It suddenly occurs to me that || Gyan || might have been asking for video footage of Jobs previewing the technology.
That can be found here.
If they could, then I don’t think that AMD could produce microprocessors that can execute the normal Pentium 4 instruction set. I’m by no means an expert, but I think that the law on the subject is that you can only protect implementations, not interfaces. So AMD can come up with a chip that does the same thing, so long as it’s not done in exactly the same way.
Err… no, but thanks.
From that link:
I agree – endianess can be a real PITA, especially if you’re a Mac-only developer who’s never had to deal with it, and stores binary data un-abstracted rather than in a container like XML.
Also an issue: CodeWarrior is widely used, and suddenly to develop Universal Binaries everyone will need to switch to XCode. And given Metrowerks’ attitude the last couple of years, its becoming clear that they’re not all the motivated to improve Codewarrior. But that’s just my speculation, Metrowerks certainly has most of the in-house expertise to make a fat-binary-aware compiler, so it may happen, but it’s unclear if and when.
And what kind of apps are likely to have Altivec code in them? Anything doing high performance imaging, including games.
Here. You need to be a member of ADC (I believe the free membership is adequate). The kit is $999. You get XCode 2.1, and a prototype Mac+Intel machine on loan (you give it back to Apple later), plus probably other stuff. Also note the Universal Binary Programming Guidelines (PDF!) that discusses such things as endian issues and fat binaries for Intel-Macs.
OTOH, just as WINE emulates the Windows APIs without emulating the x86 processor, it’s now possible for the community to develop a tool for Linux/Windows that will run MacOS X software at full speed without having to emulate the PowerPC.
This makes me wonder. The PowerPC chip can run in either endian mode: big or little. Can the newer Pentiums do that also? If so, then low-level data formats might not be a real issue, if Apple builds Leopard to use the big-endian mode. (For user processes at least, though possibly not kernel processes.)
Also, do the Pentiums use the IEEE floating point standard?
A compiled program does not consist entirely of calls to the operating system’s API. You still need to emulate a PPC chip on an Intel (to be done in Leopard via Rosetta, apparently) if you’re going to keep the old software running. Which I certainly hope Apple’s planning to take seriously, as they did with the shift from the 680x0.
Only the G3 and G4 can do that. The G5 cannot, and this was a big reason why ti took a year to get VirtualPC running on the G5. Reportedly, at the time G5-compatible VPC was released, the fastest G5 could only achieve parity with the fasted G4 at emulating a Wintel setup, which says a lot bout how important harware support for byte-order swapping is.
No.
Interesting article today in the Register. A quote that tends to support some of my earliest suspicions about IBM dropping Apple:
Yeah, I can imagine it wasn’t worth IBM’s trouble for a mere few hundreds of million/year. Still, you’d think it would have to be a bit of a black eye for a company to lose a very visible product line like that.
I’m waiting for Apple to take a page from the software companies’ playbook and slip into the paperwork a statement that, though a buyer may think he has purchased his new MacIntel, in fact he is only renting it for an indefinite period and that any unapproved use could cause the owner (Apple) to repossess the machine.
And don’t tell me Steve “Megalomaniacal Control Freak” Jobs hasn’t thought of this already!
So wait, apple comprises of 3% of the desktop market yet only 2% of the fishkill plant? Wouldn’t that mean the fishkill plant is making 1.5 times the number of CPU’s of both Intel and AMD? Something doesn’t sound right there.
Apple doesn’t get its current PowerPC processors exclusively from IBM/Fishkill (what kind of name is “Fishkill,” anyway? :dubious: ). IIRC, they’re still getting G4s for their laptops and Mac Minis from Freescale.
And speaking of laptops: Another possibility for Apple’s PowerPC → Intel switch could be plans by Apple/Jobs to try to compete more in the laptop and portables arena. IBM’s inability to deliver a low-temperature G5 has been grating Apple for quite a while already, and recent reports of laptop sales outpacing desktop sales indicates this is a market ripe for the picking.
