Apple's CPU lead Jim Keller heads back to AMD
venturebeat.com
venturebeat.com
1. Keller's company, PA, was bought by Apple mainly for its engineering talent. Getting Keller was a huge deal for Apple.
2. Keller has been hired away by Mark Papermaster, who was a disaster at Apple (all of Bob Mansfield's reservations about him from the internal email discussed in IBM v. Papermaster came true: cultural, pace of development, etc.).
I would anticipate an outside hire from Apple to fill his shoes in the next 6-9 months.
Jobs tells the VP that if the garbage in his office is not being emptied regularly for some reason, he would ask the janitor what the problem is. The janitor could reasonably respond by saying, "Well, the lock on the door was changed, and I couldn't get a key."
It's an irritation for Jobs, but it's an understandable excuse for why the janitor couldn't do his job. As a janitor, he's allowed to have excuses.
"When you're the janitor, reasons matter," Jobs tells newly minted VPs, according to Lashinsky.
"Somewhere between the janitor and the CEO, reasons stop mattering," says Jobs, adding, that Rubicon is "crossed when you become a VP."
http://www.businessinsider.com/steve-jobs-on-the-difference-...
"When the sales guys run the company, the product guys don't matter so much, and a lot of them just turn off. It happened at Apple when [John] Sculley came in, which was my fault, and it happened when [Steve] Ballmer took over at Microsoft."
1. iPod Touch + Camera. Apple starting doing big production runs on this and it fucked up, big time. There was a problem with the cameras and the devices were unusable. There was an entire generation of iPod Touch which, as a result, shipped with the necessary connectors for a camera, but without the physical camera or the casing for a camera. As far as I know, this and the antenna were the main reasons.
2. White iPhone was hugely embarrassing.
3. iPhone antenna. Jobs handled it really well: common problem, nobody's perfect. That's not Apple though. Papermaster's team squarely blamed him for this.
4. Bob Mansfield and others identified problems with the cultural fit before he joined, which seemed to be true. He wasn't lazy or dumb, he just wasn't proactive and smart in the way Apple VPs are.
Is ironic that AMD hired Keller since Intel too found it's way out of the Netburst fiasco by going back to Tualatin, which at the same time had more in common with the good old Pentium Pro than it did with coppermine and katmai.
The Core microarchitecture was designed by Israel's Intel Israel (IDC) team that previously designed the Pentium M mobile processor[1]
The Pentium III Tualatin (mentioned above) and the Pentium M are closely related.
[1] http://en.wikipedia.org/wiki/Intel_Core_(microarchitecture)
[1] http://www.forbes.com/sites/roberthof/2012/07/30/viewers-giv...
Do you really think that some silly advert made by the marketing department at Apple has any relevance to whatever the reasons were for the head of CPU design for iOS devices had for leaving?
Things might be changing for the worse inside Apple, but taking these two datapoints and assuming that something big is going wrong in the company seems to be a conclusion that isn't supported by the facts.
Falling short with national TV ads shown during the opening ceremony of the Olympics is a big deal for a company that until recently showed mostly amazing ads on TV. Losing a person who led CPU design for the company's fastest-growing businesses is also a big deal for a company that considered having that person a major coup.
Could you imagine either of these things -- lousy ads on a major TV event, key person departing -- happening while Jobs was running the show?
--
Edit: changed "only amazing" to "mostly amazing," which better reflects my views.
Avie Tevanian, Jon Rubinstein, Tony Fadell and Bertrand Serlet, all key people at various points in the history of Apple, left while Jobs was still in charge. I wouldn't read too much into this.
I'm not sure this is a major impact for Apple. Firstly, I haven't seen anything groundbreaking come from Apple on the chip side. Secondly, there are plenty of talented people from AMD, Intel,HP, etc who would love to work for Apple. He can be replaced.
Compare the message of PC vs. Mac to this message: "We have tools. Tools that work. Oh, and you need a genius to use them right" ?
Not if it is XCode.
That is not a tool that "works".
To me the only Mac computer ad that has stood out in the last decade or so has been the MBA out of the envelope commercial.
When you have that kind of reservation looming about a company, you need to see encouraging signs, not the recent flow of tepid-to-bad news we've been seeing lately.
Between the ads, the loss of the CPU designer, the less-than-inspired iPhone 5 product image leaks (assuming they are real) and insignificant incremental product updates, one shouldn't be blamed for not ignoring the writing on the wall.
