Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

joe-h2o

macrumors 6502a
Jun 24, 2012
997
445
ARM OEMs already thrash Intel in terms of price/performance/power. Its testament to Intel marketing prowess that its able to fool people into thinking otherwise.

Here's a simple question: If no-one can compete with Intel, why does Intel lose $1 billion on each $1 million of sales into the market into which they compete with ARM OEMs?

Intel is a monopoly player with a power hungry legacy architecture. Like all monopolists they aren't defeated so much as become irrelevant. Intel is already irrelevant as parallelism and lower power take over, while Intel doubles down on hotter processors with superior single-thread performance.

Show me the ARM CPU that competes with an i5 on performance.

Intel is in the business of selling non-mobile chips. They have a token presence in the area where ARM chips dominate - as you say, at a loss - but if ARM vendors want to make that step up and start competing in Intel's arena it'll be a different story.
 

nutjob

macrumors 65816
Feb 7, 2010
1,030
508
Well, ARM offers two licenses and up until I think the A6 Apple used the ARM Cortex cores and designed an SoC around that. From then they used the ARM ISA license, and like Qualcomm makes their own micro architecture. At the time Chipworks made some analysis on the core which revealed that it's not only their own design, but that the layout is done by hand which pretty much conclusively proves you wrong.

http://www.chipworks.com/en/technic.../apple-iphone-5-the-a6-application-processor/

In any case, if we leave that behind, there is very little, if anything that can be predicted about performance from the ISA alone.

Layout by hand proves nothing. They're using modules within the design and placing them manually for whatever reason. (As they say in the article "large digital blocks of logic") Designing chips is the same as designing software in one sense: you have libraries and other existing code which you customize and adapt according to the the requirements and specifications.

No-one is re-writing the entire code for ARM ISAs. There is only one sensible way to do some things, many other things you can't beat the ARM engineers' code.

You can think whatever you want, but that's how it's actually done in the real world.

----------

Show me the ARM CPU that competes with an i5 on performance.

Intel is in the business of selling non-mobile chips. They have a token presence in the area where ARM chips dominate - as you say, at a loss - but if ARM vendors want to make that step up and start competing in Intel's arena it'll be a different story.

You have missed the point entirely. Intel may have better single thread performance but it comes at the cost of power usage and cost. Most users think single thread performance is important but it's not. Right now legacy software makes Intel CPUs attractive, but the future is multiple cores and parallelism, because that uses much less power.

So keep on taking as much as you like about "performance" because it's irrelevant. Power and price are much more important. The thing you call "performance" will be fixed in software going forward.
 

subsonix

macrumors 68040
Feb 2, 2008
3,551
79
Layout by hand proves nothing. They're using modules within the design and placing them manually for whatever reason. (As they say in the article "large digital blocks of logic") Designing chips is the same as designing software in one sense: you have libraries and other existing code which you customize and adapt according to the the requirements and specifications.

No-one is re-writing the entire code for ARM ISAs. There is only one sensible way to do some things, many other things you can't beat the ARM engineers' code.

You can think whatever you want, but that's how it's actually done in the real world.

Apple also shipped the first ARM64 chip about a year before ARM them selves had their 64 bit Cortex part ready (Google it) if that did not convince you. This is just hand waving on your part, I have two references you can actually check.
 

nutjob

macrumors 65816
Feb 7, 2010
1,030
508
Apple also shipped the first ARM64 chip about a year before ARM them selves had their 64 bit Cortex part ready (Google it) if that did not convince you. This is just hand waving on your part, I have two references you can actually check.

LOL, Apple used ARMs design! That's how they did it quickly, they licensed it of course.

What they did is like building an application. You use various libraries and APIs from the OS and you write the code that glues it together and provides your functionality.

ARM provided the core logic for the design. To say that Apple did it before ARM is a joke. You're suggesting Apple created something before its creators did. It makes no sense whatsoever.