Bingo. That’s it right there. IBM/Freescale promised Apple they’d have a portable G5, and they haven’t come close. Add in the fact that PPC’s future appears to be the not-suitable-for-PCs Cell processor, and suddenly Apple’s switch makes perfect sense.
Right, but any new software compiled for MacIntel only needs to have the API calls emulated. When the next Final Cut/DVD Studio/iLife/(your favorite “reason to buy a Mac” software) comes out and includes an x86 binary, consumers may be able to run it on their existing Windows PCs.
In response to brianjedi:
IBM and Motorola/FreeScale have virtually nothing to do with one another, and any association remaining is a vestige of the old AIM alliance that was defunct almost as soon as it was conceived due to constant fighting between Apple and its suppliers, as well as between the chipmakers themselves. Jobs reportedly once said to a Moto exec “I’ll be so happy when we don’t need you guys anymore” (or something to that effect). Some alliance. Apple initially antagonized Motorola by killing off the Mac clones, which Moto hoped would expand the Mac platform, and hence actually make them some money for a change (sound familiar?). Apple’s subsequent non-priority with Moto led to a terrible period of stagnanation on the Mac platform, which forced Jobs, (who at that juncture opted not to switch to Intel so soon after switching from the 68k line) to go to IBM for help.
Anyway, IBM eschewed the major contribution FreeScale (at the time Motorola, which spun off its microprocessor division to birth FreeScale as a subsidiary) made to the “G” line of processors early on, being a vector/SIMD unit called the “Velocity Engine” (a.k.a. AltiVec on flavors of the G4, and VMX on the G5s, produced by Moto/Freescale and IBM, respectively). The only reason IBM tacked VMX onto the 970/G5 line of processors was to preserve backwards compatibility with the G4 (no vector unit is present on any iteration of the G3, despite considerable overlap, both temporally, and in terms of design, between the G3 and G4). Remember, the 970 is a stripped-down version of IBM’s Power4 core, with a “Velocity Engine” tacked on especially for Apple. IBM does, or at least did, use the 970 line for lower-cost servers of its own, but its raison d’être is Apple.
The 970 is a fantastic desktop processor, just as the Power4 is a wonderful processor for big servers. It’s too watt-hungry and hot for portables. Apple’s hope was IBM would make a low-power G5. It now is becoming increasingly apparent IBM simply blew them off. The processors that will find themselves in both the Xbox360 and the PS3 are low-watt, high-speed Power-based chips. Just what Apple needs. IBM followed the money, end of story. Not having Jobs to deal with probably didn’t hurt (remember, Jobs publically dissed IBM on a number of occasions for failing to deliver…whilst Intel showed equally lackluster performance increases over the same period), but I imagine the primary reason for leaving Apple to hang was there was nothing in it for IBM financially.
Now, FreeScale is imminently going to release a single core upgrade of the G4 (the 7448) that uses half the wattage, has twice the level 2 cache, and probably at least 50% more speed at the upper-end than the present iteration of the G4 (the 7447A). This is a plug-and-play drop-in upgrade, requireling literally no redesign of Apple’s hardware to accomodate. This processor will put Apple not far behind the Pentium M speedwise, and is probably even less power-hungry. Some time during the furst half of next year, around the time the first dual-core Pentium M is released, what is essentially a dual-core G4 (the 8641D) will be released by FreeScale, at even faster clock speeds, with power consumption levels at about what the fastest G4 needs today. Again, this processor will go toe-to-toe with Intel’s offerings. It may not be faster, but it won’t be much slower. Apple hasn’t had much of a speed advantage for at least ten years (and often lagged well behind Wintel), so if it didn’t hurt them before all that much, there’s no reason to think it would have hurt them badly into the near future.
Jobs’ aims only make sense if, beyond a couple revisions of IBM’s and FreeScale’s 970 and G4/e600, IBM and FreeScale essentially stop developing further for processors Apple can actually use. All indication are this is precisely the quandry Apple is faced with. The PowerPC as a desktop processor has no future; not because it couldn’t, but because it makes no financial sense for the suppliers. IBM and FreeScale clearly couldn’t give a rat’s tail about Apple at this point, and there’s really no good reason why they should.
