Rumors are again swirling that Apple and Intel are in discussions about Chipzilla baking the chips Cupertino uses to power its iDevices. "A source close to one of the companies says Intel and Apple executives have discussed the issue in the past year but no agreement has been reached," Reuters reported on Thursday. This isn't …
Pretty sure Intel would sooner push them towards x86, would be a major win for them.
But ARM is more efficient.
The analyses done by AnandTech seem to point towards the new x86 chips being very close to competitive with ARM. If there's any company that can pull off the integrated software and hardware approaches necessary to get enough battery life out of an x86 chip it's Apple: just look at MacBook battery life vs all of their competitors.
That said, it still won't happen, but efficiency isn't the reason. Apple have spent billions setting up their ARM design wing and building iOS specifically for ARM. That's not an investment one can easily throw away. Plus you'd end up with all sorts of fragmentation nightmares - remember Apple never quite got POWER applications running well under x86 OS X, they pretty much just rode the problem out until everyone had switched to the new architecture.
Tom's Hardware were of the opinion that ARM were efficient when doing nothing - good for smart phones - but when they looked at power conspumption by task, the race was far from over. They did note that Intel were taking greatly pains to invite analysis of their power consumption.
I don't know if this rumour is true, but Apple are x86 customers of Intel's, and having their ARM chips made using Intel's fab processes would give them an edge over their iDevice competitors.
"seem to point towards the new x86 chips being very close to competitive with ARM"
But how much of that is down the the chip design (i.e. logical structure) and how much of that is down to Intel having a ~2 year lead on advanced manufacturing processes?
If they are matched simply because a 2 year advance makes the x86 equal to the ARM for power-critical tasks, then ARM is still a winner for most, and if Apple do get access to the best fab service then they will have a major advantage over all others.
That alone, far more than not paying Samsung for fabbing, would help Apple in the premium smartphone arena.
Intel vs. Arm
The ARM architecture is intrinsically more efficient than x86. If one could compare an ARM CPU against an x86 CPU implemented using the exact same fab process, ARM would win. (OK, for flat-out number-crunching the design of the FPU also becomes a very significant factor and Intel may have a sufficient edge to win here - thought Nvidia does it even better if you can run on a GPGPU instead of a CPU).
At present you can get an intrinsically inefficient x86 CPU implemented on Intel's proprietary process, or an ARM CPU handicapped by an inferior process, and the result is something close to a tie.
It would be a very interesting turn for Intel to fab Apple products. Here are some questions to think about.
o The risk for Apple is that Intel is not in the fab business and could lose interest in fabricating Apple products pretty quickly if demand turns up for Intel mainline products. It has happened a number of times with other Intel products that got squeezed out when Intel optimizes it's profit and revenue per wafer.
o Intel fabricating Apple products would mean that Intel has determined that it can NEVER make a competitive x86 mobile product so it must either build the Apple products or develop it's own ARM products to grab a slice of the mobile pie relegating Intel to just another Fab house or ARM vendor with lower margins and smaller profits.
o If Intel where to build Apple products it would improve Apples competitive position and could potentially ensure the lockout of Intel x86 mobile products.
o Intel believes that it's process technology is it's secret sauce and giving that away to Apple would the admission of Intel's failure in mobile.
For all these reasons I think that the likely hood of Intel fabbing Apple products is less than 5%. Intel does not want to cut it's own throat, give away the family jewels or admit failure.
the altera deal...
It might be good for Apple... but it will turn to be a curse and suicide for Intelfidals ... Intel no more will be able to protect its x86 exclusive jargon and companies will taunt them so hard that u make x86 but no ARM for Intel servers that at one stage in approx 1-2 years they will throw in towel and will also "unwilingly" launch ARM servers .. which is what AMD will be found doing for years. This wil be the point where x86 tech will actually spit on Intelfidals face and jump down the drain... Nature keeps creating environment for the downfal of empires and bring weaker ones as the new leaders...
Swift cores on 22nm tri-gate anyone?
If Apple could get Intel to produce it's Swift core ARM chips on Intel's latest and greatest 22nm (future 14nm) tri-gate technology, it would give the company a huge leap ahead of the current competitors using 28nm ARM A15 cores that are quite hungry and expensive.
I feel that this would take a LOT for Intel to produce anything other x86 based chips but, you know what? The fat lady did sing: Apple moved from Power PC to Intel... Stranger things can happen!
Re: Swift cores on 22nm tri-gate anyone?
