Why don't intel just buy Qualcomm instead of allowing them to become the major threat

[shady28]

But Apple would still be making new designs for their own SoCs, so there'd be no justification for them to dismantle ARM's CPU business. I think this would fall under attempt to monopolize via willful acquisition.



There's still a large swath of the Android SoC vendors that have the capability to develop their own CPUs (Qualcomm, Samsung, and nVidia at the very least), and they won't need an update to the architecture any time soon, so I don't think they'd all be too quick to flock to Intel.

Well, in one paragraph you're saying Apple would be monopolizing. In the other you're saying that the other vendors would develop their own CPUs and it would not be a monopoly. Anti-trust laws are for preventing monopolies, not preventing someone (Apple) from going from 12% of the market to 30%.

To be clear, this is worldwide smartphone sales. Hardly a monopoly situation :

I think the ARM SoC vendors, after 4 or 5 years, would either have a fractured set of "ARM based" platforms which would greatly harm android, or they would fall behind and slowly fade into obscurity (see x86 licensees, the only one of note left is AMD).

If I were a phone manufacturer and I saw a competitor acquire ARM, I think my immediate reaction would be to run to one of ARMs competitors. In this hypothetical example, that would be Intel.

 

[Qwertilot]

Think you'd just find some very similar alternative to ARM appearing if they vanished somehow. They're just so convenient to have around for so many companies. Fundamentally replaceable though, which stops them getting too greedy.

Apple could always satisfy themselves by buying a few small countries

 

[Exophase]

Well, in one paragraph you're saying Apple would be monopolizing. In the other you're saying that the other vendors would develop their own CPUs and it would not be a monopoly. Anti-trust laws are for preventing monopolies, not preventing someone (Apple) from going from 12% of the market to 30%.

I meant to say that other vendors would eventually recover, and not go to Intel, at least not permanently. None the less, such an acquisition on Apple's part would still clearly be nothing more than anti-competitive, and thus of intent to monopolize.

I think the ARM SoC vendors, after 4 or 5 years, would either have a fractured set of "ARM based" platforms which would greatly harm android, or they would fall behind and slowly fade into obscurity (see x86 licensees, the only one of note left is AMD).

What are you envisioning that would fracture and harm Android more than it already is? Arch fractures, uarch doesn't (at least not really).

If I were a phone manufacturer and I saw a competitor acquire ARM, I think my immediate reaction would be to run to one of ARMs competitors. In this hypothetical example, that would be Intel.

If I were a company licensing ARM CPU IP to make in custom SoCs I'd be running to MIPS, if I had to run somewhere.

 

[shady28]

I meant to say that other vendors would eventually recover, and not go to Intel, at least not permanently. None the less, such an acquisition on Apple's part would still clearly be nothing more than anti-competitive, and thus of intent to monopolize.



What are you envisioning that would fracture and harm Android more than it already is? Arch fractures, uarch doesn't (at least not really).



If I were a company licensing ARM CPU IP to make in custom SoCs I'd be running to MIPS, if I had to run somewhere.

What I'm envisioning is a ton of binary translation. The performance impact of that is devastating to many android applications - specifically the ones that go to C/C++ native code for performance reasons. This is why reviews of x86 Android tablets always seems to note some kind of occasional performance issue. About 1/4 of the top 100 Android apps have native code in them.

And if the thought is that we don't need native apps on android and can just rely on Java :

 

[Exophase]

What I'm envisioning is a ton of binary translation. The performance impact of that is devastating to many android applications - specifically the ones that go to C/C++ native code for performance reasons. This is why reviews of x86 Android tablets always seems to note some kind of occasional performance issue. About 1/4 of the top 100 Android apps have native code in them.

Yes, but you'll get that whether you go with x86 or go with MIPS, so you have to pick your poison - and MIPS is still the one with the licensing model more like ARM's. And trust me I know, my app is NDK and the cost of using binary translation to x86 is huge, much worse than the averages given in that table (probably because it's NEON heavy)

On the other hand, you're missing the fact that for a few years now NDK has, by default, included x86 binaries as well. It's a mystery why old so many apps today still lack it, although it's possible that ARM's analysis was due to a flaw in Samsung Galaxy Tab 3 10.1 that prevented native x86 binaries from being used over translated ARM ones.

