Only insanity would lead Apple to make a mobile chip play
Forget the iPhone. PA Semi's an embedded affair
Comment So, here's the deal. If Steve Jobs can actually turn the acquisition of chip start-up PA Semi into a fruitful mobile endeavor, then he's an even more fantastic genius than the world has guessed to date. Because this deal seems to make almost no sense at first blush if you swallow the souped-up iPhone line.
The conventional wisdom cropping up as a result of the Forbes story which broke word of the acquisition is that Apple will use PA Semi's low-power processor designs in future iPhones and other mobile devices. Using that as the base logic, most reports covering this tie-up then chalked the deal up to a blow against Intel, which figured to get its Atom chips into the iPhone one day.
Today, the iPhone runs on an ARM processor design, which affords Apple certain luxuries. Numerous semiconductor companies create variations of the ARM chip, so Apple can bring these people in and pick the most suitable parts at the most suitable prices. In addition, many of these ARM designs demonstrate solid performance while consuming low amounts of power - the key ingredients for mobile chips.
But when we talk about low power in the mobile market, we're really discussing things in the milliwatt range. PA Semi's processors - based on a variant of the PowerPC design from IBM - do low power on quite a different scale. They eat up between 5 watts and 13 watts at 2.0GHz, which is marvelous for the company's target markets in the server and storage sector but less fantastic if you're trying to put a cell phone that won't fry your genitals into a trouser pocket.
Could PA Semi come up with an even lower-power design for the mobile market? You're damn right it could. The company has some of the top processor minds on the planet, including low-power and multi-core chip experts.
The rub, however, is that such a chip has never appeared on a PA Semi roadmap. So, we'd have to assume the company will take a couple of years at minimum to prepare such a part - on paper.
Even if PA Semi had a proper mobile part working in the labs, Apple would be gambling its entire iPhone business by picking the processor. First off, Apple would need to go through a software overhaul to move onto PowerPC. Secondly, it would need to bet that a chip start-up, even with Apple's cash piles behind it, could produce such a processor on time and continue producing subsequent chips on time. Over and over, we've seen chip start-ups die believing they could pull off such consistent magic, since creating and manufacturing these semiconductors proves very difficult.
Next page: Bad Grief
@ Daniel B.: '[..] then saying some time after that "Intel is the way!"' -- well, there were 8 years between the introduction of the G4 and the switch to Intel. Switching back to PPC within within 2 years after going through a huge amount of trouble to switch to Intel in the first place might look a little ludicrous. Whereas at the time it came out, the PPC 7500 (aka G4) actually beat Intel's chips lengthwise and broadside in performance per watt, throughput per cycle and actual just plain performance in practical applications like image or video editing.
@ Geoff: Good point, which I had overlooked. That might really be an interesting thing to observe -- though AFAIK the current line of XServe RAIDs is not manufactured by Apple but rebranded OEM ware.
@ Niall: Compared spec-for-spec, I find Apple not all that overpriced -- no, really. I guess it depends on where in their product life cycles you compare prices, but at least with the notebooks, they're not all that expensive if you compare same-spec machines from the likes of Dell, Acer, Toshiba, et al. (I last compared about half a year ago, then decided to install Linux on a MacBook Pro because it was about 250 quid less than what the others were offering me at the time in the same performance range).
But that aside, I agree about the "cool" factor -- plus, since they already have the well-tested PPC version of MacOS X out there anyway... *shrug* how could they resist? And the bonus is that MacOS X scales pretty well with the number of cores involved (it's BSD on a Mach kernel after all), so they actually _could_ go for the big iron if they wanted to. They already have one PPC installation in the upper ranges of the top 500 supercomputer list after all.
Disclaimer: I am not a fanboi of any hardware or software supplier. I don't drink Kool-Aid of any kind. But I can still be fascinated by developments...
whilst I think "sleepy's" comments are good, that Apple could license a variant of PPC to Intel, I think it's more likely that Apple want to own their own Arm variant processor instead of buying 3rd part versions, and who better to craft one that a low-power specialist?
now, if I was being really cynical, I could suggest that Apple were going to roll a custom processor into iPhone so that noone outside Apple would have a working development platform, and therefore make the thing relatively unhackable; as it is, anyone with Arm SDK can have a go!
These chips are used in NEC's storage line. At the moment Apple produces the XServe RAID, which is a low end storage system, what if Apple is looking to produce a Midsize / Enterprise level storage system based on OSX? If they could build something easier to run than a NetApp, then it would sell like hot cakes, practically zero training requirements, simple setup and management, ZFS filesystem to do snapshots and point in time restorations, time machine interface to do restorations. I think these could well end up in an Apple branded NAS device, if not a NAS/SAN hybrid device.
There is definately a market for storage at the moment and it's one of the few places where other manufacturers have an entry into media organisations where Apple has a traditional customer base. These devices are also high margin, which fits with apples sales philosophy.
I was thinking the exact same thing. You get your low energy boxen, and not only are they cool because they produce shit all heat, they're cool because they're Apple. That said, for them to be taken up in any vast quantities, Apple will have to disregard their "pay for the label" pricing scheme.
There were some other interesting points made in some of the other comments too. It will be interesting to see what this eventually turns out to be...
Re: There is no way back to PPC Macs
"-- that would be a complete marketing nightmare, and Apple is highly unlikely to do that on the consumer and workstation fronts."
Oh really? Like Jobs doing presentations on how the G4 0wns the Pentium processor because its a better arch, and then saying some time after that "Intel is the way!" Yeah, right.
They might actually solve the bootcamp trouble by patching in a "daughter board" with the x86 basics. This was already possible 15 years ago with OrangePC. These days, that option might be actually cost-effective now.