Apple to expand CPU design group beyond iPad A4 - Page 2

Originally Posted by SockRolid

Not sure which universe you're in, but in this universe there were several Windows emulators for PowerPC Macs. Google "windows emulator powerpc mac" for some examples.

Originally Posted by NasserAE

MS Windows is made to run on any PC and Macs are PCs. Even when Macs used PowerPC you could have Windows installed through Virtual PC. Right now we have Parallels, VMWare, and Sandbox VM. This is not an issue.

Originally Posted by technohermit

I get more use out of Fusion than I ever did of Bootcamp. Just saying that it might not be the doom and gloom you are trying to put out there.

Originally Posted by Tsadler

Um, ok so basically you're saying you won't be able to install Mac OS on unsupported devices. Who, besides a fraction of a percentage, gives a damn?

Have you ever owned a Mac? Do you know what Bootcamp is? Get educated.

Originally Posted by solipsism

So are we talking about all this cost and effort to prevent a few Hackintoshes? And I cant see how allow Windows as dual-boot or VM has increased Mac sales. So the reason for Apple working on its own chip designs tells me that its something considerably more profound that what you seem to be inferring.

Originally Posted by herbapou

I fail to see any mention in the article that would suggest this ARM design will be use in Mac's. Its for mobile iOS devices...

Originally Posted by wizard69

The whole reason Macs have been so successful as of late is that they have no restrictions on what can run on the machines. Why would Apple screw with that?

Even though I tire of dilger writing style he was pretty clear that this was about mobile device.

Originally Posted by wizard69

I think you missed something here, no one has indicated that they are getting rid of i86 hardware in Macs.

Originally Posted by SockRolid

Exactly what I'm thinking. $278 million for PA Semi plus $121 for Intrinsity is too much to spend on a slightly warmed-over single-CPU ARM design. The purchases prevent competitors from acquiring that IP and engineering talent, but that's only a small bonus.

I think Apple's long-term goal is to create their own proprietary SoC for not just their iDevices but for Macs as well.

This would help lower their hardware costs since they won't be paying off-the-shelf prices for one of the most expensive components in their products. And lower hardware costs will help Apple maintain their margins.

But there are two more gigantic benefits. First, Apple could conceivably transition Mac OS back from Intel to their custom multi-core ARM. They have already transitioned Mac OS through several CPU changes: 68k to PowerPC, then PowerPC to Intel. Been there, done that, got the developers to come along too. And that could enormously benefit iDevices in the next decade. Eventually mobile device CPU power will exceed that of today's desktop computers, and Apple could prepare for that future by transitioning Mac OS to run on their ARM-based mobile CPUs.

Second, and this is perhaps the most important benefit for Apple, using a custom ARM chip on all their computing products would free them from dependence on an outside chip designers. For decades, from the 6502 to Intel Core i7, Apple has been at the mercy of the Motorolas, IBMs, and Intels of the world. Each of which have different goals than Apple. Motorola and IBM were more concerned with the embedded versions of their PowerPC chips than efficient and speedy desktop and laptop computer versions. Intel is more concerned with optimizing Windows performance than anything else. (And the CISC design of their CPUs uses vast areas of silicon for the execution of obscure backward-compatible x86 instructions generated only by Microsoft's compilers.)

None of those chip makers really wants to build a bespoke chip just for Apple. Intel, just after Apple completed the PowerPC-to-Intel transition, gave Apple their newest chips first. The original MacBook Air had an avant-garde chip that eventually was used in other laptops. But that was presumably because Apple paid them for that privilege, an unsustainable tactic, and the honeymoon ended.

Apple could eliminate their co-dependence on other chip designers now that they have acquired PA Semi's and Intrinsity's intellectual property. And that will set them up for their next decade or two of innovation. No other tech company in the world will be as well prepared.

Originally Posted by Prof. Peabody

I used some colloquial language, but the meaning was clear. There's no need to be so dismissive.

The last I heard, Apple has a license to customise the Cortex A8 and A9 and so far, all they have done is do some very minor tweaks on the A8, put it in an SoC and called it the A4. My argument was only that given the IP they have, the acquisitions they have made, and the talent they acquired thereby, it makes sense that the A5 or whatever the next chip is called might be a more customised version of the actual core silicon.

If what you are saying here is accurate:


Then I'm off by a year or two and it might take until the A6 to see some real differences between Apple's silicon and the competition.

Originally Posted by Morky

Does anyone know whether it's possible that Apple could hard wire in silicon certain functions from their core libraries for huge performance gains? That could serve as one of the reasons they are pushing developers to use only Apple's tools.

