Apple to the Core: Why Steve Jobs is buying PA Semiconductor(pbs.org)
pbs.org
Apple to the Core: Why Steve Jobs is buying PA Semiconductor
http://www.pbs.org/cringely/pulpit/2008/pulpit_20080425_004775.html
8 comments
Perhaps it's a cheap shot, but he apparently doesn't know what ISA stands for - it's glossed as "Industry Standard Architecture". I think you're off base on Apple's ability to port to a new (or old, for them) architecture and handle app translation on the fly. They've done it twice before with very good success and OS X has the infrastructure to make it as easy for them as it's likely to be for anyone.
He also talks a lot about x86 having disadvantages as a CISC that are just wrong. He says:
"If there are instructions that will never be used, why spend the silicon real estate to hard code them? CPUs optimized for OS X would be smaller, cheaper, and use less power than any Intel or AMD alternative simply because they could be simpler overall."
The problem with this statement is that PowerPC is a RISC architecture, and it clearly was insufficient. Despite having tons of unnecessary instructions and backwards compatibility issues, x86 has huge economies of scale which is why very few people buy SPARC or PowerPC anymore. The core of an x86 is a RISC processor anyways, and the complex instructions get translated into micro-instructions for execution. The only way the x86 is going to be upset is from embedded processors or from massively parallel, but it's definitely not going to be from a processor "optimized for OS X", whatever that means.
"If there are instructions that will never be used, why spend the silicon real estate to hard code them? CPUs optimized for OS X would be smaller, cheaper, and use less power than any Intel or AMD alternative simply because they could be simpler overall."
The problem with this statement is that PowerPC is a RISC architecture, and it clearly was insufficient. Despite having tons of unnecessary instructions and backwards compatibility issues, x86 has huge economies of scale which is why very few people buy SPARC or PowerPC anymore. The core of an x86 is a RISC processor anyways, and the complex instructions get translated into micro-instructions for execution. The only way the x86 is going to be upset is from embedded processors or from massively parallel, but it's definitely not going to be from a processor "optimized for OS X", whatever that means.
"The problem with this statement is that PowerPC is a RISC architecture, and it clearly was insufficient."
PowerPC's shortcomings had nothing to do with being a RISC processor. After all, you said it yourself: x86 is really RISC at its core anyway.
The "economies of scale" argument has nothing to do with the technical merits of the chip design, it has everything to do with the fact that every PC save PPC Macs has used an x86 chip for decades.
Architecturally, x86 does have disadvantages. It's been designed over the years to maintain backwards compatibility -- meaning it's still living with mistakes that were made decades ago. There are a lot of reasons why its unlikely that Apple will start manufacturing custom chips to run OS X, but none of them are because x86 is a great chip design, or that it would be impossible to design a custom chip that would run OS X well.
PowerPC's shortcomings had nothing to do with being a RISC processor. After all, you said it yourself: x86 is really RISC at its core anyway.
The "economies of scale" argument has nothing to do with the technical merits of the chip design, it has everything to do with the fact that every PC save PPC Macs has used an x86 chip for decades.
Architecturally, x86 does have disadvantages. It's been designed over the years to maintain backwards compatibility -- meaning it's still living with mistakes that were made decades ago. There are a lot of reasons why its unlikely that Apple will start manufacturing custom chips to run OS X, but none of them are because x86 is a great chip design, or that it would be impossible to design a custom chip that would run OS X well.
The economies of scale comment was simply that back before the Pentium Pro, most people thought that RISC was going to take over since it is such a better system, but because Intel is able to sell so many chips it can afford to throw extra engineers at the problem and add things like out of order execution to the x86. Or rather there are much more important things than the ISA's technical merit when considering what is the best chip to use. Plus I don't think that Apple has a chance, or they will have anything to gain by making its own chips for desktops and laptops.
The absurdity of "running OS X well is that OS X is not that special. If he means "running everything better" than it makes sense, because these are general purpose processors, and there is just a small set of instructions that are basically the same for all architectures.
x86 is not the best ISA, but the Core 2 Duo is an excellent chip, and it's hard to make a better chip no matter what ISA you are using.
The absurdity of "running OS X well is that OS X is not that special. If he means "running everything better" than it makes sense, because these are general purpose processors, and there is just a small set of instructions that are basically the same for all architectures.
x86 is not the best ISA, but the Core 2 Duo is an excellent chip, and it's hard to make a better chip no matter what ISA you are using.
The author gives 2 reasons for the acquisition:
1. The short term reason is to force Intel to give Apple price cuts for fear that Apple will make their own chips.
- I do not think that Intel needs to worry about Apple manufacturing their own PC chips as Apple already went through that phase and AMD already provides the necessary pressure on Intel to lower their prices.
2. In the future, software and OSes will not be tied down to a specific chip so Apple will start manufacturing their own processors to increase their margins.
- I think the author is on to something regarding the future of processors but I do not think the PC market will change that drastically. Apple will probably start making their own chips for the iPhone and their new gadgets but I doubt they will do the same for the PC market. Maybe at one point we will see an Apple server.
1. The short term reason is to force Intel to give Apple price cuts for fear that Apple will make their own chips.
- I do not think that Intel needs to worry about Apple manufacturing their own PC chips as Apple already went through that phase and AMD already provides the necessary pressure on Intel to lower their prices.
