Although early
erroneous reports placed Apple's
custom A4 chip at the leading edge of ARM-based design, things are slowly falling back to earth. First, we learned that the graphics subsystem was almost certainly the
same PowerVR SGX component found in the iPhone, and now
Ars Technica reports that the actual CPU is the familiar single-core Cortex A8 also found in Apple's handset. That makes the A4 seem an awful lot like an tightened-up, overclocked iPhone 3GS chip, which makes sense, seeing as it was
actually in production in September of last year. So why the need for a custom part? The answer may well be efficiency and power savings: by cutting out extraneous Cortex A8 features and I/O that go unused in the iPad, Apple can further reduce the A4's size and energy draw -- which could be why Steve Jobs said the iPad's chips
"use hardly any power." That might not make a huge difference when tied to a large LCD in a device like the iPad, but
Ars speculates that this strategy combined with some of P.A. Semi's dynamic power optimization tech could result in a hyper-efficient chip for the iPhone somewhere down the line. That would certainly be interesting in the future -- but right now we've got the A4, and we can't wait until the end of the month to properly put this thing through its paces.
posted March 1st 2010 12:37PM
Charmin Ultra....
@bigdonny
Sorry guys, but more is more. Always has been; always will be.
@Evan Less is always gonna be more to people who can only run one application at a time. But as you said, more is always more to us folks who like to multi-task!!
@Evan
Really? Give me one reason the processor in the iPad needs to manage an RS232 serial port.
You want a car analogy? If you have two cars with the same power and gearing, the one that weighs less will be faster.
@Jack
Yeah but that's because it has a better hp/weight ratio.
Although I do get what you're saying
@Jack Yea... ok, right... xcept your analogy doesn't even come close to applying. Speed:Weight/HP ratio is analogous computing speed:clean code/bloated programing ratio.
The proper relationship here is much closer to A4:iPad is analogous to econo-engine:Prius
@bigdonny
"Although early erroneous reports placed Apple's custom A4 chip at the leading edge of ARM-based design, things are slowly falling back to earth."
I wonder who was responsible for the hype about the chip?
@Evan Mmm.. Sure, it happens all the time when “competition” is overshadowed by high barriers to entry and near-monopolistic strangleholds on markets. Hopefully, it'll still be the case for this chip: http://bit.ly/a4-examined-on-ipad
god i hate nuke boosters...
The author of that story doesn't know what he is talking about and obviously doesn't have much experience with ARM based SoC's.
@TheLostSwede I'm happy to correct the piece, just list the mistakes and cite to sources, please.
@TheLostSwede Please, do enlighten us.
@Nilay Patel Sorry, I meant the ArsTechnica author, not Engadget
@TheLostSwede Regardless who you're claiming that about, you still need to list what's actually wrong with it in order for that to carry any weight.
@TheLostSwede I haven't talked to Jon about his sources, but I can tell you that I didn't just hit this without doing some checking first. We'll find out for sure at the end of the month, I suppose.
@TheLostSwede You mean Jon Stokes. He is one of the best connected journalists in the world of processors and architecture. It is more likely you are talking out of your ass...
@TheLostSwede PWNED....
@Nilay Patel I'd hit it.
@TheLostSwede well, I guess it's good that you all know what I do for a living then, but never mind, I don't know what I'm talking about, happy?
@TheLostSwede ....yes.
@TheLostSwede
What I know about ARM-based systems-on-a-chip could be written on the back of a postage stamp in a large font but right now I am considered as much an authority on this subject as you unless you prove otherwise. Unless you actually backup your statements with the reason why you are correct then you'll simply be dismissed as another Internet blowhard because, frankly, it's so easy to be one.
@Kelmon Ok, how's this for an answer then?
It doesn't make any sense for Apple to cut out just about any functionality out of the A4 SoC as claimed by Ars Technica, as most of the external devices uses standard I/O interfaces such as USB, SDIO etc. Cutting off these would have a negligible effect on the power consumption as if an I/O interface isn't used, it doesn't really draw any power.
Yes, Apple is using a custom graphics design courtesy of PowerVR and this would mean some changes compared to a generic ARM Cortex-A8 design (as this is presumably what Apple has based the A4 on) and this would potentially give them more control over certain power aspects.
