Slashdot Mirror
TSMC Preparing To Manufacturer A6X Chip As Apple Looks to Ditch Samsung
An anonymous reader writes with reports that TSMC is preparing to do a first test run of Apple's A6X chipset currently manufactured by Samsung. The TSMC manufactured chips will feature a process shrink from 32nm to 28nm, and there's a good chance Apple will grant them the contract for the next generation A7 chip. From SlashGear: "The test will kick off in Q1 2013, The China Times reports, with TSMC producing a new, 28nm version of the existing 32nm A6X that Samsung has been producing for the full-sized iPad 4th-gen; the smaller chip, which will likely be more power efficient as well, will debut in a new iPad 5th-gen and iPad mini 2."
Now we can wait for the hardware repetition of the Google Maps fiasco. Whose head is going to roll this time when the shit hits the fan?
I'm so excited to hear about every minor thing Apple does. They sure are a groundbreaking company! Imagine, switching to a different supplier. What insight! What killer business acumen they must have!
I've never in my life heard of such a thing. An electronics company sourcing a different supplier for components!
HOLY SHIT APPLE IS SO FUCKING AMAZING! PLEASE POST MORE INTERESTING STORIES ABOUT APPLE!
I heard a rumor that they are stocking their stationary cabinets with scripto pens, and are no longer using Bics. Is this true!?!?!??!?!?!!
This could sour the cozy relationship between Apple and Samsung.
Have gnu, will travel.
You posted the same thing twice. Well done on being a competent internet user.
+1 IDisagreeSoHeMustBeATrollOrAnAstroturferOrAShill
So that's thermonuclear now.
Better go on a fruitinarian diet now Samsung, it will help to cleanse the radiation & "detox", whatever the fuck that means.
You posted the same thing twice. Well done on being a competent internet user.
Sorry, I meant to paste the pro-Apple one that time. Now I'll only get paid by one side of the war :(
I had been planning to purchase an iPad 4 for a while, but I guess I had better do it soon. I don't really want to be a beta tester for Apple/TSMC. There have been serious problems with TSMC's 28nm process and I don't trust them to get it right. And during the past year or so, Apple has shown a disturbing trend of prioritizing screwing over their competitors (Samsung and Google) above providing a good customer experience, as demonstrated by the Apple Maps fiasco and the myriad of problems with LG displays on the Retina MacBook Pro. I'm very concerned that corners will be cut in the rush to TSMC fabrication.
I thought it was Samsung that pulled this contract.
Getting rid of partners like Samsung will hurt them in the long run. The only reason why Samsung became one of the top suppliers of parts for Apple is due to Apple's long history of problems and failures with smaller partners unable to produce significant quantities with the quality expected by Apple. Of course TSMC is not a small company, but Apple dropping Samsung for parts is about pride not intelligent planning or business strategy. Why drop a relationship that works for something less predictable?
Apple is going to have a very tough year in 2013. They blew their wad last year for product updates and except for minor product revisions will not offer anything interesting until at least the fall if Apple TV is not actually a myth. In the meantime news like this will only scare investors at a time when there had already been a loss in faith with Apple's business strategy.
I think the problem with Apple is they are still trying to follow in Steve Jobs' footsteps. It's only Steve Jobs that had a hate on for Android, Google, and anything connected to them, so pursuing this prideful vendetta against Samsung is like Steve's dying wish. The problem is that Apple is going to have to eventually start making their own decisions and continuing a strategy to remove Google and Samsung as business partners is not in their best interests.
I haven't thought of anything clever to put here, but then again most of you haven't either.
Yeah, lets just just send more money into China. Great idea.
At least Korea wasn't China.
Apple demands "Do it for us cheaper or we go elsewhere!".
Samsung are merely stating that if they aren't going to be generating goodwill from Apple by kowtowing to that demand, why the hell give them preferential prices on their product?
Samsung may not be able to sell quite as many chips as if they had Apple on board, but they'll make more per chip because Apple isn't gouging their profit margins.
28 at TSMC is the same 'process' generation as 32 at Samsung. The smaller number represents a 'half-node' where certain components have a smaller 'geometry' on a given process. Essentially, you use optical 'tricks' on the mask to make certain components or interconnects a little bit smaller.
In reality, the half-node designation is more of a marketing trick to imply a more advanced process. This issue also applies to Intel. Intel claims a full process lead over plants that fabricate similar chips, but there are often aspects of Intel's process that are similar to the previous generation process from other fabs.