Encouraging signs:
* In their recent "disappointing" quarter they grew faster than Google.
* Mountain Lion has just become the fastest selling version of Mac OSX ever, shipping over 3 million copies in the first 4 days (from memory, those numbers may be a little out).
* The latest figures show iOS market share increasing despite the fact that the iPhone is due a refresh (which has historically meant sales are at their slowest).
* In a market where PC sales as a whole are decreasing, sales of the Mac are increasing.
But hell, they did some bad ads, the writing is on the wall.
Anecdotes, schmanecdotes and all that, but if one believed online enthusiasts every time Apple did something, the iPod was a joke, iPad was a flop, and iPhone 4S was an underwhelming epic fail. The actual opinion of world does not often reflect the Internet's bizarre jumble of fragmentary opinions.
- Apple's strength is not in its chips. It's in its software and tightly integrated stack
- It seems, from the outside, AMD needs an inspiring leader to give its design teams a clearer goal.
- I know first hand what it feels like when your work no longer lines up with where your company is going. I wish him the best of luck on his new challenges.
Huh? Every time an iOS device is released, it usually is the industry leader for mobile processors. Apple is huge and can be great at many things at once.
SoC design at that level really that innovative. You take some IP, throw it onto a floor plan and contract with a fab. Apple having their own designs is more about optimizing their production chain than it is about driving innovation.
[citation needed]
I know someone who works on processor stuff at Apple. More detail would obviously be bad for him, but they have digital design expertise which they are actively using to design (presumably) their future generations of chips.
Apple's A5X SoC is 50% bigger in surface than Tegra 3. Considering the Tegra 3 SoC contains 4 CPU cores, and Apple's A5X only 2 CPU cores, plus taking into account all that extra space, I could see how A5X's GPU easily doubles Tegra 3's GPU in size, if not more. That's where the performance increase is coming from. Hopefully Nvidia and others will be smart enough to build bigger chips with in the next generations.
Emphasis below mine:
http://en.wikipedia.org/wiki/ARM_architecture
Fabless licensees, who wish to integrate an ARM core into their own chip design, are usually only interested in acquiring a ready-to-manufacture verified IP core. For these customers, ARM delivers a gate netlist description of the chosen ARM core, along with an abstracted simulation model and test programs to aid design integration and verification. More ambitious customers, including integrated device manufacturers (IDM) and foundry operators, choose to acquire the processor IP in synthesizable RTL (Verilog) form. With the synthesizable RTL, the customer has the ability to perform architectural level optimisations and extensions. This allows the designer to achieve exotic design goals not otherwise possible with an unmodified netlist (high clock speed, very low power consumption, instruction set extensions, etc.).
http://en.wikipedia.org/wiki/Apple_A5
"The A5 contains a rendition of a chip based upon the dual-core ARM Cortex-A9 MPCore CPU[10] with NEON SIMD accelerator and a dual core PowerVR SGX543MP2 GPU.[11] ... Among the customizations that Apple has done to the chip is the inclusion of an image signal processor unit (ISP) that will do advanced image post-processing such as face detection, white balance and automatic image stabilization[14] and an "earSmart" unit from Audience for noise canceling"
Look I'm no expert but it's been often reported in the press about the A4 and moreso the A5 that Apple made optimizations to them. PA Semi was originally often attached to such reports but I think later reporting often cited the Intrinsity acquisition.
Here's a comparison of the A5 and the reference ARM Cortex A9 design:
Looking at the CPU core itself can be a little complicated. IIRC, Apple was using the Intrinsity version of the Cortex-A8 which can be viewed either as customized (since it wasn't the vanilla Cortex-A8) or as off-the-shelf (since Apple/Samsung apparently didn't make any additional modifications to the Intrinsity version). http://www.anandtech.com/show/3665/apples-intrinsity-acquisi...
Now that Intrinsity is part of Apple, if we assume that they're still applying Intrinity's dynamic circuit technology to the Cortex-A9 then you could say that Apple has a customized version of the Cortex-A9.
Also, Apple has an ARM architectural license (from the Newton days AFAIK) that allows them to add new instructions and such, but they have apparently never used it.
First link says that A9 is sold as synthesizable IP (duh), not that Apple made modifications. Second that Apple integrated a separate DSP core (a completely unrelated subject). Third just shows that they did indeed synthesize it (any change of floor plan is going to produce an completely different layout for the same circuit).