2. In the future, software and OSes will not be tied down to a specific chip so Apple will start manufacturing their own processors to increase their margins.
- I think the author is on to something regarding the future of processors but I do not think the PC market will change that drastically. Apple will probably start making their own chips for the iPhone and their new gadgets but I doubt they will do the same for the PC market. Maybe at one point we will see an Apple server.
Apple will probably start making their own chips for the iPhone and their new gadgets but I doubt they will do the same for the PC market.
Agreed. I am an electrical engineer (circuit designer) and Intel is a monster when it comes to fabricating processors (their fabs are super optimized for processors alone, unlike generic fabs). I don't think Apple could do much better costwise by designing and manufacturing their own processers rather than buying Intel (even factoring in Intel's profits). This assumes that AMD continues to exist as competition.
Everyone in the circuits industry is very interested in how Apple will use PA Semi's processors.
Agreed. I am an electrical engineer (circuit designer) and Intel is a monster when it comes to fabricating processors (their fabs are super optimized for processors alone, unlike generic fabs). I don't think Apple could do much better costwise by designing and manufacturing their own processers rather than buying Intel (even factoring in Intel's profits). This assumes that AMD continues to exist as competition.
Everyone in the circuits industry is very interested in how Apple will use PA Semi's processors.
First, ISA refers to a particular kind of hardware interface. I have never heard ISA refer to the Wintel duopoly.
Second, there actually is a reason to stick with Intel when it comes to fabbing - only Intel and a few other large players can deliver 45nm process. As long as Intel has the lead on process technology their chips will almost always be faster.
Multiple cores - maybe. DSP-like or 3D acceleration in hardware - more likely.
Second, there actually is a reason to stick with Intel when it comes to fabbing - only Intel and a few other large players can deliver 45nm process. As long as Intel has the lead on process technology their chips will almost always be faster.
Multiple cores - maybe. DSP-like or 3D acceleration in hardware - more likely.
It looks like someone who might actually have some insight into this purchase doesn't understand it: http://www.cnet.com/8301-13512_1-9928659-23.html
I also find it hard to believe that Apple wants to get into the microprocessor business.
Of course, one of the biggest traditional problems with device manufacturers owning a chip supplier is that losing out on other chip customers (who don't want to buy from a competitor) means they lose economies of scale, and Apple probably has enough economies of scale in the device market internally.
Well, I suppose we'll see what the plan is with time.
Of course, one of the biggest traditional problems with device manufacturers owning a chip supplier is that losing out on other chip customers (who don't want to buy from a competitor) means they lose economies of scale, and Apple probably has enough economies of scale in the device market internally.
Well, I suppose we'll see what the plan is with time.
like a lot of things cringely writes, this strikes me as kind of dubious. on the other hand, i hope it's true.
the fact that apple machines used to have a totally different architecture than PCs felt right, in a psychological way. it drew a clear line in the sand between Us and Them. now that apple machines are pretty much the same under the hood as PCs ... it's sad.
cringely's reasoning sounds specious, but i haven't heard anybody come up with anything better to explain why apple needs a chip company.
the fact that apple machines used to have a totally different architecture than PCs felt right, in a psychological way. it drew a clear line in the sand between Us and Them. now that apple machines are pretty much the same under the hood as PCs ... it's sad.
cringely's reasoning sounds specious, but i haven't heard anybody come up with anything better to explain why apple needs a chip company.
the fact that apple machines used to have a totally different architecture than PCs felt right, in a psychological way. it drew a clear line in the sand between Us and Them. now that apple machines are pretty much the same under the hood as PCs ... it's sad.
I'm a bit of an Apple fanboy myself, but I simply don't get this attitude. I want the best processor in my machines. I admit that I have an emotional attachment to my personal computers, but it somehow does not encompass the brand of processor (or hard drive, graphics chip, etc.).
I'm a bit of an Apple fanboy myself, but I simply don't get this attitude. I want the best processor in my machines. I admit that I have an emotional attachment to my personal computers, but it somehow does not encompass the brand of processor (or hard drive, graphics chip, etc.).
I think this talk about intel is crap. They wanted PA Semiconductor because Dan Dobberpuhl and his team are incredibly brilliant at design and because he has a strong interest in techniques to lower power.
You can read an interview with Dobberpuhl where his interest in low power comes through here: http://www.acmqueue.org/modules.php?name=Content&pa=show...
He's all about low power. Making iPhones that have processors that work great but low power in an essential step to the coming revolution, and will further distance iPhone from it's competitors. Therefore, acquisition. Pressure on Intel is not what this is about. This is much more visionary than that and has to do with continuing to push what is possible with iPhone as far as it can go.
You can read an interview with Dobberpuhl where his interest in low power comes through here: http://www.acmqueue.org/modules.php?name=Content&pa=show...
He's all about low power. Making iPhones that have processors that work great but low power in an essential step to the coming revolution, and will further distance iPhone from it's competitors. Therefore, acquisition. Pressure on Intel is not what this is about. This is much more visionary than that and has to do with continuing to push what is possible with iPhone as far as it can go.
Another perspective: http://www.roughlydrafted.com/2008/04/24/why-did-apple-buy-p...
Perhaps multi-platform multi core chips will become the norm? Who knows? Probably not Cringely.