Intel is thus the only game in town. I guess AMD lacks the stability needed. Essentially Apple is getting aboard with the producer that will make desktop and portable chips for somebody else whether Apple lives or dies, and hence will always be making something useful for somebody wishing to make desktop and portable computers. This is obviously a risky proposition for Apple, as market differentiation has always been seen as either a millstone around its neck, or its sole saving grace, depending on who you ask. Many pundits seem to feel Apple should have jumped onto Intel ages ago; others think it was a rotten idea then, and an even worse one now. What argues against the naysayers is the simple fact that Apple really has no choice. Staying with the PPC platform is now clearly a bigger risk than getting diluted in the Intel platform. Apple may be leaping from a burning building with a flaming parachute, but they at least might survive hitting the ground, whereas staying put means amost certain death.
I hear that AMD’s manufacturing capacity is even less than IBM’s. Intel, on the other hand, presumably could keep Apple swimming in processors without batting an eye.
Yeah, that’s probably it.
Capacity has nothing to do with it. Apple is a 2% player in the Desktop market, AMD has around 15 - 30% of the market share depending on who you choose to believe. AMD could quite comfortably accomodate Apple without batting an eyelid. The reason Apple is going with Intel is because of branding, everyone’s heard of Intel, not as many people have yet heard of, and regard, AMD as being in the same league, even though their processors are arguable technically superior, especially for the mac platform.
So what say y’all about Robert Cringely’s speculation that Intel is planning to purchase Apple to mount a direct challenge to Microsoft’s dominance? Malarky, or effective way?
I think Cringley had his tongue firmly in cheek in the last couple of paragraphs of that article. His “five questions” are interesting, though, and I’d been privatelt mulling several of them since Apple’s announcement. There are several things about Apple’s move that don’t make sense, sans information we haven’t yet been told.
Cringly is either trolling or grossly misinformed about pretty much everything. Some of the more obvious mistakes that were glaringly obvious from that article are:
-
Apple’s benchmarks of the G5 are remarkably flawed and rely on a very limited subset of applications specifically targeted for the G5. Of course they would conveniently forget about this trumpeting when they announced their switch to the Pentium. And with job’s personal reality distortion field, so would the rest of the mac community.
-
Intel HAS a 64 bit desktop technology. It’s calle EMT64. Theres no possible way Cringly could not know this unless he’s completely ignored the desktop processor market for the last 6 months. In fact, intel has just recently announced that even the new celerons will now be 64 bit, making a 64bit celeron imac a real possibility.
-
Jobs specifically mentioned the 3.6Ghz system was a pentium, it’s not “unnamed”. In fact, you can see a freaking screenshot of the damn presentation here.
-
As has been speculated in this thread so far, there are a number of reasons why apple decided to go intel. One reason is the prestige value as I speculated. Another is to gain access to the Pentium-M which is a much better mobile platform than anything AMD has produced so far.
-
The reason why the development system was so slow in benchmarks was because it was emulating mac software. The entire point of the figures was to show that the emulation would run with a significant but not unusable speed hit. It wasn’t to try and prove that pentiums would run faster than G5’s with programs compiled for G5’s.
-
The reason it was announced a year ahead of time is because it’s freaking hard to port applications to a new platform. They went through all that with the whole motorola->IBM thing so theyre committed to handling it much better this time around. Maybe 90% of the applications will be ported in 1 month but without the other 10%, people are going to be afraid to move. Theres no possibility of “keeping it secret”, thats just stupid. Even if the hundreds of thousands of developers could all simultaneously keep a conspiracy that would rival the moon landing in complexity, theres no way you could contact every maker of a 2 bit shareware program to keep them in the loop. A lot of people are developing mac applications that never see the light of day outside of their office or personal computer. All of these people have to also port their programs.
The conclusions he draws are “interesting” but given the ultimate premise is so laughably wrong, theres no real way to say anything about the validity. In short, Cringly either has an assload of intel stock and would like to see it go up so he can make a quick buck or he’s so hopelessly misinformed even the slashdot crowd could see through his act.