All Apple did is take the ARM's source code, modified it, combined it with its own and compiled it. It's just a variant or adaption of ARM's design. All Apple did is come out with their own specific design before ARM released a more general design for a wider set of requirements and applications. Apple's job is obviously much easier in that case.
 

subsonix

macrumors 68040
Feb 2, 2008
3,551
79
LOL, Apple used ARMs design! That's how they did it quickly, they licensed it of course.

What they did is like building an application. You use various libraries and APIs from the OS and you write the code that glues it together and provides your functionality.

ARM provided the core logic for the design. To say that Apple did it before ARM is a joke. You're suggesting Apple created something before its creators did. It makes no sense whatsoever.

That's not what I'm suggesting at all, if the instruction set spec is established they can implement it. It's not depending on ARM being done with the micro architecture or their own Cortex cores.

All Apple did is take the ARM's source code, modified it, combined it with its own and compiled it. It's just a variant or adaption of ARM's design. All Apple did is come out with their own specific design before ARM released a more general design for a wider set of requirements and applications. Apple's job is obviously much easier in that case.

This is pure speculation. Apple acquired P.A. Semi, which have designed a 16 core PPC chip, so they also have skilled engineers for this.

P. A. Semi concentrated on making powerful and power-efficient Power Architecture processors called PWRficient, based on the PA6T processor core. The PA6T was the first Power Architecture core to be designed from scratch outside the AIM alliance (i.e. not by Apple, IBM, or Motorola/Freescale) in ten years.
 

nutjob

macrumors 65816
Feb 7, 2010
1,030
508
That's not what I'm suggesting at all, if the instruction set spec is established they can implement it. It's not depending on ARM being done with the micro architecture or their own Cortex cores.



This is pure speculation. Apple acquired P.A. Semi, which have designed a 16 core PPC chip, so they also have skilled engineers for this.

LOL no! A specification is fine, but it's open to interpretation. Everyone releases at least a reference implementation even if they know it will be customised, because it is a concrete implementation of the specification and it resolves any issues or ambiguities.

I know that Apple can design chips, they've been doing that for many years. But no-one with the slightest bit of intelligence or good sense would try to do what you're suggesting (starting from scratch). They'd be fired!
 

subsonix

macrumors 68040
Feb 2, 2008
3,551
79
LOL no! A specification is fine, but it's open to interpretation.

The spec itself is hardly open to interpretation! If it was, how do you imagine one would write assemblers and compilers for it. The implementation of the spec is obviously open to interpretation, that's the point!

Everyone releases at least a reference implementation even if they know it will be customised, because it is a concrete implementation of the specification and it resolves any issues or ambiguities.

I know that Apple can design chips, they've been doing that for many years. But no-one with the slightest bit of intelligence or good sense would try to do what you're suggesting (starting from scratch). They'd be fired!

They can use existing designs and IP obviously, even if it's not from ARM.
 

nutjob

macrumors 65816
Feb 7, 2010
1,030
508
The spec itself is hardly open to interpretation! If it was, how do you imagine one would write assemblers and compilers for it. The implementation of the spec is obviously open to interpretation, that's the point!



They can use existing designs and IP obviously, even if it's not from ARM.

A specification cannot describe everything in sufficient. Why is it that devices that conform to the same specification don't work (well) with each other or are incompatible (eg Bluetooth, WiFi, SQL databases, C compilers). Anyone who's actually written anything according to a specification knows this. Don't take my word for it, google it and see.

They can't use existing designs and IP from someone other than ARM if ARM is the only one who has them, since they just created them! LOL!
 

DesterWallaboo

macrumors 6502a
Sep 7, 2003
520
726
Western USA
Show me an ARM CPU that can compete with a Xeon.... and then convince all of the developers of the software and plugins we use in our studio that they should rewrite/recompile all of their apps/plugins for my new machine.