Originally Posted by wizard69

Lets face it they need to deliver far better hardware on iPad simply due to the competition. It would be pretty pathetic if iPad 2 arrived with a single core processor.

Originally Posted by solipsism

Seriously?! Youre now arguing that you cant have gradation in how much something has been optimized or customized?

Originally Posted by solipsism

You keep saying the iPad is slow and that it cant keep up with the competition, yet everything coming from the iPad introduction stated was how surprisingly fast it felt for an ARM-based system and no other tablet OS competitor on the market has anything comparable, only minimal changes to Android 2.x.

If you honestly think that the iPad 2 (which is likely to ship months before PlayBook) has to use a much faster CPU than other tablets using ARM processors despite power management, drivers, apps, software and other aspects that can affect usability then youre in for a shock.

Originally Posted by MacTel

We're all hopeful that the A4 isn't the end-of-the-line for Apple's processing interests. A family of A4s with varying clock speeds and then a family of A5s with new architecture that allows mulitple cores would be ideal.

It would be most telling if the Lion OSX version supports the use of ARM processors. That would solidfy the idea that Apple will take the processor into their Mac lines. With the XServe gone (why???) there's not as much need for high-powered processors. The ARM architecture will eventually catch-up in performance with Intel server offerings (read 3 years time).

Last month TI announced it was the first to license ARMs next-generation Eagle core. Today, ARM is announcing the official name of that core: its the ARM Cortex A15.

Architectural details are light, and ARM is stating that first silicon will ship in 2012 at 32/28nm. Heres what we do know. The Cortex A15 will be a multi-core CPU, designs can have as few as a single core but most will have 2 - 4 cores depending on their target market.

ARM's latest processor core announcement was the Cortex-A15, previously codenamed Eagle. The A15 complies with the ARMv7 instruction architecture but with support for 40-bit virtualization. The next ARM processor to be announced will support 64-bit and could be unveiled as soon as next week, the report said.

One possibility - which would be very reminiscent of Intel's marketing style - would be if the basic Cortex-A15 design already supports 64-bit processing and ARM has quietly kept that detail back from the original announcement to give it more publicity. As ARM is licensor of IP it might be possible to allow chip partners to choose whether to opt for full 64-bit processing or opt for 32-bit depending on application and as they are designing their implementation of the chip.

Originally Posted by wizard69

IPad is slow as it is. The competition is a problem because many of the coming tablets will have either dual core ARM processors or ATOM processors in them.

Originally Posted by wizard69

.Actually I think it needs a much faster processor just so it can begin to reach some of the goals Apple envisions for iPad. Yes I believe Apple, (Steveo and company) have a vision as to what iPad can be. iPad 1 isn't even close to what they have planned long term.

Originally Posted by wizard69

It looks like their intentions are to supplement the ARM instruction set with instructions that accelerate the execution of Objective C.

Originally Posted by Dick Applebaum

Maybe he's been running overclocked and is suffering stability issues

Originally Posted by wizard69

IPad is slow as it is. The competition is a problem because many of the coming tablets will have either dual core ARM processors or ATOM processors in them.

However it is not that competition that makes iPad seem slow, rather it does that on its own when anything non trivial is attempted on the platform. The problem is when these tablets are [out] it will become very evident as to how slow iPad is.

iPad 1 isn't even close to what they have planned long term.

Originally Posted by Tallest Skil

Here we go. What better way to completely shut the user out of the computer than making the whole thing proprietary? Mac OS will only work on Apple's architecture (read: Apple computers), and Windows, et. al. won't ever be installable because they'll have no need to build Apple versions.

It'd take forever, but it may happen.

Originally Posted by wizard69

Actually I think it needs a much faster processor just so it can begin to reach some of the goals Apple envisions for iPad. Yes I believe Apple, (Steveo and company) have a vision as to what iPad can be. iPad 1 isn't even close to what they have planned long term.

Originally Posted by ascii

It seems to me they're dreaming if they think they can do better than a dedicated CPU manufacturer. Their current approach of tweaking existing designs seems like the best of both worlds. But I guess Apple can afford to try these kind of experiments and if they lose a few hundred million it's not too life threatening.

Originally Posted by solipsism

If we consider a five or even a ten year time frame for Apple making an x86-compatible chips that leverages their OS and we use the average YoY average growth rate for the last five years how would that compare to the number of CPUs AMD currently sells? I couldnt find how many CPUs they sell but they noted in October that they sold 25M DirectX11-capable GPUs over the past year, a number that Apple may match n 2011 in the number of Macs sold, so why couldnt Apple also make a CPU?

Originally Posted by ascii

It seems to me they're dreaming if they think they can do better than a dedicated CPU manufacturer. Their current approach of tweaking existing designs seems like the best of both worlds. But I guess Apple can afford to try these kind of experiments and if they lose a few hundred million it's not too life threatening.