Taking into account the features that Apple have integrated into the iPad, they wouldn't really be able to trim to many features out of the SoC and I very much doubt that the A4 lacks camera support as suggested by Ars Technica as the space taken up on the die by that functionality will be tiny. Since the iPad incorporates Wi-Fi, Bluetooth, a digital compass, TV-out, audio, USB 2.0 connectivity and so on, there's really very little that could be removed to make the chip more power efficient.
However, most of the things I mentioned, plus of course the 3G module in the 3G versions, would be handled by third party chipsets that require the various I/O interfaces to be operation for them to function. This alone proves that Apple couldn't have cut off these features. Sure, some of them, or at least parts of them could be integrated into the A4, much like ST Ericsson has done with its upcoming U8500 http://www.stericsson.com/platforms/U8500.jsp which has built in HSPA functionality. However, most of the other things are still using external chips.
He's making a lot of assumptions, but considering that Apple will most likely run the 3G module over USB and possibly even the Wi-Fi implementation, then how come they'd only need one USB port? Bluetooth would most likely go over SDIO and it's possible that Wi-Fi is using this interface rather than USB, but either is possible. A lot of the other interfaces are used for chip to chip communication, which the author seems to have completely forgotten about. He also states that Apple doesn't need to use UART for serial connections, but then goes on to say that the touch screen might use a serial interface.
Comparing the A4 with the i.MX515 is also odd, as the i.MX515 isn't exactly a widely used ARM implementation in Smartphones, at least not as far as I'm aware. Sure, the iPad isn't a Smartphone as such, but close enough. The bump in the core speed and a tweaked PowerVR SGX implementation will handle the higher resolution screen quite easily and Engadget even had a story up about the iPhone 3GS being able to play back 1080p video http://www.engadget.com/2009/08/20/iphone-3gs-totally-capable-of-1080p-video-playback/ so a performance bump would be all it takes for the iPad to handle its higher screen resolution.
If anything, the power saving would be because of a manufacturing process shrink if anything, say from 65nm to 45nm, which would also make for a smaller, cooler running chip. This is normal praxis among all CPU manufacturers and anyone that follows this should know that this is the most common way of creating a more power efficient chip.
@TheLostSwede
Fair enough - I'll let others decide whether that's valid or not because, honestly, I didn't understand it. Why didn't you just post that in the first place? For all I know that might be technological gold and your comment would have been Highest Ranked with article writers updating their stories across the intertubes. Odd...
@Kelmon Maybe he thought people would understand that he was talking about the source author and not Nilay.
@N900
I don't think that matters. What matters is that simply posting something along the lines of "you are wrong", regardless of who "you" actually is, is guaranteed to be frowned on unless it's backed up by the reasons why, which can then be digested, verified, and either agreed or disagreed with.
@Kelmon Agreed. I guess if that marathon above us both was in his first comment to begin with, he wouldn't be in oblivion.
@N900 He's on his way back up
@oldpass56 PWNED BACK!
@TheLostSwede It really depends. A USB 2.0 end-point can eat up its fair share of die real estate -- the last one I worked with took up ~1x1mm @ 90nm. That may not seem like much but if cost is really an issue (and a $500 iPad sure pushes the price limit) then you shave what you can.
I do agree that the ability to remove extraneous IO blocks isn't something worth hiring a chip team for. But as Jon Stokes pointed out, the most likely reason is the back-end work the PA folks can do in the way of power gating and -- I would assume -- dynamic frequency and voltage scaling similar to what they did with the PPC SoC they released before.
Now mind you, at the time of release (2005), such an SoC was impressive. But in this day and age, you'll be hard pressed to find a mobile SoC that *doesn't* utilize granular voltage islands and DFVS. So while I'm sure at the time of purchase, Apple though they'd revolutionize the chip industry based on silicon optimization alone, they're pretty behind the curve on this one.
That being said, I suspect the A8-based A4 is just a stop-gap solution. Around early 2009, Apple hired a bunch of CPU designers. It's too early for a from-scratch core to be finished -- which would explain the use of the tried-and-true A8 -- but I suspect most of the work by the SoC team would be to make it so that their custom designed CPU could be dropped in at a later date, whenever it's done.
