apple is buying ARM?!!!!!

[OBLAMA2009]

this would really suck. do you think maybe they're thinking about buying AMD as well?

http://www.theregister.co.uk/2010/04/21/apple_arm_acquisition_rumor/

 

[Vesku]

Would Apple continue ARM's current business model of licensing to all? I just don't see that fitting with Apple's style.

 

[Lonyo]

I doubt anyone would want to be at the behest of Apple, so ARM as a licensing company would disappear. No one would want to license an Apple controlled ARM architecture, even if Apple did want to keep it going.

Of course, I can think of many other companies who wouldn't want to see it happen, and a massive bidding war could potentially ensue.

 

[Sylvanas]

The Apple juggernaut is really on a roll recently. This would really shake things up.

 

[OBLAMA2009]

what do you guys think about apple buying amd. there were rumors that amd people were seen on the apple campus and everyone thought it was because apple might use their chips, but in light of this do you think maybe they might be interested in buying amd. it would make sense for both companies

 

[Lonyo]

what do you guys think about apple buying amd. there were rumors that amd people were seen on the apple campus and everyone thought it was because apple might use their chips, but in light of this do you think maybe they might be interested in buying amd. it would make sense for both companies

Maybe except for the x86 license transfer issue.
Sure AMD can now fab x86 chips on other people's fabs without worry, but that doesn't mean a company which buys AMD suddenly gets the x86 license.

 

[Hacp]

Apple and AMD don't make sense. Maybe Apple and Global foundries. AMD, the chip company is valuable only b/c they have an x86 license. Apple doesn't really need x86 for osx.

 

[Lonyo]

Apple and AMD don't make sense. Maybe Apple and Global foundries. AMD, the chip company is valuable only b/c they have an x86 license. Apple doesn't really need x86 for osx.

Apple and GF makes less sense than Apple and AMD.
Apple and Hynix would make more sense, especially since they are, or at least were, for sale.

 

[Fox5]

Considering how dependent the entire world is on ARM processors, I couldn't see regulators (remember, it's a British company) allowing this to go through. Apple may acquire stock in ARM though, but ARM is too important now to let the ISA be closed.

On the other hand, Microsoft and Google are both lucky to have used interpreted code for their phone platforms, they can switch to another architecture if need be. X86 is pretty much the only other game in town, unless Freescale starts pumping out low power PPC chips.
A move like this would probably hurt cheap dumb-phones more than anything else (and aren't a lot of those based on java mobile anyway?).

 

[pjkenned]

Why would they? Sure Apple uses ARM architecture, but they also use x86. If Apple were to buy into a big architecture, it would leave them back in the PPC days almost because there would be a huge investment in an architecture that may end up at a disadvantage one day. Plus, selling components is not a super high margin business like software and designing sleek cases/ UI's.

There's a reason Dell/ HP/ Acer don't make disk drives even though they buy huge amounts of them.

 

[Fox5]

Why would they? Sure Apple uses ARM architecture, but they also use x86. If Apple were to buy into a big architecture, it would leave them back in the PPC days almost because there would be a huge investment in an architecture that may end up at a disadvantage one day. Plus, selling components is not a super high margin business like software and designing sleek cases/ UI's.

There's a reason Dell/ HP/ Acer don't make disk drives even though they buy huge amounts of them.

The iphone is apple's future, and buying ARM would shut out/raise costs on a lot of competitors.
(though I would think ARM architecture licensees, like Marvell, Qualcomm, and Samsung would be rather immune to this, just companies that license the cores directly would be hurt, and diverging ARM instruction sets would hurt compatibility in the industry)

It would also make Intel pretty happy, and Apple does like to have complete control of their platform. Still, I can't see this being allowed to happen even if Apple is interested.

 

[ElFenix]

did apple ever sell off its original stake in ARM? they owned 43% of it when acorn spun it off (acorn retained 43% as well).


edit: apparently so

 

[pjkenned]

The iphone is apple's future, and buying ARM would shut out/raise costs on a lot of competitors.
(though I would think ARM architecture licensees, like Marvell, Qualcomm, and Samsung would be rather immune to this, just companies that license the cores directly would be hurt, and diverging ARM instruction sets would hurt compatibility in the industry)

I would venture to guess (actually having read some ARM license agreements) that existing products would be fairly insulated. If Apple did this, Intel would have a serious x86 competitor, or a big guy would buy a startup and supply the rest of the industry withing 1-2 product cycles. Apple competitors would not use ARM IP so production of cores with ARM IP would drop off a cliff and Apple would be looking at a large writeoff of the acquisition value.

There would probably be a secondary request in the US and probably some interest from EU regulators on the acquisition, despite the fact that you could claim ARM has healthy x86 and other competition. Regulator interest on a deal like this would probably hold up the closure by 6+ months beyond what would probably be a 4-6 month minimum. In the cell phone land, that is a lot of time for a next-gen product cycle.

Fun to dream, but really tough to execute.

 

[EarthwormJim]

If ARM dissapears, really the only alternative would be Atom from Intel. Even then Atom is a couple of generations away from being power usage competitive.

 

[jhu]

If ARM dissapears, really the only alternative would be Atom from Intel. Even then Atom is a couple of generations away from being power usage competitive.

What about MIPS? My WRT54GL is using a MIPS processor.

