Slashdot Mirror
Apple's A6 Details and Timeline Emerge
MojoKid writes "For a CPU that hasn't seen the light of day, there's a great deal of debate surrounding Apple's A6 and the suggestion that it may not appear until later in 2012. The A6 is a complex bit of hardware. Rumors indicate that the chip is a quad-core Cortex-A9 CPU built on 28nm at TSMC and utilizing 3D fabrication technology. While the Cortex-A9 is a proven design, Apple's A6 will be one of the first 28nm chips on the market. The chip will serve as a test case for TSMC's introduction of both 28nm gate-last technology and 3D chip stacking. This is actually TSMC's first effort with an Apple device. The A4 and A5 have both historically been manufactured by Samsung."
Steve's not dead two weeks and already Apple fumbles the ball. STACKED chips? How is the next iPad going to be as thin as it can possibly be when they start stacking chips?
Awkward, off-topic and pathetic rabble. Troll credentials rescinded for immediate review by troll committee.
The Admin and the Engineer
I love my quad core desktop processor, but I find myself scratching my head at the idea of quad core CPU in a tablet. Even with iOS 5's enhancements there's no true multitasking in it or any other tablet/phone OS - every application is interacted with in a full-screen monolithic manner.
Dual core CPUs allow the OS to do one thing in the background and not bog down the device for the running application, but what on earth are you going to do with 4 CPUs when you can only interact with 1 program at a time? This seems like it would only be of benefit to games and a couple other niche uses, otherwise a processor with fewer cores and higher per-core performance like the A15 mentioned in the article would be far more beneficial.
I bet they'll try to patent this "innovation" -- even though they clearly stole the idea.
For goodness sake, Pringles has been stacking chips since the 1960's.
Required reading for internet skeptics
Apple has already had problems in the past with low-stock at launch. Why would they risk having even worse problems using unproven tech at a fab they haven't used before? There's always problems with supply when dealing with smaller fab tech, which will probably be worse with 3D being thrown in.
The A4 and A5 are not even that old.
GENERATION 25: The first time you see this, copy it into your sig on any forum and add 1 to the generation.
Is TSMC now into doing assembly, in addition to wafers? Since when did it get into the packaging business? I thought that their business model was to ship their wafers to the assembly houses approved of by their customers, in this case, Apple, and that the assembly houses involved would do the packaging for them. From 3D stacked chip, I'm assuming that they'll be stacking multiple die on each other, like in an MCP. What's it in case of an A6 - 4 basic CPU's just stacked one over the other? Some of the signals, like data & address could be easily routed, but quite a number of the control signals would have to be multiplexed so that more than one CPU ain't accessed @ any one time.
The other part of the question - iOS - is it something that's as SMP enabled as OS-X is? From what I've seen of i-PADs, they are not multi-tasking OS's at all - all they do is save the state of an app once you exit it, and resume from that point if you return. If that's the case, how does multiple cores help for this case?
Finally, Apple can make this chip even better for themselves by moving their macs and airbooks to this processor, so that they have just one CPU platform of their own, making it easier to have a common code base for their apps, like Safari, Mail, et al.
FTA:
" Given the iPhad's dominant market position, "
I wonder who slipped that in there?
"National Security is the chief cause of national insecurity." - Celine's First Law
this could well not be true - these days Apple takes one sustainable step forward (like incremental development of iPod) - rather than one giant leap ahead that well turn to a giant fall (like Newton).
I believe that energy consumption of a transistor is the square of its speed. With double the microprocessors, instead of a faster processor running the less urgent threads, multiple, slower processors can. Thus, power savings.
What this boils down to is jack squat. Every damn time some new chip comes out, there's more marketing diatribe surrounding it than actual functionality. Need I not remind us all of the P4 boner everyone had at the time? Give this thing a month (if that) and the hype will be plaguing something else...
So TSMC's 28nm is going to be what is behind AMD and nVidia's next gen GPUs, despite their poor handling of 40nm for both companies. Those guys (nVidia in particular) also have a large first dibs on the production.
So if they are planning on the A6 from there later in 2012, well I could see it. Both nVidia and AMD want to launch new GPUs soon. I'm sure they want a Christmas launch though realistically it'll probably be early next year. Ok well they do those, tons o' chips are made with the 28nm process, the big surge of demand is met, then things are good. Mid to later 2012 comes along and the 28nm TSMC process is stable and the kinks worked out, good to go.
However if they are going to try and do it early 2012, well I think that'd be bad. They'd be seriously supply constrained fighting with nVidia and AMD. While no doubt TSMC would love to give Apple what they want to get their business, they have standing contracts with nVidia and AMD (which aren't small either).
All user interface operations have to be done on the "main thread", that is, the thread that was running when your program starts up. If you have a lot of CPU intensive work to do, or will be blocking for I/O to complete, Apple actually recommends that you put that on a background thread. There is support for for threads in both Cocoa (desktop) and Cocoa Touch (iOS).