@metafor I never said that Apple wouldn't have made a custom implementation of the A4, but Ars Technica made it sound like this was something unique to Apple, when in reality it's something that every ARM SoC manufacturer has the option to do if they so chose to.
Neither was that story arguing anything about cost, it was all about power saving, which you seem to agree with me that cutting off a few minor bits wouldn't result in any major power savings.
Still, no-one but Apple really knows what the A4 is, but it's not as special as some are making it out to be from what I can tell, based on the details available. It's by no means a bad SoC solution, but I'm sure other ARM based SoC's using the PowerVR graphics core will offer similar performance and power envelope.
With regards to power usage, I'm sure Steve Jobs comment was most likely comparing the iPad to other "similar" devices based on x86 architecture and in that comparison, the iPad draws next to no power. Compared to other similar ARM based devices, I'd expect it to offer similar battery life, although there aren't really that many ARM powered devices with as large screen as the iPad, so it'd be very hard to compare.
You might very well be correct with regards to Apple developing its own custom solution, but I have a feeling that they'll still use an ARM core for the base of it, as developing a brand new architecture from the bottom up would just be too costly. But now we're just speculating on something that may or may not happen, so I'll stop here.
@TheLostSwede There's really no other reason to higher CPU designers other than to make your own CPU. Apple has an ARM architectural license; I can't imagine they'll just sit on it and not use it.
That will likely be at least a year out though. A custom CPU takes quite a bit of time and even longer to integrate.
I wonder how the A4 will compare to the chip in the iPhone4G. It sounds like the 4G will be more powerful
@JXCGunrunna First off, it probably wouldn't be 4G, I don't think the networks are ready yet. Fourth generation, yes, but the current 3G reffers not to the generation, but to the network type.
Second, I would think it would use the same processor. Using the same one cuts cost, and also it wouldn't really make sense to have the weaker processor in the larger device.
Even more reasons im not getting this.
@CJisohsocool You were never getting one. The mere fact it has an Apple logo on it sealed that deal long before any specs were announced.
yawn...
@Postal
http://www.engadget.com/exclude/apple
You're welcome. But thanks for adding to the post count anyway.
@Jack
I like apple news. the keyword being: news, which this is clearly not.
Also, why do you care what the post count is? Strange.
@Postal
Higher post count = more Apple articles. Posting just to say "yawn" is basically saying you want more of these articles.
haha !!
I remember all those fanboys who pretend this marvel was way faster than an ATOM chip :-P
@smover Well it certainly *looks* faster in the demo videos than most atom based netbooks do. Even when they're running a single app (had to get that in there!)
@smover
it is faster than an Atom chip. A LOT faster. Were you under the impression that had changed?
@darksharpie because on the iPad side you have a smartphone OS, and not a computer OS... You can't compare it with a netbook running 7 ^^ You could compare it to a nebook running WP7S, which may exists very soon.
@Jack Atom=x86 (Intel Architecture-32) while A4=Cortex A8 (ARM architecture. Apple vs. oranges, dude.
How's oblivion, by the way?
@aschettler
Nothing you said negates anything I said. You're welcome to try again though.
@Jack
it doesnt negate what you said, just makes it irrelevant.
@smover
The Apple A4 is just a half Cortex A8...
/kidding
@smover
I admit confusion on this one: are you arguing that it is a good thing or a bad thing that the iPad is running the iPhone OS? One the one hand you seem to imply that it is a "cut down OS" and that a "full" desktop OS is better but on the other hand you note that its performance is much better because of it. I would have thought that as long as it does its job well it doesn't matter in the slightest what hardware or OS it is running.
While I would quite like to see what a Windows Phone 7-based tablet might be like, or even the mythical Courier, putting the full Windows 7 on a tablet device is just daft. I'll argue this point until I'm blue in the face (in which case I'll look like a Smurf) but a touchscreen device absolutely needs to be running an OS and applications that are solely designed for a touch interface, which Windows 7 and its legacy of software applications clearly are not.
I like Jargon
Isn't the whole point of the ARM architecture is to power down components not it use? Why build custom silicon unless they are going to tie stuff specifically to their hardware? Normally the power used to run the display and radios (3G, WIFI, etc..) far outweigh the power used by the ARM processor.