Or was this your explanation for how these different "ARM based" platforms would hurt Android? If they're still using the ARM arch no translation would be involved. I doubt anyone's in a serious crunch to expand ARMv8 at all costs, and if they were they could possibly come to a consensus on extensions. But just look at ARMv7, it was standard for over 5 years, and ARMv8 was really only necessitated by the need for 64-bit.

If someone really acquired ARM and shut down future cores I think what you might see is some other company start licensing their custom ARM core like ARM themselves have, assuming that the architectural license allows this.

 

[shady28]

Yes, but you'll get that whether you go with x86 or go with MIPS, so you have to pick your poison - and MIPS is still the one with the licensing model more like ARM's. And trust me I know, my app is NDK and the cost of using binary translation to x86 is huge, much worse than the averages given in that table (probably because it's NEON heavy)

On the other hand, you're missing the fact that for a few years now NDK has, by default, included x86 binaries as well. It's a mystery why old so many apps today still lack it, although it's possible that ARM's analysis was due to a flaw in Samsung Galaxy Tab 3 10.1 that prevented native x86 binaries from being used over translated ARM ones.

Or was this your explanation for how these different "ARM based" platforms would hurt Android? If they're still using the ARM arch no translation would be involved. I doubt anyone's in a serious crunch to expand ARMv8 at all costs, and if they were they could possibly come to a consensus on extensions. But just look at ARMv7, it was standard for over 5 years, and ARMv8 was really only necessitated by the need for 64-bit.

If someone really acquired ARM and shut down future cores I think what you might see is some other company start licensing their custom ARM core like ARM themselves have, assuming that the architectural license allows this.

That's just it. I think you're right - on all counts. MIPS might get more licensees, someone would make enhanced ARM derivatives and license that out, x86 would get more of the market, etc. etc. etc.

ie, fractured.


Maybe put this in a bigger picture. Where would Android be today if someone had bought ARM 6 years ago and had then frozen ARM development at the Cortex-A8?

Would android phones even still be running on ARM?

And how would they compare to an iPhone 6?

 

[R0H1T]

That's just it. I think you're right - on all counts. MIPS might get more licensees, someone would make enhanced ARM derivatives and license that out, x86 would get more of the market, etc. etc. etc.

ie, fractured.


Maybe put this in a bigger picture. Where would Android be today if someone had bought ARM 6 years ago and had then frozen ARM development at the Cortex-A8?

Would android phones even still be running on ARM?

And how would they compare to an iPhone 6?

It's just a case of ARM being at the right place & just at the right time, kinda like Wintel & how the dominance of Windows lead x86 to where it is right now. Android could just as well have been launched on MIPS but the fact that there weren't any smartphones, at least not that I know of, running on MIPS processors at that time didn't help its cause.

 

[jhu]

Maybe put this in a bigger picture. Where would Android be today if someone had bought ARM 6 years ago and had then frozen ARM development at the Cortex-A8?

Would android phones even still be running on ARM?

And how would they compare to an iPhone 6?

If the agreements were all to remain intact, I don't think much changes. Back in 2009, most Android phones were running Qualcomm chips (my HTC Aria has Qualcomm's MSM7227 chip in it). Most Android phones now are running Qualcomm chips (even many of Samsung's phones), and Qualcomm's been designing their own ARM CPUs since at least Scorpion.

 

[scannall]

Apple approached Intel first about a SOC for the as yet unreleased iPhone, and Intel turned them down. That would have been the easiest route for Apple, since OS X already runs on x86, and iOS is just a subset of OS X. It wasn't too long after that that they started assembling their own SOC team. The smartphone SOC landscape would certainly look a lot different if Intel had taken Apples offer.

 

[Sheep221]

Apple approached Intel first about a SOC for the as yet unreleased iPhone, and Intel turned them down. That would have been the easiest route for Apple, since OS X already runs on x86, and iOS is just a subset of OS X. It wasn't too long after that that they started assembling their own SOC team. The smartphone SOC landscape would certainly look a lot different if Intel had taken Apples offer.

Everything that runs iOS is ARM based.

 

[scannall]

Everything that runs iOS is ARM based.

Yes, that is correct. Meaning they had to update Xcode to do ARM as well as x86. Had Intel accepted the job to make SOC's for Apple, that wouldn't have been necessary.

 

[jhu]

Everything that runs iOS is ARM based.

Except the iOS simulator. That's x86.