Technology size has become the property of marketeers now as when people look closely at what is actually being manufactured then intel's 22nm seems to be very similar is scale to everyone else's 28nm ... to such an extent that seems that what would have been 20nm is now being renamed 14nm!
iPhone / iPod only
Intel's been happily licensing, adapting and baking silicon using non Intel designs for years e.g. The ARM based XScale range and the amd64 derived Atom and Core i cores.
Re: iPhone / iPod only
Yep, right since the 4004. However, your examples aren't great. Intel sold off XScale to Marvell (just as the smartphone processor market was taking off) and amd64 is a set of instructions, not silicon design, which they licensed begrudgingly from AMD because of the success of Opteron embarrassing their Itanium business. But AMD is likely licensing far more IP from Intel.
Hopefully, this will mean Intel would bake ARM-based chips, instead of the awful alternative that is having yet-another-PC-disguised-as-something-else in the iPhone.
Maybe it is time for Intel to switch to more promising architectures.
Intel cannot move away from x86 for PCs and servers. There's too big a backward compatibility problem. I mean, they tried with Itanium and it was a complete failure (although it succeeded at killing of HP and DEC's CPUs, which was a win for Intel).
Intel already does bake ARM-based chips, especially after acquiring Infineon's mobile division and their X-GOLD, S-GOLD etc. range.
They could move away, just as AMD managed a transparent shift from i686 to x86.
The easiest route might be something like a bigLITTLE architecture with ARM cores and x86 cores on the same die, sharing cache, memory controller and maybe other internals. Gradually shift the balance of core provision towards more ARM cores and less x86 ones, as software takes advantage of the new feature.
Not quite expecting Intel to jump to the cold water, but at least not force Apple to switch over their ARM stuff (the iGadgets) to x86.
Their Itanium (Itanic) venture actually failed because they were originally trying too hard to keep the x86 compat layer, and their first chips were real slow. Apple had a better way to jump from one arch to another, and they've done so twice: Motorola 86k -> PowerPC -> Intel. Maybe they should've relied more in the OS manufacturers to port stuff instead of trying to emulate everything?
Probably depends on what Apple want to pay Intel ... Intel get a huge amount of $/wafer from x86 chips
True, but if it is a case of money from Apple that is less than x86 or Apple staying with Samsung and getting nothing, not so simple.
I suspect the bigger worry for Intel is if this indirectly leads to more interest in ARM for servers running massively threaded code where lots of cores at high MIPS/Watt is far more attractive than high-profit high-end x86 chips.
Otellini don't mean no ARM
Apparently Apple averaged 10 ARM's per second over the last three months. Intel would have to be pretty stubborn to not want a peice of that.
"This isn't be the first such report."
Stop the presses. Do I speak English? Maybe not this variant.
Intel needs to fill fabs if they want to go to 450 mm wafers
Those wafers are 50% larger than today's 300 mm wafers, thus you get ~2.25x as many chips per wafer. Assuming the 450 mm equipment processes wafers at the same rate, you need half the number of fabs or twice the volume of chips. Intel really only has 4 full sized fabs today, with only two they'd lose a lot of the economies of scale that have helped them for years.
Apple would be a good pick for them because they are highly unlikely to ever go x86, because they're expanding their custom SoC designs to include designing their own GPU (they reportedly hired away an entire team of top ex-ATI guys who mysteriously left AMD last fall) You don't make moves like that if you want to go back to buying off the shelf SoCs designed by someone else.
One other advantage for Intel is that Apple has a lot of cash, and they've demonstrated they're willing to pay in advance for guaranteed capacity. Supposedly they tried to cut a deal with TSMC for guaranteed capacity last year but TSMC turned them down. Problem is, TSMC already has more customers than they have capacity.
This deal can only happen as long as will.i.am signs a contract promising not to be associated with any idevices
Oh yeah, Sammy put their prices up didn't they
By, if accounts are to be believed around $1BN which was the same as that court judgement amount handed down.
Which has just been reduced by around $450M.
Apple are obviously afraid that Samsung will be nuked by North Korea...
"but it would also free them from reliance on Samsung"
Unfortunately, Samsung has taken an interest in a Japanese manufacturer making screens for Apple.
So much for freedom.
I will stick my neck out and say that Apple are developing a i86 version of its mobile processor.
Because Apple want to ditch Samsung,
Because Intel needs to take on Arm on at Mobile, Winning Apple would be a watershed for intel and beginning of the end for Arm dominance (in that Market),
Because Apple has done it before, it would not be a surprise if the code was not compiled already.
Because Intel has seen off every competing architecture that has threatened it.
Because the technological challenges are solvable.
Because of the blurring of the lines between mobile portable and desktop, and its easier to support one architecture than two