Originally Posted by esummers

I think that multi-core ARM chips will one day make it to the Mac. It is more work on the software side to make applications use multiple cores effectively, but Apple has made progress in this and I'm sure there is a lot more coming. I'm very interested to see if there will be more multi-core programming techniques coming out with Lion. They already have Grand Central Dispatch and OpenCL. When you look at the execution unit density of ARM vs Intel or AMD, the ARM chip is capable of a lot more operations per second in the same size die. It just can't perform a single stream of operations as fast, so things need to operate more in parallel on the software side. I think that Apple sees this as the future. Intel is already going multi-core because they hit the ceiling. If you are going multi-core, it makes more sense in the long run (when you don't have legacy software) to have chips with denser execution units like the ARM. The other way things can go in the future is more execution units customized for a particular task. The ARM design still makes sense as part of the stack in that case.

Originally Posted by TenoBell

You do realize that you are arguing that what is available right now is inferior compared to something that is imaginary and exists in the future?

I don't see a general consensus out in the market that the iPad is slow. Yes indeed it will be slower than the newer technology that is not here yet. But that is the true for any technology that exists right now.

Originally Posted by TenoBell

Again this is the state of evolution of all technology.

Originally Posted by Dick Applebaum

Dave


Can you amplify on that?

I can't remember where (my mind is long past overflowing) but I read that Apple was planning on using some technique or construct that allowed direct execution of Objective C code (without compilation or linking?).


TIA Dick

BTW, how goes the battle?

Originally Posted by christopher126

...

BIGGEST DISAPPOINTMENT: Tetris...talk about not taking advantage of the whole screen an washed out colors! Certainly not using the Retina Display to its advantage...very disappointed!

Originally Posted by Quadra 610

A natural move. Apple is in control of their own destiny these days. They can go as proprietary and as in-house as they like, with the assurance that consumers will buy.

Originally Posted by wizard69

It looks like their intentions are to supplement the ARM instruction set with instructions that accelerate the execution of Objective C.

Originally Posted by Dick Applebaum

Dave
Can you amplify on that?

I can't remember where (my mind is long past overflowing) but I read that Apple was planning on using some technique or construct that allowed direct execution of Objective C code (without compilation or linking?).


TIA Dick

BTW, how goes the battle?

Originally Posted by wizard69

You have to dig through the various patents but for some reason Apple has been patenting techniques for enhancing microprocessors over the recent years. I don't have any links handy but I believe the patents came around 2008 or 2009.

What I find interesting here is that Apple has been doing any research at all into this level of CPU architecture.

Apple is Working on Multi-Core Processor Snoop Filtering for Macs

One common cache coherence technique involves bus snooping, in which processors broadcast memory references to each other on a dedicated bus so that data can be transferred between caches rather than accessing main memory. While bus snooping may enable cache coherence, bus snooping may also consume resources, such as power and time, and thus may reduce processor efficiency. Moreover, as the number of processors in a multi-core or multiprocessor system increases, the amount of snooping and broadcasting may increase exponentially, reducing the efficiency of such systems accordingly.

We'll agree to disagree on Sandy-Bridge as it's design doesn't remotely resemble the ARM schematic that Apple has lifted in this patent for presentational intentions, not to mention Apple won't be designing custom controllers and patenting them to use within the Intel IP vicinity, nor with AMD for that matter.

This is a mirror of the ARM Cortex-A9 Multicore 4CPU design.



Apple has access to this Hardware IP, because they paid heavily to license it [Intel did as well which probably explains why their Atom is trying to match the ARM] and it's reasonable this patent is w/ regards to Darwin leveraging this design for the iOS Platform and their future numbered revisions of the A# Processor that includes their own GPU integrated via IMGTech.

I like how they've generalized the patent for future desktop, laptop or other general purpose systems, but it seems rather clear it's immediate impact is for the A# [A5, A6 or whatever they call it] Apple Processor of the future with 4 Cores; hence the ARM A-9 Cortex as baseline to draw upon.

Posted by: Marc J. Driftmeyer | September 17, 2010 at 07:06 PM[/IMG]

Originally Posted by wizard69

It looks like their intentions are to supplement the ARM instruction set with instructions that accelerate the execution of Objective C.

Originally Posted by Hiro

I would turn this around a little. Changing and adding instructions can be a huge undertaking on the order of having to redesign an entire core because of the domino effect the new transistor placements would drive.