In the end, the real metrics are transistors per mm2 and the power dissipated by these transistors. In Apple's case a slightly bigger die is no issue if it means the chip uses much less power. Indeed, traditionally, extremely low power chips have frequently been made on old processes that have long been obsolete for mains powered desktop CPUs.
As chips shrink, for instance, interconnects (wires) get shorter, but the 'volume' of these interconnects (third-power) gets smaller still, increasing the electrical resistance, and thus the power wastage (thru heating effects) for a given material at a given voltage. Dropping the voltage used by a chip is an option, but this option has largely already been used up in previous generations.
A modern ARM SoC (system on a chip) that powers a phone or a tablet attempts to SEEM low power by switching off most of its functional units most of the time. For instance, sound and video playback are always done by tiny dedicated logic units, designed to be very power efficient. OTOH, run a desktop like benchmark for the CPU/GPU on a tablet for long, and many tablets will get too hot to handle, as the battery is exhausted in a fraction of the usual time.
It is not a coincidence that Apple made its first tablet so large- at that size you can fit a large-capacity battery in the case. The 4:3 aspect ratio helps here too. 7" tablets with a 16:9 ratio are extremely compromised with respect to case space for the battery. As the SoC gains 4 64-bit cores and a very powerful GPU, battery capacity is going to become a very important factor.
What are you talking about? They already had a hardware version of the Google Maps fiasco - it cost them money to bail out Sharp.
This is a second time of doing the same - which makes things even more idiotic. Apple is determined to cut off their nose to spite their face, apparently.
The A6X puts everything else in the mobile industry to shame.
http://images.anandtech.com/graphs/graph6472/51764.png
http://images.anandtech.com/graphs/graph6472/51759.png
The first graph is a masure of memory bus bandwidth; while this has historically sucked on ARM in general, and Apple has had the lead in this area because they started life with their own memory controller design replacing the one normally supplied by the ARM folks, this lead has been significantly narrowed in the Samsung Exynos line. There are still some optimizations to be had to match the A6X speed, but it's close enough that for an optimized pipeline, it's not going to matter as much as the graph shows. nVidia's ARM offerings still have pretty sucky memory bandwidth, as do the Qualcomm chips, like the SnapDragon series.
The second graph measures the GL pipeline, and since the Linux stack moves copies of surfaces rather than GL over the user/kernel boundary, there's significant overhead in the protection domain crossing, as well as in amount of data being moved (which devolves, again, to the memory bandwidth issue).
Unless you go direct GL across the user/kernel boundary, and run the surface processing entirely in kernel space, if you are doing software compositing, even in the case of 2D, which applies to the vast majority of surface transforms, since people simply do not watch YouTube videos on rotating cubes or on spheres, you are talking 3 protection domain crossings to get the data from user space to the GPU, process it in the GPU, bring it back to user space, and then push it again across the boundary to get it into the frame buffer.
Basically you are comparing a graphics stack that sucks with one that generally doesn't, except under specifically defined circumstances (and in those cases DirectX beats GL in terms of reduced system vulnerability to unbounded texture processing for both the Linux and MacOS X/iOS stacks).
Either way, the comparisons in those graphs are not straight apples-to-apples unless you happen to be running Android on all your devices, and so have the additional expensive-on-ARM copy and protection domain operations on all the platforms.
If it comes out only 6 months after the last ipad refresh, Apple is gonna have a riot on their hands.
I wonder of TSMC will be up to task, or will be another Sharp, who's still teetering on bankruptcy despite Apple's contract.
Taiwan Semiconductor Manufacturing Company.
Google is hard.
Slashdot still doesnâ(TM)t support Unicode after it was added to the HTML standard in 1997.
If it comes out only 6 months after the last ipad refresh, Apple is gonna have a riot on their hands.
The iPad 4 was really only a minor update. No-one cared when it came out six months after the iPad 3, and no-one will care when the iPad 5 comes out in sixth months with further refinements.
"There is more worth loving than we have strength to love." - Brian Jay Stanley
Perhaps Apple can pump enough money into TSMC to get some manufacturing lines running that can compete with Intel's last-generation process technology. This will keep AMD's CPUs in a not-quite-competitive state for another 6 months, at least!