No thanks.
 

coolspot18

macrumors 65816
Aug 16, 2010
1,051
90
Canada
Mac OS may never be ubiquitous, but iOS already is.

iOS only has <20% of the global market share, it's not as ubiquitous as we like to think it is.

----------

And from where have obtained the performance specs of processors that are years away from development? Can you please link to your source, I'd like to read..

Processor road maps are laid out years in advance - if companies can magically build processors that are as fast as a mid-range Intel I5 processor, someone, especially the Chinese, would have done so already. Processor design takes decades of evolution and refinement ... there are no upcoming ARM processors that will be as fast as a mid-range Intel processor within the next 5 maybe even 10 years.
 

Michael Goff

Suspended
Jul 5, 2012
13,329
7,421
Layout by hand proves nothing. They're using modules within the design and placing them manually for whatever reason. (As they say in the article "large digital blocks of logic") Designing chips is the same as designing software in one sense: you have libraries and other existing code which you customize and adapt according to the the requirements and specifications.

No-one is re-writing the entire code for ARM ISAs. There is only one sensible way to do some things, many other things you can't beat the ARM engineers' code.

You can think whatever you want, but that's how it's actually done in the real world.

----------



You have missed the point entirely. Intel may have better single thread performance but it comes at the cost of power usage and cost. Most users think single thread performance is important but it's not. Right now legacy software makes Intel CPUs attractive, but the future is multiple cores and parallelism, because that uses much less power.

So keep on taking as much as you like about "performance" because it's irrelevant. Power and price are much more important. The thing you call "performance" will be fixed in software going forward.

Intel has better multi-core performance as well.
 

maflynn

macrumors Haswell
May 3, 2009
73,486
43,410
Here's a simple question: If no-one can compete with Intel, why does Intel lose $1 billion on each $1 million of sales into the market into which they compete with ARM OEMs?
Its simple because people are buying less and less PCs, Intel has largely failed to get on the mobile CPU segment (tablets/phones). ARM does this very well, Intel has been about raw performance, with a nod to power saving.

The bottom line for me is that Intel CPUs have not transitioned over to the mobile platform very well, and likewise I think the ARM CPUs will not transition over to the desktop computing segment very well.
 

Keirasplace

macrumors 601
Aug 6, 2014
4,059
1,278
Montreal
iOS only has <20% of the global market share, it's not as ubiquitous as we like to think it is.

----------



Processor road maps are laid out years in advance - if companies can magically build processors that are as fast as a mid-range Intel I5 processor, someone, especially the Chinese, would have done so already. Processor design takes decades of evolution and refinement ... there are no upcoming ARM processors that will be as fast as a mid-range Intel processor within the next 5 maybe even 10 years.

Hope your taking notes here on your prognostic... Because you'll have to come and make amend within this 5 years...
 

coolspot18

macrumors 65816
Aug 16, 2010
1,051
90
Canada
Even Microsoft seems to be agreeing with that now. It's own software seem to be more and more platform agnostic. 2014 is NOT 1990

Windows NT has a long history of multi-platform support, so it's nothing new for Microsoft. Some platforms that it supported/supports include: IA-32, x86-64, DEC Alpha, MIPS, PowerPC, ARM, Itanium... So the fact that Windows now supports ARM is not out of the ordinary for MS.


But if Microsoft is making Windows compatible with ARM (which I think they are - isn't Windows 8 and Office able to run on ARM?) then this may not be a problem after all.

There are pretty serious rumors that Microsoft is killing ARM support - sales of Windows RT have been lacklustre to say the least.
 

Michael Goff

Suspended
Jul 5, 2012
13,329
7,421
Windows NT has a long history of multi-platform support, so it's nothing new for Microsoft. Some platforms that it supported/supports include: IA-32, x86-64, DEC Alpha, MIPS, PowerPC, ARM, Itanium... So the fact that Windows now supports ARM is not out of the ordinary for MS.




There are pretty serious rumors that Microsoft is killing ARM support - sales of Windows RT have been lacklustre to say the least.

