We’re continuing our detailed coverage of all the bits and pieces dripping from the Worldwide Developers Conference in San Fransisco. As you all might recall, Apple bought a chip company called P.A. Semi not too long ago, a maker of low-power PowerPC chips, fuelling rumours Apple might use them in its own products such as the iPhone and iPod. This has now been confirmed by His Steveness himself.The New York Times caught up with Jobs after his keynote presentation at WWDC, and the most interesting bit is a few words Jobs shared with the New York Times on Apple’s acquisition of P.A. Semi. Jobs said:
PA Semi is going to do system-on-chips for iPhones and iPods.
This has been pretty much a dead give-away for most of us, but it’s nice to finally hear a confirmation from the best source you could ever wish for. I think we can now also safely put to rest any of the crazy ideas some people had about Apple using P.A. Semi’s chips in laptops or desktop computers. I’m with you on it being exciting, but it won’t happen.
This may also shed some light on the recent news that Apple might – possibly maybe – drop support for PowerPC chips when Snow Leopard comes out. P.A. Semi currently makes PowerPC chips, and if those were to end up in iPhones, Mac OS X cannot drop support for PowerPC, since Apple points out – ad nausea – that Mac OS X-proper is the same as Mac OS X-iPhone. PowerPC users, do not rejoice just yet: who is to say P.A. Semi isn’t simply going to make ARM or low-power x86 chips?
What happened to Samsung? I remember reading a while back that Samsung chips would wind up powering iPods, did that deal ever happen?
I hope they can come up with a better headphone amp. The amplifier in my iPod Touch is VERY noisy with high-end earphones, although I have an external portable amp so it’s not a big deal for me.
I believe the SoC that currently powers the iPod is in fact a Samsung part.
Also note that P.A. Semi have no fab of their own, so unless Apple plans on buying up a chip foundry also, they’ll still need a manufacturing partner, which would likely remain Samsung.
Edited 2008-06-12 01:05 UTC
In the keynote his Stevenes said that he will reduce the cost of iphones.
Can this possibly reduce them further?
Edited 2008-06-12 03:10 UTC
I believe he was talking about reducing ‘his’ cost, not ours.
Anyone know what the potential response from the other 90% of the software market is? Does this announcement even warrant a response or is it just some basic tools with a lot of hype?
For instance, how does it compare to Intel’s Thread Building Blocks library (www.threadingbuildingblocks.org), which was open sourced (GPLv2) last year? It is described as “C++ template library that simplifies the development of software applications running in parallel.”
I haven’t found any information on the OS community’s stance on this OpenCL language. Anyone know anything here? Would any software on a default install of a Linux distro (or OS X, for that matter) have a use for this?
Nobody has much idea what OpenCL is so they can’t have any useful opinion.
BTW, I think you replied to the wrong thread.
Apple has not released any information on OpenCL, which I find rather amusing for something touted as “open”
Intel’s threading building blocks would not be suitable for general OS X development. While it is a nice library, it is too reliance on templates and it borrows heavily from the STL. That’s a really good thing, if you’re targeting C++ developers. Since the majority of development under OS X is in Objective-C*, that’s not very useful.
Because of that, OpenCL has to be C based. In fact, I’m going to guess it’s just a C library and if they manage to do in C what Intel have done in C++, that’s still going to be very useful. Not everyone uses C++ and language bindings (for Python, Perl, etc) are a lot easier to write for C libraries.
*You can use C++ code in Objective-C++ but it’s not that common as it’s very complex.
As far as I understand OpenCL is oriented at GPU-based acceleration. In the area of parallel processing a lot of interesting work is going on in the development community. For instance, OpenMP (supported by the latest gcc versions and Visual Studio) is being adopted steadily. I did not have time to try it extensively yet, but from a quick glance it seems to make parallelization over constructs like loops fairly easy. Other interesting work is carried out in parallelization of the C++ standard library, e.g. in GNU libstdc++.
Wrt. to OpenCL: I read some rumors that Apple wants OpenCL to become a standard, so with a bit of luck they will make the source available.