Remember when Motorola was developing the 88xxx series and announced that the 68060 would be the last of that series, thus forcing apple to migrate quicker towards the 88xxx series. To appease Apple, consequently, Motorola designed the 68070 to allow for a slower transition, yet in 1990 IBM announced the availability of the POWER architecture for servers, and the RSC for low wattage embedded applications. With introductions of the POWER architecture IBM raised the bar substantially with record breaking performance however competitors such as HP, Sun, and DEC complained that IBM cheated because the POWER1 architecture was divided amongst six chips as opposed to the competitors one. Not only was IBMs POWER1 setting new benchmarks levels of performance but also the RSC for low wattage application was defining a new level of performance for embedded applications. Apple being forced to rewrite their code to either a 88xxx which is just coming out of development and is unproven, or to rewrite to the RSC which has finished over four years of development and several years of testing. So apple decided to screw Motorola and whatever effort they've put into 68070 because now they want IBM's RSC. So a partnership is formed and it's referred to as the A.I.M. alliance with the objective to promote and popularize IBM's intellectual property;{Embedded}Motorola licenses the PowerPC from IBM and gets a well tested 32-bit cpu architecture complete with compilers and testing tools and even guaranteed future sales to apple, {Personal computers}Apple gets a cpu that is nearly three times the performance of Intel's Pentium and a discount on purchases from Motorola and provisions that neither IBM nor Motorola will enter the PC market with PowerPC for a limited time,{Enterprise} IBM gets reassurance in knowing that Motorola the largest manufacture of microprocessor has adopted its intellectual property and is committed to promoting it while apple just two years before was the largest manufacture of personal computers is committed to promoting IBM's intellectual property thus total dominations of the 32-bit personal computer markets is assured leaving IBM to just focus on the 64-bit enterprise market--perfect plan what can go wrong. Well surely Apple believes that a PowerPC system that is over twice the performance of a Pentium system should also cost twice as much, and fuk market share that only benefits Motorola/IBM its all about the margins baby. After several years of shipping low volume and high margin personal computers apple needs new cpu's to compete with x86 which is closing the gap in perfmance between PowerPC, yet Motorola's other customers which provide over 50 million unit purchases of 130nm PowerPC are completely content with the performance why should Motorola build a new fab to provide apple with new cpu's when apple unit purchases are less then 5 million annually. Apple reluctantly gives Motorola money to help in the development of higher frequency processors. Motorola to get this cheap bitch(Apple) off its back asked IBM to also help in the development of 74xx series. So apple get its 74xx processor and decides to use the same style of naming that IBM is using for its main-frame cpu, and call it a G-something, thus providing Motorola with less brand recognition and hopefully tarnishing/confusing it with IBM's main-frame cpu. Nice move apple, screw your partners and get those margins. Apple begs IBM for a 64-bit processor, so IBM design's two one for desktops and another for laptops, yet IBM said the desktop cpu should reach 3Ghz, yet only reached 2.8Ghz by the deadline so apple wants out of the contract. Next on Apples screw list is TSMC, Intel, then ARM.
So what you've said, is that the architecture, both software and hardware, that Apple is using, is better for this type of work.
Got it.
But there is not that much difference in newer hardware like the Exynos, and the software architecture is a matter of implementation decisions which can be reverted, if in using Linux, one is willing to work around the spirit of the GPL in exchange for performance.
The person posting about the Raspberry Pi didn't actually mention it, but they have a rather radically different user/kernel interface for the GL in order to support feeding the Broadcomm VidCore; most Open Source user space graphics stacks couldn't talk to it, since the two X-on-GL projects I'm aware of have basically ceased operation.
So the Raspberry Pi has addressed both of the issues, although they've done so on a chip with pretty poor memory copy bandwidth anyway, so don't expect to use the chip for high performance video on much beyond the set-top boxes that BroadComm initially targetted the chip at in the first place.
Personally, I held off on buying my Raspberry Pi until such time as they promised to open source the video drivers, and immediately bought after getting that promise.
Transport the Raspberry Pi model to a GL kernel space on an Exynos, though... and then you've got something meaningful relative to the Apple stack.
Either way, the graphs are telling a non-Exynos hardware story and a non-optimized software story to play up the Apple hardware and software stack. It's good (disclosure: I worked on the Apple Core OS kernel team), but it's not magic which no one else can replicate; as a friend of mine commonly says: it's a mere matter of typing.