 

[Fox5]

What about MIPS? My WRT54GL is using a MIPS processor.

MIPS hasn't had serious development on a performance solution (the types high end cell phones/small computer devices need) in a while, it's an option, but it'd take some time to more toward it.

 

[Modelworks]

I'm a huge fan of ARM and a ARM developer as well. The thought of Apple controlling the company makes me physically ill. ARM strength is in the way the company is set up. They allow basically anyone to design a processor using the technology, even a hobbyist at home if they want to, they make their development kits open to the public. An engineer is free to add or remove features for what they need. If apple gets control they will steer the company away from that into what apple needs and it will become nothing more than another intel or amd where the cpu they ship is the only version that exist. All that flexibility and freedom in the design will be lost.

 

[Modelworks]

What about MIPS? My WRT54GL is using a MIPS processor.

MIPS has a very convoluted design in how the processor functions. When you start working on MIPS systems you basically have to forget everything you learned about any other architecture and start over , code, interfacing , it is all different . ARM while different from x86 has a lot of similarities and getting things to work cross platform is much easier. MIPS brings a lot of unnecessary complication to the design process and that is why its use is declining. Really the only companies still using MIPS is the ones that have used it for years . I haven't seen a single company start up and choose MIPS. I have seen a lot of companies drop MIPS for ARM though.

 

[thilanliyan]

This would be bad if true.

 

[Fox5]

MIPS has a very convoluted design in how the processor functions. When you start working on MIPS systems you basically have to forget everything you learned about any other architecture and start over , code, interfacing , it is all different . ARM while different from x86 has a lot of similarities and getting things to work cross platform is much easier. MIPS brings a lot of unnecessary complication to the design process and that is why its use is declining. Really the only companies still using MIPS is the ones that have used it for years . I haven't seen a single company start up and choose MIPS. I have seen a lot of companies drop MIPS for ARM though.

Heh, back in school, MIPS was used as our architectural equivalent of Java. So simple, so clean, every architecture should follow it. I wasn't a big fan, but I didn't 'get' any major deficiencies compared to other isas. What's wrong with it?

BTW If this goes through, I bet on this happening:
The other ARM licensees continue their designs, take control of most of the market.
X86 (Atom) gets a big foothold.
There's some minor attempts to bring MIPS and PPC back into the field. Those old low power, low performance G3 and G4 chips would make rather attractive cell phone chips now if shrunk to modern processes.

 

[yh125d]

psh, aint gonna happen

 

[RaiderJ]

I don't really see Apple being interested in any hardware companies - keeping their options open is what drives competition between their potential hardware vendors.

 

[Modelworks]

Heh, back in school, MIPS was used as our architectural equivalent of Java. So simple, so clean, every architecture should follow it. I wasn't a big fan, but I didn't 'get' any major deficiencies compared to other isas. What's wrong with it?

If every architecture starts to follow MIPS you would have engineers jumping out of windows

Some of the problems:
All instructions are 32 bit, that means if I want to process a 32 bit constant, I have to use multiple instructions. I can't have an instruction , the constant and only use what memory they need. I have to use multiples of 32.

Instruction have to fit in the pipeline and finish in 1 clock , if I want to do a multiply or divide I have to do it using a special pipeline for that task. Cannot do math on variables in cahce because they upset the pipeline. Everything has to be pulled from cache, set up for the pipeline, processed through using multiple clock cycles and then put back into cache.

jumps are horrible if you want to access large amounts of memory often having to use multiple instructions to get a single jump completed.
There is no stack to speak of.
subroutines support is almost zero
The cpu keeps zero track of registers, cache or the pipeline, you have to do it all for it.


Mips is like an old design that instead of innovating it in hardware they continued to try to patch the problems without really changing anything. Need 64 bit support,? instead of adding hardware to support it people can just use software tricks to address it. Need to handle longer instructions ? Add the ability to chain together old ones using more clock cycles and coding tricks.

 

[Ross Ridge]

All instructions are 32 bit, that means if I want to process a 32 bit constant, I have to use multiple instructions. I can't have an instruction , the constant and only use what memory they need. I have to use multiples of 32.

All RISC instruction sets a like this, ARM included.

jumps are horrible if you want to access large amounts of memory often having to use multiple instructions to get a single jump completed.
There is no stack to speak of.
subroutines support is almost zero
The cpu keeps zero track of registers, cache or the pipeline, you have to do it all for it.

None of these problems exist in the MIPS architecture. You must be thinking of something else.

Of the top of my head, the only odditities with the MIPS instruction set that the ARM instruction set doesn't also share are branch delay slots and the weird way its multiply and divide instructions work.

 

[Modelworks]

All RISC instruction sets a like this, ARM included.

No they are not. I can mix and match as needed under other micros.

None of these problems exist in the MIPS architecture. You must be thinking of something else.

Of the top of my head, the only odditities with the MIPS instruction set that the ARM instruction set doesn't also share are branch delay slots and the weird way its multiply and divide instructions work.

They do exist in MIPS unless they changed it in the last two years. Pick up a copy of See Mips run. I see people saying things like this often but they usually use the chips through an OS or something like Java running on top where all the hardware underneath is hidden. They are not aware of all the code that is MIPS specific that is required if you are writing code in assembly.