I doubt that Apple will use the Power arch again! My guess is that they will put the team to work on a Arm clone that’s super low power and very fast.
Krreagan
Time to market says otherwise.
In what way? Apple has already had to come out and say they will continue the PPC support at P.A. for the Military, not their commercial products. I think SJ felt burned by Moto and IBM for their (lack of) support for low power and fast PPC processors. After all that has been the main reason given as to why Apple left the PPC arch in the first place.
They won’t develop for a new PPC device in the foreseeable future. they may or may not continue support for their legacy products in the White Cat edition!, that is up to them.
Krreagan
You are probably right that it will be ARM, but I doubt P.A. Semi will make their own core. More likely, they will design a SOC around an existing ARM core. They _might_ add some simple processors around the ARM core to make a Cell-like architecture, but I really can’t see why they would need to — the newer ARM cores should have plenty of compute power for iPhone-like devices.
What could be more interesting is if Apple wants to go into the micronotebook market (to compete with Asus Eee and the other newcomers). Putting MacOS on one such would be quite a selling point. If they do so, an ARM-based SOC at heart might not be a bad idea.
I’m willing to bet that Apple will still maintain a PPC port of the desktop version of OS X seeing as they now maintain OS X for 3 architectures, not counting different bit counts (or whatever the technically inclined call it
). Keeping the OS ported to multiple architectures should yield a leaner, higher-quality codebase, or so the theory goes.
i cant shake the idea that apple will go iphone on their non-pro desktop and laptop products.
if done right, people can get the stuff they “need” from either apple or some third party via the app store, itms and so on.
the pro products will still be around, with a matching price tag, for the professionals that still want to use photoshop and other media products that have stayed available on apple over the years.
This is a strange move. They moved away from their (almost) custom made chips for notebooks, because their price was too high for the mass market and couldn’t stand the competition from Intel.
Now, they do the opposite: they move from a common design to make their own, for a mass market. Will they stand the competition this time?
Except that Apple didn’t own IBM and Motorola…
I think the story is different with SoC chips.
For these types of devices what they’ve done with P.A. is really just an extension of what they do with their custom hardware in general. They take a generally available CPU (nVidia chose ARM chips for it’s SoC, Apple will use P.A. chips), and integrate it in silicon with their own parts, or licensed third party parts. That’s remarkably similar to their current model – take the chip, Intel, ARM or whatever, and put it on a circuit board with other licensed chips (and maybe some custom/proprietary ones, like nVidia does).
The only difference is the package they are putting it in. They’ve shifted to micro chip level integration, in addition to circuit board integration.
Having PowerPC chips in iPods and iPhones won’t stop them from dropping PowerPC support on the desktop.
I must say that I’m quite sad with the demise of PA Semi, and I don’t understand why Apple did that.
Their PPC chip was brilliant and had very fast acceptance in high performance/embedded applications (for example mil, aero…). There are many applications where no consumer-grade Intel x86 is suitable.
ARM licences CPU cores, Apple can design custom ASICs without having to actually develop an ARM CPU, like nVidia and Samsung embedded chips, for example. In all cases, you’ll have to get a licence from ARM to create an ARM clone.
The PPC has shown numerous times that it can reach higher performances than ARM. PA Semi has shown that a high performance PPC core can be a relatively low power device. But, if Apple would need high performance/low power CPUs, they could as well use Intel Atom chips.
I can’t belive that the high profile guys at PA Semi will spend their time doing ASIC assembly with IP bricks. If Apple does not have something really innovative, they will leave.
Hello,
The PA Semi core team worked on StrongARM before they worked on PPC
. These guys know their stuff about ARM, probably moreso than Samsung, as seen here:
http://blogs.computerworld.com/apple_buys_p_a_semi
If Apple wants this technology fabbed, they can go to TSMC or one of the other fabs out there who will gladly fab the chips for them (even IBM).
The other plus (for Apple) is that Apple gets to play keepaway with technology that would actually help IBM make fast PPC chips that won’t need water cooling.