Assume by "stacking" they are referring to (and the article alluded to) something similar to Intel's Tri-Gate transistors?
http://hothardware.com/News/Intel-Announces-New-22nm-3D-Trigate-Transistors/
And not simply stacking and interconnecting like this?
http://www.tomshardware.com/news/rochester-3d-processor,6369.html
It seems that nvidia has a quad core A9 with GPU, with devices shipping October 2011.
Given that many people are using TSMC, I suspect a lot of fabless chip companies have plans for 28nm in 2012. And Apple is not TSMC's biggest customer by a long shot.
Apple can afford to bring out iPad3 with a CPU that is not much faster than the current one.
What they can not afford, is stalling GPU performance.
If rumours are correct, and iPad3 will have a retina display, it will need a lot more shader performance to fill that screen with 3 million pixels. As it is now, it is hard enough to get 60fps on non retina displays with moderately complex OpenGL ES2 shaders.
Bram Stolk http://stolk.org/tlctc/
Apple has been twisting Intel's arm (that IS a pun) about power consumption and threatening to dump their chips in favor of ARM. Another way Intel limits Apple is that their product cycles are tied to Intel's product cycles, which constrains Apple to a parity with other laptop vendors. By moving to a homebrewed CPU, it would give Apple even more architectural control / freedom which would assist in differentiating Apple products from their competition.
Funny how it all comes full circle. Apple suffered from having its unique RISC architecture for many years. Then Apple conformed to X86 for just a few years and leveraged that to get enough marketshare that they can move back to an independent architecture again.
Seth
$5 / month hosted VPS on linux = awesome!
Going to a second supplier makes sense for most companies including Apple.
"Most"? Actual it does NOT make sense most of the time for simple reasons of economics. Virtually all manufacturing has large fixed costs (tooling, engineering, setup, salaries, etc) which have to be recouped somehow. If you produce a small number of units, your per-unit cost climbs steeply. This is 100% of the reason for volume discounts.
The problem with using a second supplier is that you are replicating all of these fixed costs but you can only amortize them over half the number of units. Worse, both suppliers have to be able to scale production in the event the other supplier cannot meet demand. This means your equipment utilization is going to be quite poor since you are keeping production capacity intentionally idle. It is a very rare circumstance where you can second source (on custom products) without incurring very significant extra costs.
Second sourcing only makes sense under one of three circumstances. Huge volume, huge risk or a commodity product. It might make sense for Apple to use a second source supplier but the costs are very real and very significant.
im sure more dogs/cats eat biscuits than humans.
Only humans eat cakes, and even then not more than 1-2 times a month. Its hardly a daily purchase, unless you consider bread to be a different cake, which in theory it is, a bland plain cake.
Why dont we tax drugs instead, more $ in it.
Liberty freedom are no1, not dicks in suits.
(1) Apple isn't getting two substantially different parts from two suppliers; they are getting two identical parts from two different suppliers as Apple designed the chip themselves.
That has no bearing on the economics of the situation. Both suppliers still have the same fixed costs to amortize. The fact that the product is identical is irrelevant. They both have to buy equipment, hire staff, engineer the build processes, etc. These are fixed costs that have to be paid even if they never actually produce a single unit. With a second supplier, Apple pays many of these costs twice but neither supplier can amortize them over as many units. This drives the price up. It HAS to cost more, the only question is whether the additional cost is worth it. Sometimes it is worth it. If you have a unreliable supplier or if the product is difficult to build or if using a single supplier causes strategic pricing issues it can be worth the cost.
(2) Apple doesn't manufacture anymore; that has been outsourced to Foxconn with engineering and design remaining with Apple so their costs of using two suppliers is substantially diminished.
It doesn't matter whether Apple makes it themselves or not because they still have to pay for it. Foxconn can realize efficiencies Apple cannot because Foxconn specializes in assembling products whereas Apple specializes in designing them. Apple's not a manufacturing company. But if Apple uses a second supplier for ANY product, Apple will be forced to pay a higher price because of the supplier's fixed costs. This isn't speculation, it is basic cost accounting.
Really this is no different than Apple getting flash memory from two different suppliers. Apple like other companies does not have to re-design the board for two different suppliers.
The fact that Apple does not have to re-design anything is irrelevant to the fixed cost amortization problem. The supplier still has to buy equipment, hire people, pay for utilities, etc. Companies buy from multiple suppliers for only two reasons. Either they are trying to manage risk and the associated costs OR because neither supplier can supply their entire purchase. There are very real and demonstrable costs associated with using multiple suppliers.
Disclosure: I'm a certified accountant and I specialize in corporate cost accounting. I do this stuff for a living.
The problem with hardware acceleration chips is that they are typically single purpose. Maybe as an iPhone owner who only downloads iTunes videos using a single codec, that's great but once you step outside the box and want to view a video in one of the other dozen common formats then 4 cores become much more important for decoding on the fly.