assets decline.

Real reason is probably tweaking designs to their needs in a way that competitors can't access.


Yes. This is like a chicken processor who also owns a feed mill saying that he controls the supply chain. In actuality, he delivers a load of feed to somebody else's chicken farm and then picks up the chickens a few weeks later.


Fully control the supply chain? Please don't editorialize like that.


Maybe he meant "fully control the portion of the supply chain that matters".

Which I'm not sure is the case either...


Yeah, from the looks of it Apple is running some seriously beefy graphics hardware that somehow sips power. We know the CPU is in-house, but what about the GPU (which IMHO is the bigger part of the equation).


The A4 is an System-On-Chip, so no separate GPU.

From http://news.cnet.com/8301-13924_3-10442684-64.html :

  By definition, the A4 is a system-on-a-chip, or SOC, that
  integrates the main processor, graphics silicon, and other
  functions like the memory controller on one piece of silicon...


I know there are HN posting standards, but not all editorialization is bad. I find some opinions interesting and it gives you an angle to confirm or falsify when reading the article.


But the opinion can be given in a comment.


"Fully control the supply chain?" that was the question in my mind so it appeared . But I agree with your concerns.


Do they even own a plant or is production outsourced?


Apple bought P.A. Semi, a small chip manufacturer, last year, so it is technically under the Apple umbrella.


P.A. Semi are designers, not manufacturers. I expect fabrication is outsourced to an outfit like TSMC.


Samsung would be an obvious guess at the fab, given Apple's history of working with them.


P.A. Semi is fabless, meaning manufacturing is done by someone else.


As a circuit designer, I find it really surprising that Apple has rolled their own processor for the iPad. We finally confirm why Apple bought PA Semi. But it also says a lot about Apple and the processor industry:

* Apple predicts that sales will be high enough to justify building an iPad-specific processor. I assume nobody outside of Apple will get to use the A4 or derivative chips.

* Apple feels that there was no external vendor that could cost-effectively get the exact power-performance trade-off they wanted for the suite of applications that they want to optimize.

* Tempering the previous point, this may be a net loss in the case of the iPad alone in terms of pure costs versus going with an external vendor, BUT they now have much better negotiating power for all their other processor purchases from Intel, Samsung, ARM, etc. They can always threaten to roll their own. So even if they lose some money on the A4, they may be in a better position overall

* EDIT: Let me also add that despite what some people are claiming this is likely NOT an ARM, but based on PA Semi's power-efficient flavor of the Power architecture. I believe the Power architecture is in the same family as the Motorola PowerPCs that Apple formerly used and dumped for Intel because of the PowerPC's lower power-efficiency.


The price is tempting. But I don't like the mindset of bigger ipod touch (app store and the closed OS).


The first thing that comes in mind is, if it's capable of running OS X. (or any other regular OS)


No, because the chip is ARM and therefore can only run operating systems that have been ported to ARM (this does, however, include various linux distributions).


from other news reports it appears the A4 is just an Arm processor.


It would kinda have to be to run random pre-existing iPhone apps.


It could have an ARM emulator, but I doubt that's the case.


i'd love to see a vmware/parallels app for the iPad. imagine being able to run os x/windows/linux on this thing.


It'd have to do machine-code translation, and so would be slow as hell. Try running windows in qemu some time to see what I mean.


To be fair, qemu is appallingly slow. Windows in Microsoft's Virtual PC for Mac (on Power PC macs) was several orders of magnitude moe performant.

Machine-code translation can be done well, but that still doesn't mean trying to pack a desktop OS onto a mobile platform is going to be a good idea.


This makes Apple the new Commodore.

(Commodore really took off when they acquired MOS Technology. It enabled them to develop the custom silicon they needed to make the VIC, 64, and Amiga world-class game-changing machines.)


While I love Commodore, the analogy breaks down. MOS had their own fab, which allowed them to do lightning-fast turnaround. If Apple bought a fab today, you should short their stock because that is a low-margin, cash-hungry business.

The VIC-II was apparently built so quickly without simulation tools because they fabbed custom test runs of individual parts of the chip with debug logic added. This "unit-test" style development had chip spins coming out in 2 weeks.

Nowadays, you do this with simulation or an FPGA.


This isn't really new for Apple. They used to own a sizeable chunk of ARM (26% if memory serves) back in the days of Newton but ended up selling out.

Now they own an ARM licensee.


They also used to be a major player in the PowerPC line.


I agree that the A4 chip is interesting. But to me the biggest news is that Apple is supplying the iPad with 3g access and an unlocked SIM. Not that the majority of people will buy an iPad, but they are preempting what I think we'll see in the future with data connectivity, namely that we'll pay a monthly subscription for wireless connectivity the same way many people now pay for cell phones instead of landlines. We'll expect a carrier of choice and we'll expect the service to be available almost everywhere and on almost every device.


