Apple Threatened to Abandon Intel Chips over Power Consumption

[techsmith]

Arm

I think ARM does have the ability to take over Intel one of these days. The nice thing with ARM is that anyone (even you or I) could buy a license to make ARM chips, so you have far more competition. With Intel, the competition is AMD, and Intel is definitely dominant at the moment, and they have a tendency to behave monopolistic. The Pentium 4 days were the dark ages. Their chips were greatly inferior to the AMD chips but instead of using their money for improving the chips, they spent it on "Intel Inside" advertising and punished suppliers that sold AMD by giving late shipments of Intel chips and no discounts. Now Intel is rocking with their tick-tock strategy, but how long is it before all these ARM licensees catch up? The next Windows will have an ARM version, iOS already works on ARM and it is built on top of the same core libraries as OS-X, so the pieces are falling into place for the replacement to happen, even on the PC side. Frankly I'm quite excited about the advancement of ARM. They are improving at such a fast rate.

 

[ViperDesign]

Hmm, I feel like microsoft walked into the building and said, I will call him mini me. Mini me being apple.

Apple has become what it seems people left microsoft for, something different. in the end, they are the same

Sent from my ADR6400L using Tapatalk

 

[alent1234]

I think ARM does have the ability to take over Intel one of these days. The nice thing with ARM is that anyone (even you or I) could buy a license to make ARM chips, so you have far more competition. With Intel, the competition is AMD, and Intel is definitely dominant at the moment, and they have a tendency to behave monopolistic. The Pentium 4 days were the dark ages. Their chips were greatly inferior to the AMD chips but instead of using their money for improving the chips, they spent it on "Intel Inside" advertising and punished suppliers that sold AMD by giving late shipments of Intel chips and no discounts. Now Intel is rocking with their tick-tock strategy, but how long is it before all these ARM licensees catch up? The next Windows will have an ARM version, iOS already works on ARM and it is built on top of the same core libraries as OS-X, so the pieces are falling into place for the replacement to happen, even on the PC side. Frankly I'm quite excited about the advancement of ARM. They are improving at such a fast rate.

laptops come with 4GB of RAM. once you start using normal RAM in ARM devices most of the power savings will vanish

 

[KnightWRX]

laptops come with 4GB of RAM. once you start using normal RAM in ARM devices most of the power savings will vanish

Once ARM catches to Intel in terms in instructions per second, 64/32 bit hybrid architecture, etc.. etc... the power savings won't really be there. Intel is quite aggressive on their performance per watt figures.

The plain fact is, there's probably a lot of headaches involved in changing over to ARM with very little benefits to be made. An ARM MBA would be mostly atrocious, Apple would be better served going with AMD's APU or even Intel's Atom if it were to transform the MBA from what it is to simply a netbook.

 

[cmaier]

Once ARM catches to Intel in terms in instructions per second, 64/32 bit hybrid architecture, etc.. etc... the power savings won't really be there. Intel is quite aggressive on their performance per watt figures.

The plain fact is, there's probably a lot of headaches involved in changing over to ARM with very little benefits to be made. An ARM MBA would be mostly atrocious, Apple would be better served going with AMD's APU or even Intel's Atom if it were to transform the MBA from what it is to simply a netbook.

Exactly. Even if an ARM processor supported 4 cores and 64 bit, and even if it ran at the same frequency as intel's offerings, it wouldn't have the same IPC. And if the clock frequency were increased to compensate, it would not longer be competitive on power consumption.

 

[Trudy]

G5 was/is a real 64bit CPU. X86 uses 64bit extensions.
That is the reason why you don't see any performance gains in X86 when using 64bit program. Actually, on X86 performance decrease in 64bit by 3%. Why use 64bit when it is slower?

Actually it's the other way around. The PPC was designed 32/64 bit from the beginning, so creating a 64-bit version inherently didn't change much for direct performance unless you were using 64-bit math. In fact, using 64-bits without taking advantage of the 64-bit math can slow you down since pointers and other things are larger, meaning fewer can be held on the CPU.

The x86 was still going off of limitations of the original architecture from the 1970s, especially in the number of registers, fixed at a paltry eight (PPC has 32). x86-64 increases the number of registers to 16, providing up to a 20% performance boost, and also doubles the SIMD registers.

Why haven't one single PC company ditched BIOS?

Backwards compatibility. Windows didn't get EFI support on x86 until Vista SP1. EFI is sweeeeet.

Apple was fist in graphic accelerate its operating system back in 2002

I'm not sure, but I think that title would go to the Amiga in the mid 80s. It had a built-in blitter chip that the OS leveraged to accelerate graphics. I believe the Atari had an optional blitter chip available back then too, and it was included in the late 80s.

ARM was initially funded by Apple. Not founded. Venture capitalists are not founders; just money providers.

It was founded as an Apple/VLSI/Acorn joint venture because Apple didn't want to deal directly with a competitor, Acorn. Apple then funded the founding of the new company, and gave direction for the new ARM chip so it would fit Apple's needs for the Newton.

 

[AppleScruff1]

Why Intel want Apple:

Small marketshare but this scenario is very likely:

Apple: Your chips stink. Fix them.
Intel: Sure and who's going to fund our R&D to fix them?
Apple: We will, take some cash and get us the product we want.

When you have a customer who can help fund your R&D or is some cases pay all of the R&D for a particular product you don't say no. You say anything you can to keep them.

You don't think that Intel can finance their own R&D? They need Apple to design better chips?

 

[justinfreid]

I think the future will be ARM based. A new architecture and a brave new world where the only apps are those blessed by Apple and only available through the Apple App store.

Current trends certainly point to a tightening of Apple's control of app deployment, even as OS X remains open. The melding of the mobile and desktop conventions favors a consistently more iPad-like experience as in Lion.

 

[mdelvecchio]

Yep, maybe monkeys are gonna take over the world. One never knows.

the irony here is unbelievable. you do realize that, right? where was apple 11 years ago? bankruptcy, my friend.

 

[stevep]

The only reason Intel is slightly interested in Apple is because it wants Apple to switch iOS devices to Atom. This will happen eventually but might take some time (remember, switch from PowerPC also did not happen over night).

Brilliant. Comedy at its' best.

 

[alent1234]

Brilliant. Comedy at its' best.

once they switch atom to the 22nm process it will kick ARM's a$$ and be just as energy efficient. Atom is already being used in a lot of consumer electronics, just not advertised

 

[darkplanets]

I find this hard to believe at face value. Perhaps it's true, but from my Armchair Analyst™ viewpoint I don't see Apple being a large enough customer to influence road-map decisions that much. If they were a larger portion of Intel's business (let's say 10%+ for an arbitrary value), perhaps, but right now the only reason Intel has incentive to capitulate is to potentially capture the iDevice market down the road, something that's not likely to happen.

I instead propose that this is a much more subtle strategic move that should have been taken long, long ago-- some pencil pusher finally realized that there's market potential in ARM wattage processors, namely for non-Apple mobile products. Phone manufacturers have no allegiance over mobile chips, and while Apple spent money in-house for their chips, I'm sure Intel would love to take business from the TEGRA platforms, etc. The tablet push only exacerbates this point-- Intel should obviously see that for highly mobile devices low wattage is the paramount factor.

I'm personally rather in favor of this move-- ARM can NEVER compete with x86-64 chips on a performance level, especially once you count the power requirements to compensate with Intel clockspeeds and usage of regular RAM, and was never really intended to. Like it or not, but Intel's R&D is much better than any Apple in-house chip session with ARM standards, and Apple is not about to get into the chip business outright... and that gap will only widen once Intel's "next-gen" chips come down it's pipeline. All Intel needs to do is reshuffle chip architecture some more and reduce it to 22nm (or under) and now Intel can compete with ARM while providing all the bells and whistles ARM cannot. It's a logical progression of their technology.

 

[Hyper-X]

"Bully Intel" into making more efficient chips?? That's like saying someone is trying to bully Porsche into making faster cars. Making a better product is in Intel's best interest. A more efficient chip would be more attractive to ALL of Intel's customers, and would thus improve Intel's bottom line.

Hardly what I would call "bullying".

Right... like Porsche needs someone to tell them that they need to make faster cars, as if that's not already happening without you telling them. Do you think Ferrari needs you to tell them that their cars need to perform better than their current lineup? I'm sure Toyota needs you to tell them that they need to produce more fuel efficient cars right?

Apple making threats to Intel is quite laughable considering Intel's amazing track record for producing very high quality products. Intel has been doing just fine the way how they've been doing things without Apple telling them how to do what they do best.

 

[lilo777]

Why Intel want Apple:

Small marketshare but this scenario is very likely:

Apple: Your chips stink. Fix them.
Intel: Sure and who's going to fund our R&D to fix them?
Apple: We will, take some cash and get us the product we want.

When you have a customer who can help fund your R&D or is some cases pay all of the R&D for a particular product you don't say no. You say anything you can to keep them.

Intel's R&D budget is way bigger than Apple's one.

 

[cmaier]

Actually it's the other way around. The PPC was designed 32/64 bit from the beginning,

No it wasn't.

 

[Rodimus Prime]

The x86 was still going off of limitations of the original architecture from the 1970s, especially in the number of registers, fixed at a paltry eight (PPC has 32). x86-64 increases the number of registers to 16, providing up to a 20% performance boost, and also doubles the SIMD registers.

That is more a RISC vs CISC argument. RISC chips need more registers to work compared to a CISC chip. They relay much more heavily on the registers. CISC uses memory a bit more.

PPC was a RISC chip. Pentium is a CISC chip. ARM is RISC.

 

[KnightWRX]

Pentium is a CISC chip.

The P6 architecture that originated with the Pentium Pro (and went on to be known as the Pentium II, Pentium III and the Core architecture) is actually a RISC core that is used to "emulate" a CISC architecture running on top of it.

 

[cmaier]

That is more a RISC vs CISC argument. RISC chips need more registers to work compared to a CISC chip. They relay much more heavily on the registers. CISC uses memory a bit more.

PPC was a RISC chip. Pentium is a CISC chip. ARM is RISC.

This is so oversimplified that it probably borders on incorrect. Modern x86 chips (and AMD64 chips) are essentially built like RISC chips, with some additional gunk in the instruction decoder to intercept "complex" instructions and replace them with an appropriate sequence of microcoded RISC-like instructions.

More registers would generally be a good idea for most CISC architectures - it's not that RISC chips inherently need more registers to work. More registers is better for CISC and RISC because using registers can be a thousand times faster than using memory.

The downside is that if you have too many registers, then whenever you have a context switch you need to save them aside, and that can be expensive (either in terms of time or die area, depending on how you implement it). This, for example, is why SPARC uses a giant register file with rotating windows - it's faster to switch to a new window than to save aside all the registers back to memory.

All that being said, x86 would benefit mightily from doubling the number of registers (on average speed would likely be increased around 20%). AMD64 could probably benefit from increasing the number of registers as well, but the number of registers there is much closer to the sweet spot.

The P6 architecture that originated with the Pentium Pro (and went on to be known as the Pentium II, Pentium III and the Core architecture) is actually a RISC core that is used to "emulate" a CISC architecture running on top of it.

I'm going to assume, because of the quotes, that you don't really mean "emulate"

 

[stevep]

once they switch atom to the 22nm process ...

And how long will that take? By which time Arm will have moved the goalposts.

...it will kick ARM's a$$ ...

Arm have already carved out the mobile market for their designs. It would take more than a re-hash of an old chip design to persuade the big players in the tablet market to move away from the Arm roadmap.

...and be just as energy efficient.

I doubt that somehow, but even if they were, the same point above applies.

Atom is already being used in a lot of consumer electronics, just not advertised

And Arm chips are being used in a hell of a lot more. In fact you probably have at least a dozen in your home right now.

Unfortunately for Intel, the future is in low power, mobile, cloud, long battery life, small form factor. At the moment they have nothing for this market, and won't have for at least 12 months. Of course, they could always take out another license from Arm to produce the necessary hardware (I believe they already hold a few Arm manufacturing licenses, though I'm not sure what for).

 

[cmaier]

Unfortunately for Intel, the future is in low power, mobile, cloud, long battery life, small form factor. At the moment they have nothing for this market, and won't have for at least 12 months. Of course, they could always take out another license from Arm to produce the necessary hardware (I believe they already hold a few Arm manufacturing licenses, though I'm not sure what for).

They took over strongarm from DEC, then sold it off to Marvell.

 

[alent1234]

And how long will that take? By which time Arm will have moved the goalposts.

Arm have already carved out the mobile market for their designs. It would take more than a re-hash of an old chip design to persuade the big players in the tablet market to move away from the Arm roadmap.

I doubt that somehow, but even if they were, the same point above applies.

And Arm chips are being used in a hell of a lot more. In fact you probably have at least a dozen in your home right now.

Unfortunately for Intel, the future is in low power, mobile, cloud, long battery life, small form factor. At the moment they have nothing for this market, and won't have for at least 12 months. Of course, they could always take out another license from Arm to produce the necessary hardware (I believe they already hold a few Arm manufacturing licenses, though I'm not sure what for).

22nm is already up and running at Intel's oregon fab. if they are serious about Atom they will start with Atom CPU's as the first ones. biggest problem with Atom now is that it's a generation behind in the manufacturing process which is why it's not energy efficient. nothing to do with the architecture or x86 which is such a tiny part of modern CPU's

 

[Mousse]

As if 7 hours isn't good or something, I know consumers want more, more, more, but somethings gotta give.

It is. It's just Apple probably wants shrink the battery to make their machines thinner and lighter. MBA powered by a pair of AAA. It might happen in my lifetime.

 

[Trudy]

No it wasn't.

Yes, it was, although we may be splitting hairs here. The architecture was designed with an optional 64-bit specification that was compatible with the standard 32-bit specification. The 64-bit specification just wasn't implemented in hardware until the second-generation PowerPC 620 (contemporary with the 603). Because of this foresight, the PPC provided a smooth, fully compatible upgrade path to 64-bit.

Meanwhile, the 8-bit 8008 really began the x86 architecture. They tacked on some 16-bit ops for the the 8085, and rushed a rushed 16-bit version in the 8086. Then memory management was added for the 286 (remember real mode for compatibility that required a CPU reset?), 32-bit extensions were added to it for the 386, and then came 64-bit extensions. It's a kludge.

 

[GS17]

Spin it however you want, but power consumption pretty much dictates how good a machine is these days.

How good a machine for whom?? if you're a professional (read that any profession that uses mail, excel, word, safari, and other low performance apps and are a road warrior and all that) I guess you're right, you need medium or low power and a lot of battery time and like thin laptops.

But if on the other hand you are a Professional (at least video editor, in my case video composer/colorist and CG Artist - Modeler - Simulations - VFX) you need all the power you can get no matter how much it consumes on real desktop machines, no hybrids of laptop parts alas iMac. Faster, More Storage, Better video cards, expansion ports, etc... That's what we want, but I guess in Apple's eyes we are dinosaurs and no longer fall under the Consumer market Apple is after.

Definitely Pixar is not creating their next blockbuster in an iPad2 running Maya on an ARM chip, and they will definitely never will.

 

[KingCrimson]

Geo Metro vs BMW 5-series. I know which one I prefer.