Is there any evidence at all that Apple made secret-sauce changes to the A9? There certainly aren't benchmarks of any kind to show it. Any new features? Anything at all that the posited "Apple A9" does differently than all the dozen others out there in the wild? No.
Recap of this subthread:
Someone says they know someone at Apple and they are working on "designing future chips"
You: Apple doesn't design anything, it's all off the shelf.
Me: I'm pretty sure he means Apple's modifications to the base stuff.
You: No, there's nothing in wikipedia that indicates Apple modified anything.
Me: here's two wikipedia links and other citations showing the changes Apple makes and how that might be expected for some ARM licensees.
You: They maybe didn't change one part I want to talk about, I win.
For the record I think you are correct that it hasn't been shown that Apple has made radical changes to the A9 CPU. It also hasn't been shown to my knowledge that they haven't made changes. There certainly are benchmarks where Apple's chip does better then other A9's but there's no evidence this is hardware based.
Most importantly, I don't care if they changes they're making are to other parts of the SoC or to the core A9 CPU design. Changes to the SoC perfectly qualify under what you originally replied to "future generations of chips" before the gratuitous display of nitpicking and goalpost moving.
See: http://en.wikipedia.org/wiki/Apple_A4 or http://en.wikipedia.org/wiki/Apple_A5
They did after buying PA Semi, but after Agnilux and Keller they may not now. In any case it appears they didn't use that talent that way.
Frankly if you really want to argue (and for the record I think such an argument is ridiculous), ARM and NVIDIA have much more "elegantly scalable" solutions that don't require high level brute force duplication of the core.
I've always had the opinion that Apple's hardware and industrial design were top notch, but I hated their OSes and software. Then again, I was never the target market for it.
The iPhone 5, might be the last phone with custom Apple chips.
It's more likely that Apple has and is satisfied with a 3-4 year roadmap for the A* line that looked pretty boring to Keller, and the AMD job appeared much more interesting (and potentially much more lucrative).
All phones are basically moving toward the exact same shape and size.
In that time frame, they could have already selected a future Intel SoC.
Why would they need to continue upgrading the hardware at this point? It seems like a useless effort given that they have a software technology advantage and marketing brand advantage.
I'm not sure what you're asking. Apple has demonstrated time and again that their strategy depends on offering the best possible technology at desirable price points. Consumers don't care about MHz the way they used to, but they know the difference between a fast and slow UI, and they care about the features and battery life that are dependent on an ever more powerful and efficient SoC.
Intel, as much as they might want to please Apple and know what a coup it would be to be inside the iPhone and iPad, is only one company, and one with a history of engineering pride, sometimes to their detriment (see Itanium[1]). Apple would have to be sure that Intel would respond to whatever crazy demands their engineers come up with, and right now a bunch of scrappy ARM vendors seem much more likely to do that. If one fails Apple, they can hire another one for the next device. Worst case scenario: they can buy one. If Intel stops playing ball, they'd be stuck with another painful ISA transition. They've gotten better at those but they're not free.
Apple's seen what happens when a supplier roadmap (Motorola/IBM's PPC) goes out of alignment with theirs. Even if Intel produced a product at parity with the best ARMs, a competitive ARM market (albeit one currently dominated by Samsung) serves Apple better than a single source.
EDIT: [1] I should mention that Itanium was actually fairly effective for Intel at clearing the field of several major "enterprise" architectural competitors, disaster though it was as a real product. But surely they would have been happier if it had actually been a success instead of a multibillion dollar embarrassment.
Deleted comment
I would keep an eye on the Apple share price over the next few weeks, I have an inkling that something is going to change regardless of how much money they have in the bank.
Given that it may be understandable for Keller to move onto something more of a challange for a person who likes to design the bricks.
Apple don't need to make there own add on's, they pick what they want and what they don't want/need and in that the level of skills needed to assemble those parts onto a workable CPU is at a level were you don't leverage the skill set that Keller has as much as you could.
I can understand Apples needs and what they don't need and in that it is non impacting to Apple and probably a good move for Keller.
To read anything else into this story is clutching at the speculation stick and has no grounding.
Is it simply a case of 'maybe he never signed a NCC?'
This is a feature, not a bug in California's laws. I wish my state would adopt a similar view.