There will be blood: why Apple and Intel are destined to clash
arstechnica.com
arstechnica.com
When it came time for Intel to introduce the successor to the Core 2 microarchitecture, they did it with the i7, a high-end desktop chip. Apple doesn't make a high-end desktop. They make small form-factor desktops and they make workstations. It shouldn't surprise anybody that the desktop i7 didn't get picked up by Apple quickly. When the Nehalem-based Xeons were released, Apple was the first to use them, in the Mac Pro. Much later, Apple was able to get the i5 and i7 into only the biggest of the iMacs, which have always (in the Intel era) been MacBook Pros at heart.
Things have been a bit different in the laptop department because of Intel's anti-competitive behavior with respect to graphics chips. First, they refuse to grant NVidia the licenses they need to make chipsets for the mobile Nehalem processors. Then, they refuse to release a mobile i7 that doesn't contain their pathetic integrated graphics that to Apple (who is trying to ensure all their customers can have GPU-accelerated video decoding and OpenCL) is just a big lump of wasted silicon. Thanks to Intel, Apple's laptops need to add an extra graphics chip to their already-crowded motherboards. That's why none of the 13" machines have gotten an i5 yet: they're too small. In order to put an i5 in a MacBook, Apple would have make a major compromise or get really clever during a complete re-design of the internal layout.
Ars tries to partially attribute Apple's continued use of the Core 2 to a lack of engineering resources due to a focus on the iOS devices, but that sounds absurd given Apple's recent financial numbers. I think it can be entirely explained by the differing and conflicting interests of Apple and Intel. They undoubtedly are watching each other's mobile strategies, but that's not necessary to explain what's happening with the desktops.
Not precisely. In the latest generation of chips this is in on-die. Apple doesn't have to add Intel's integrated graphics to their motherboard, Intel ships all their chips with their integrated graphics cores inside them, sort of (but technically slightly different than) how they ship their chips with L2 caches inside them already.
Intel may well be being dicks about licensing, but to ask for a westmere i7 without Intel's integrated graphics is asking for them to spin a different die process. (Though they did do that for Apple once IIRC.)
Overall, the fact that it's there isn't a huge issue, people who don't like it just turn it off and use an off-package graphics card. I think this is more about the licensing issues than Intel needing to offer another die.
It's a licensing issue, not a die construction issue. The fact that all the mobile i7 dies happen to contain Intel's integrated graphics is irrelevant.
Anyways, your points are all quite right and Apple has asked for and gotten smaller packages before.
I still think the licensing issues trump the package issues and that more relates to Intel absorbing the northbridge onto the die. (That one actually did go on-die not just on-package I think?) It just seems silly to blame an unrelated change that added an on-package IGP for these issues.
I mean I know people have an irrational hatred for Intel's IGPs for some reason, but blaming them for the nvidia license dispute issues that are about the northbridge going on-die? I just don't get it.
Then you haven't been paying attention to how Apple got those recent financial numbers. Apple has held operating expenses extremely low, and hasn't grown its engineering ranks significantly in almost a decade. It's not at all unreasonable to expect that engineering resources are constrained. That's how Apple likes to work.
Well - back in 2002 their R&D was about $100 million per quarter, now it's over $400 million, so I doubt your assertion.
With NEON, floating point performance is coming along, and Apple have invested heavily in getting Grand Central (multi-threading) accepted by app(lication) developers.
Cocoa apis are moving heavily towards blocks, and towards being architected with efficient exploitation of multiple cores. There are likely further changes, such as tagging enumerations with 'can-parallelize', that will further increase app performance.
I believe there is a WWDC 2010 session on how to make your code run into the future. I will watch that with interest to see if there are any more clues, but in truth, we probably already have them.
Since when has Apple outsourced R&D of anything recently? Pretty sure the iPad A4 chip is not a repackaged Intel chip. The research and design that went into the unibody macbook pros doesn't seem like they are raising the white flag just yet either.
Yep. Apple won't sit idly by. No matter what Intel does.
Is Apple concerned if somebody makes a cheap, one-size-fits-all platform to compete with the iPhone. I highly doubt it.
Isn't Apple designing its own microprocessor for iPhone now so they optimize exactly what they want optimized? Why would they care if people believe the opposite strategy is better?
Intel want x86 on the phones, since there are so many developers out there used to it, so many tools, so much software that would only need a small tweek.
I'm not sure about that. ISA is an implementation detail, a popup in your IDE.
Any existing Windows or Mac app that needs to go mobile will in the least need an interface overhaul (true for iOS and WP7) and at most need to adopt completely new mobile APIs.
Atom is a decent first step toward that niche, but I'm still not convinced about the value of x86 in mobile computing either.
Intel has money, and its strength in semiconductor architecture engineering and process technology. Eventually, x86 will be ARM competitive. The real question is, will it be in time to catch Apple?
Oh well, I guess I could always go back to Linux or BSD for my primary workstation.
Right now when you develop for iOS you're going to have the best experience on Mac with XCode. But as part of that, porting your app to e.g. Android probably means a total rewrite. If you had to use e.g. Linux for developing iOS applications then it would be easy to write in some language and have the IDE have different phone backend targets. Do you think Apple would like that?
1) I don't think anyone ever claimed Apples was discontinuing OSX, just that it seems clear that it's taking a back seat, both marketing and engineering wise, to the iPhone/iOS line of products.
2) If Xcode did run on Linux or Windows, that wouldn't reduce the lockin between Xcode and iOS. The inability to write iOS applications in a higher level language and target different phone backends has absolutely nothing to do with XCode running on OSX, and everything to do with the iPhone developer agreement specifically preventing you from doing so.
Odds are, they're headed in either the same direction with their PC line, or they're finally going to give it up altogether and start finding partners to OEM the OS. They'd have a HUGE leg up at getting into datacenters that way -- contracting companies like RackSaver, Boxx, and other similar vendors (which does, in spite of recent news, still include Dell) would help enormously with that. It's already at a point where other than the pretty case and the OS, the mac is just another PC, and I would argue that the part that matters for the non-mobile PC's isn't the hardware or the pretty case.
Apple does the industrial design, and then solicits RFP's to get the system board. The laptop OEM builds the motherboards and all that, and I'm guessing that they deliver the internals to Apple, rather than fully-assembled machines, so that they do the final integration and burn-in testing and that sort of thing.