It's just the US that's behind the curve on this. I have a 2Mbit 3g connection that works anywhere for $30 a month. Which is really great since I live on a boat, so I'm connected wherever my home happens to be.


That sounds like a great lifestyle.


Me and three other guys that live on boats are actually in the process of starting a blog with life and experiences from the boat life. It's probably not HN material though.


It most certainly is. Everyone here is working, and environment's important to work. Being on a boat is quite the change.


I can't speak for everyone else, but I can say that I personally would definitely be interested in reading your blog.


Maybe I can find some nerdy angle and write a blogpost that I would feel good about submitting :-)


It's unlocked, but the radio only supports the ATT 3G frequencies in the US, so it's effectively locked in to ATT unless you want to use EDGE for everything on T-Mobile.


Keep in mind that T-Mobile's 3G frequencies are the 'standard' ones internationally. Though, IIRC Telus (Canada) and maybe Rogers (also Canada) share the same non-standard 3G frequency as AT&T.


So, what does this chip do that, say, a Snapdragon doesn't?


Custom chip makes it hard to hack with no documentation and possible encryption built in.


You may be on to something there.


It is almost certainly based on the Cortex A9 instead of the A8 that current Snapdragons use, which means multiple cores. Apple knows how to use multiple cores effectively and the ability to stop them when not using them is a nice power savings.

Snapdragons extend by adding in the wireless circuitry. I'd wager that A4 left that to other chips and extended with someone's graphic rendering engine, say Imagination Technologies?

But I think the most important reason is that it lets Apple make the design decisions that make sense for their projected use, not what some chip house thinks will be the focus three years down the road.


The Snapdragon and Cortex A8 are both implementations of the ARM V7A architecture, but are completely separate designs. The Snapdragon is Qualcomm's version and is unique to Qualcomm, whereas the Cortex A8 is licensed by ARM to third parties.

It's the same as Intel and AMD both making x86 chips; the ISA is the same but the design is different. However, ARM is more than happy to let Qualcomm design thier own.


I thought Cortex A8 was TI's implementation of ARMv7 and Snapdragon Qualcomm's: the ISAs are the same but Snapdragon can run distinct cores at distinct speeds because of disjoint caches per core, beneficial for power savings, for example.

Check http://www.arm.com/miscPDFs/14128.pdf

for more details than are simple to read.


No, the Cortex A8 hard macro was designed with TI as a lead customer, but is an ARM design.


Ah, yes. I misread the wikipedia page. similar to A8, not A8.


Generate a reality distortion field.


True. Now we can start a countdown until they move their proprietary chip into the iPhone.


It's not that important actually. iPhone is a locked system hardware and software-wise. Whatever chip it's using now or in the future hardly matters.

On the other hand, adding these in the laptops might be interesting... But I don't think they are in a hurry to switch architectures so soon.


The iPhone 3GS already uses a custom chip. It combines an ARM processor, a SIMD FPU, a GPU, and 128 MB of DRAM onto a single die. It's pretty neat.


What's the A4? Probably some ARM variant design, right?


It's ARM. Given the "snappiness" one might speculate Cortex A9.


Where is the information coming from that it's ARM? I didn't see that information on any of the live-blogging sites. Did someone mention it in the Apple presentation?

I find it odd that Apple bought PA Semi who were apparently working on a variant of the Power architecture and had them stop what they were doing and switch to implementing an ARM.


Occam's razor. ARM provides backwards compatibility with iPhone apps.


Occam's razor doesn't apply the same to the hardware equation.

If the architecture is ARM, it would mean that Apple bought PA Semi for the purpose of getting them to dump their main architecture, which was the whole reason they were created, and got them to switch over to ARM and develop, fabricate, and test the chip in the span of a about a year.

I'll place my bets on Power architecture + ARM emulation/JIT/whatever or iPhone App recompile for the iPad.


I considered emulation, but it's very hard to imagine that a PPC running an ARM emulator could be more power-efficient than the best ARM. Keep in mind that the PA Semi team has previously designed ARM, MIPS, and PPC processors; I don't think they have any particular loyalty.


Cool. Thanks for the info that they designed other architectures. I didn't know that. I guess we'll find out soon enough!


This is what I also imagined. Anand over at anandtech posted a big review of the A9 and it seems like a big step forward for ARM chips. I'm sure someone will soak the A4 in a acid bath and take some photos of the raw silicon and tell us all. (in a few months)


I wonder what the 4 stands for? Was there an A1, A2, A3?


I assume it's a pun on paper size.


That's what I thought... but it doesn't seem very likely from a US company?


Apple has a history in design, where knowing about the A4 format is not uncommon. Besides, it's not like they could have named the processor Legal, Tabloid or Letter.


Maybe they wanted it to sound more mature than it really is :)




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: