Apple Switching To Intel Chips In 2006 1427
telstar writes "According to C|Net, Apple has officially decided to drop IBM, and will use Intel processors starting in their '06 line of systems. This change was rumored last month. The announcement is expected Monday at Apple's Worldwide Developer Conference in San Francisco, at which Chief Executive Steve Jobs is giving the keynote speech." From the article: "Apple successfully navigated a switch in the 1990s from Motorola's 680x0 line of processors to the Power line jointly made by Motorola and IBM. That switch also required software to be revamped to take advantage of the new processors' performance, but emulation software permitted older programs to run on the new machines."
Any Evidence At All? (Score:5, Insightful)
Is this yet another rumour? Is there anything to be read in Apple meeting with Intel above the idea that they might go PCIe instead of PCI-X?
Comment removed (Score:3, Insightful)
Comment removed (Score:2, Insightful)
Re:This obviously means no Powerbook G5s (Score:4, Insightful)
Looking at the iMac G5, I can't see why not. I mean, that things almost a laptop already! I'm really surprised there are no laptops with G5's yet. I thought it was the next step from that iMac...
Re:New device (Score:3, Insightful)
I think you hit the nail on the head. Recompiling regular high-level (C, Obj-C, etc) for x86 or x86-64 is (relatively) freaking easy. Rewriting hand-tuned Altivec code to run on Intels SSE-2 or SSE-3 is a major, major issue for software developers both inside and outside of Apple... and supposedly, Altivec is superior to SSE1/2/3 in quite a few regards. It won't be a direct translation.
I like your "Intel producing PPC chips idea". If licensing agreements allow for it, that could be technically feasible. But... I highly doubt that Intel wants to produce chips based on an entirely new architecture. I think we can utterly rule this out.
What seems a lot more likely to me is Intel producing x86 chips with Altivec-equivalent instructions to ease the porting process. If Intel could add Altivec functionality to an x86 chip, compatibility could basically be a recompile away. (perhaps with some relatively minor adjustments for endian issues)
That doesn't seem like it would be a technical hurdle for Intel, if licensing agreements allow for it... I know the name "Altivec" is trademarked; I don't know if the actual instruction set is covered by IP law. (The name "Altivec" obviously doesn't matter because it ain't exactly a household word anyway; only the functionality matters)
Intel knows how to make chips, not just x86 (Score:5, Insightful)
Would anybody be that surprised if Intel started making PPC-esque architecture chips? Don't be. Intel knows Si's at 14 as well as anyone and better than most.
Too many people have taken these rumors to mean Apple's going to release Macintosh for x86. I'm not quite ready to jump that gun just yet.
This is lunacy. (Score:2, Insightful)
Intel chips take ridiculous amounts of power, and give off huge amounts of heat. The current PowerPC chips use a fraction of the power. This makes no sense for the Mini or the iBooks.
iBook battery life will drop by at least a third, maybe one half. Unless of course they could use slower chips, but there's no way that's going to fly. We all know the canonical Mac argument - "yeah the clock speeds are lower but the chip is totally different so it's not a one-to-one correspondence. Trust me, we're faster". Except this time, there _will_ be a one-to-one correspondence!
In market where clock speed is the main thing marketers latch onto, putting a genuinely slower chip in an iBook is suicide.
In a product line which relies on a marketing thrust of reliability, putting a frigging heating element in a notebook is likewise suicide.
Executive summary: either this is a hoax or Steve Jobs has finally gone off the deep end.
Comment removed (Score:4, Insightful)
Re:Intel knows how to make chips, not just x86 (Score:5, Insightful)
Re:68k emulation easy, but what about PPC emulatio (Score:1, Insightful)
Fast forward to 2005. Apple's market share is in danger of slipping into third place, behind Linux, and nowadays their hardware really is vastly overpriced compared to the competition. My wife's iMac was $2000, versus $300-600 for my generic PC that's just as fast and does all the same stuff.
There are now a lot of users now who paid $2000 for a Mac that is getting pretty long in the tooth. Apple convinced them to pay $130 for MacOS X, and then $130 every few months after that for the latest 10.x version. They also paid lots of money to upgrade a bunch of their software to run MacOS X native, instead of in MacOS 9 compatibility mode. After laying out all this cash, does Apple seriously expect them to buy an intel-based machine for their next Mac, and be forced either to pay for all their apps again or run them in slow-as-hell emulation mode?