A very limited version of changing instructions could be a combination of a compiler and decode stage change. There you can break up a new "native" instruction into it's core native components for issue. That's still a hard task, but at least the domino effect is limited to a small portion of the CPU footprint. And in the end I don't know if the extra speed you would get from the decoder change would be fast enough to justify the resources spent compared to just a compiler driven change. It would take some serious low level analysis and simulation to answer a question like that with any certainty.

So I think the turnaround is that Apple has the ability on an instruction-by-instruction basis to determine if it is necessary to do a decoder+compiler change or just a compiler change, but even more profitably Apple can make adjustments in the SoC (low risk compared to messing with the core) to optimize I/O and graphics based on the same kind of iinstruction-by-instruction analysis. It is an advantage none of the other producers have or in the short to medium term could hope to match. The other guys are all producing and selling silicon to many clients, so a laser focused optimization just doesn't make sense for them.

Originally Posted by Smallwheels

The smart way to run a company is to have the ability to create something from raw materials to the finished product and have the production capabilities to do it. By doing that there would be nothing that could stop it from creating products it wants.

Some business models have all of their design and manufacturing done by others. The people in the office just do the marketing and billing. Apple is half way between the two. They design things but don't manufacture anything.

Apple needs to take the next step with its chips and create their own. They're starting out with the A4 and perhaps with their new team will get good enough at it to design chips from scratch. They've got to start somewhere.

I wish Apple would actually manufacture things. They could start off by letting Americans assemble iPods and iPhones. Then they could have Americans manufacture the cases to their computers and idevices in the USA. In time Apple could manufacture their own circuit boards and chips in the USA. With the premium prices of Apple products the company could afford to do these things within the USA. Just the good public relations alone would get them more sales from Americans who want to support US manufacturers.

If Foxcon employs one million people in China then Apple could eventually move that production to the USA and create at least that many jobs.

Originally Posted by Smallwheels

The smart way to run a company is to have the ability to create something from raw materials to the finished product and have the production capabilities to do it. By doing that there would be nothing that could stop it from creating products it wants.

Some business models have all of their design and manufacturing done by others. The people in the office just do the marketing and billing. Apple is half way between the two. They design things but don't manufacture anything.

Apple needs to take the next step with its chips and create their own. They're starting out with the A4 and perhaps with their new team will get good enough at it to design chips from scratch. They've got to start somewhere.

I wish Apple would actually manufacture things. They could start off by letting Americans assemble iPods and iPhones. Then they could have Americans manufacture the cases to their computers and idevices in the USA. In time Apple could manufacture their own circuit boards and chips in the USA. With the premium prices of Apple products the company could afford to do these things within the USA. Just the good public relations alone would get them more sales from Americans who want to support US manufacturers.

If Foxcon employs one million people in China then Apple could eventually move that production to the USA and create at least that many jobs.

Originally Posted by SpamSandwich

Until the cost to manufacture and assemble in the US falls below that of China this will never happen. Getting rid of the minimum wage and giving US manufacturers a tax holiday could be one way to increase US-based production, but in the age of taxation without representation I don't see this happening.

Originally Posted by Tallest Skil

You're blindly missing the point.

OS X may not be available for installation on non-Apple computers, but it's possible.
Windows may not be made for installation on Macs, but it's possible.

If Apple makes its own architecture, it will be physically impossible within the fundamental laws of the universe for it to work.

Originally Posted by mdriftmeyer

Designing a the CPU architecture from scratch is pointless and futile. Licensing the IP to improve upon the officially designed and released architecture specs from ARM is a wise investment.

Originally Posted by Quadra 610

A natural move. Apple is in control of their own destiny these days. They can go as proprietary and as in-house as they like, with the assurance that consumers will buy.

Further control = further strengthening of the elements that make for a superior User Experience.

Originally Posted by Hiro

I would turn this around a little. Changing and adding instructions can be a huge undertaking on the order of having to redesign an entire core because of the domino effect the new transistor placements would drive.

A very limited version of changing instructions could be a combination of a compiler and decode stage change. There you can break up a new "native" instruction into it's core native components for issue. That's still a hard task, but at least the domino effect is limited to a small portion of the CPU footprint. And in the end I don't know if the extra speed you would get from the decoder change would be fast enough to justify the resources spent compared to just a compiler driven change. It would take some serious low level analysis and simulation to answer a question like that with any certainty.

So I think the turnaround is that Apple has the ability on an instruction-by-instruction basis to determine if it is necessary to do a decoder+compiler change or just a compiler change, but even more profitably Apple can make adjustments in the SoC (low risk compared to messing with the core) to optimize I/O and graphics based on the same kind of iinstruction-by-instruction analysis.

It is an advantage none of the other producers have or in the short to medium term could hope to match. The other guys are all producing and selling silicon to many clients, so a laser focused optimization just doesn't make sense for them.