No, they're not killing ARM support, they're combining tablet and phone.

"Right now tech preview is for x86 PCs. ARM devices and phones, not 'til well into 2015," Belfiore said in a reply to a fan's tweet. "The purpose of going out this early on the PC is largely to get enterprise/corporate feedback. More later," he said in another reply.

ARM Windows isn't dead.
 

Michael Goff

Suspended
Jul 5, 2012
13,329
7,421

SlCKB0Y

macrumors 68040
Feb 25, 2012
3,426
555
Sydney, Australia
Did you have a problem running your PPC stuff under Rosetta? I didn't.

Do you understand how much CPU power that required? By the time Apple switched, Intel CPUs were easily 2x faster than their G5 counterparts.

The fastest ARM is somewhere between an Atom and a lower-power i3 (CPUs that low-end that Apple don't even make computers using them).

So how is your "Rosetta 2" going to work again?

----------

So, yes, it will be upgraded. It won't get everything, and I doubt anyone expected it to. But it isn't dead.

Wow, you could possibly be the biggest Microsoft apologist i've encountered.

If I bought an RT device a year ago I would expect it to get the latest OS when released, not some half-baked update with some features.

The platform is most certainly doomed... it's just that Microsoft choose not to put it out of its misery yet.

Hmmm.... sounds familiar..Windows Phone 7 anyone?
 

Michael Goff

Suspended
Jul 5, 2012
13,329
7,421
Do you understand how much CPU power that required? By the time Apple switched, Intel CPUs were easily 2x faster than their G5 counterparts.

The fastest ARM is somewhere between an Atom and a lower-power i3 (CPUs that low-end that Apple don't even make computers using them).

So how is your "Rosetta 2" going to work again?

----------



Wow, you could possibly be the biggest Microsoft apologist i've encountered.

If I bought an RT device a year ago I would expect it to get the latest OS when released, not some half-baked update with some features.

The platform is most certainly doomed... it's just that Microsoft choose not to put it out of its misery yet.

Hmmm.... sounds familiar..Windows Phone 7 anyone?

You must not get out much. I'm not apologizing, it sucks. Heck, I even have an RT device. But it isn't like we know what features it's going to get, and every single OS in the world does this. Will the 6+ get every feature of iOS 9? Likely not, but I'm okay with that. It's just the nature of technology.
 

Renzatic

Suspended
You have missed the point entirely. Intel may have better single thread performance but it comes at the cost of power usage and cost. Most users think single thread performance is important but it's not. Right now legacy software makes Intel CPUs attractive, but the future is multiple cores and parallelism, because that uses much less power.

So keep on taking as much as you like about "performance" because it's irrelevant. Power and price are much more important. The thing you call "performance" will be fixed in software going forward.

The hell you say. Let's do some math for a second...

1 core = 10 watts

1 core completes a process in 10 seconds. 10x10 That's 100 joules of energy, give or take.

Now take two of those same cores working on that same process in a multithreaded environment. Assuming each is able to an equal amount of work, it'd take 5 seconds for them both to do.

2 cores = 20 watts
Process = 5 seconds.

= 100 joules

Or a quad core configuration...

4 cores = 40 watts
Process = 2.5 seconds

= 100 joules.

You're completing the task faster, which will give you power saving over time, but this assumes one thing that rarely, rarely, RARELY ever happens in the real world...

...that each added core gives you an exponential increase in performance. Parallel processes are very difficult to write, and most applications have absolutely no need for it, with each thread added giving you slim gains at the cost of exponentially increased overhead.

So yeah, single thread performance is still, and always will be very important. A 5w core that takes three times as long to perform a process is going to give you a net loss in battery life compared our 10w, despite being lower powered. ARM chips aren't so much better than Intels that they can do considerably more with considerably less. I'd say the absolute best case scenario is that they're roughly average, with a 5w ARM chip doing a process in 20 seconds that a 10w Intel chip can do in 10.

So you're wrong.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.