Sorry, but it doesn't make sense. I don't think Apple is that stupid, and it makes me not believe the article. If they did this, I think their market share would drop so low you'd need to state it as a decimal.
Scary as HELL (Score:1, Insightful)
http://www.digitmag.co.uk/news/index.cfm?NewsID=4
Some high lights:
1) However, Tucker ducked questions regarding technical details of how embedded DRM would work saying it was not in the interests of his company to spell out how the technology in the interests of security.
2) Conversely, Intel is heavily promoting what it calls "active management technology" (AMT) in the new chips as a major plus for system administrators and enterprise IT. Understood to be a sub-operating system residing in the chip's firmware, AMT will allow administrators to both monitor or control individual machines independent of an operating system.
3) Additionally, AMT also features what Intel calls "IDE redirection" which will allow administrators to remotely enable, disable or format or configure individual drives and reload operating systems and software from remote locations, again independent of operating systems. Both AMT and IDE control are enabled by a new network interface controller.
Hey, OS X might run on your standard Intel / AMD box and they might compete with MS. I would rather have Apple stay on their own hardware and keep the DRM (hardware based) out of the picture.
Who really cares? (Score:3, Insightful)
I submit that normal USERS (not some geek with an odd political fetish) really don't care what the hardware is. I am sure the OS will still be "Mac OS X".
Sheesh, do I care if my snail-mail letters are carried via pigeon, car, truck, plane or train, as long as the bill is marked "paid" on time!
Re:April Fools? Right? (Score:3, Insightful)
Right now the Pentium M is Intel's most expensive CPU, and there's really no alternative to it if Apple wants an Intel CPU in the Mac mini
Re:This obviously means no Powerbook G5s (Score:3, Insightful)
Let's see... maybe because the sonofabitch weights 25 pounds? http://www.apple.com/imac/specs.html [apple.com] has the 20" at 25.2 pounds. I got one iMac G5 20" at home and five at the office and "almost a laptop" doesn't cut it.
Re:April Fools? Right? (Score:1, Insightful)
So after a long hard day at work, you like to come home and kick back with a game or two. Say a game on your new xbox 360... powered by IBM. Or you're waiting for your playstation 3, powered by IBM, to arrive next year. Or you prefer something a little more Revolutionary... something perhaps powered by IBM? Sure, IBM won't be getting $100 per unit, but they'll certainly have volume. It's little wonder that IBM found themselves with more interesting things to do than ship apple their G5 chips.
Re:April Fools? Right? (Score:3, Insightful)
Apple hasn't been able to produce enough high-end PowerMacs to keep up with demand in YEARS, due mostly to availability of the CPUs.
Re:The sky is falling! (Score:5, Insightful)
So unless this direct afront to Apple was mitigated with huge discounts, I doubt Intel will get the deal.
Of course none of this will be public except the choice of chip supplier. We'll have to read the spin from Apple, Intel, and makers of the Pandora like we were reading pig entrails for signs of what's really going on.
Comment removed (Score:3, Insightful)
Re:The Desperate Need For Validation In The x86 Wo (Score:5, Insightful)
I didn't comment on the benchmarks story earlier as I was way too busy but it is now 4:40am and I have nothing better to do...
A lot has been said about the speed of various systems and benchmarks but at the end of the day the only real test for the performance of a computer is how well it runs your work. I have been developing some pretty compute intensive software for the last three months and I have to say that the G5 is a very quick processor. In my benchmarks a 2.3Ghz PPC was able to handily beat a 3.06Ghz Xeon EM64T chip with my code. This is only with gcc at the moment, I expect using the IBM compiler will make a significant difference just as using the DEC compiler on Alpha produced far faster code than gcc could. Opteron is also a very fast chip. So is Centrino. P4 and the Xeon based on it are also fast but clock for clock they are seriously underpowered.
Do not think that because some benchmarks showed what you want (that some cheap tatty Intel box is faster than a high end PPC970) that it is in fact the case. Write your own code and give it a whirl. Heck, my G4 Mac mini (1.42Ghz) is quicker than my Athlon XP 2200+ (1.8Ghz) running my code and it isn't even using Altivec yet.
As for Apple using Intel chips, far more likely it is something derived from the iPod part of their business than the Mac. The PPC970 is not underpowered, it is very quick, very efficient and easily a match for anything Intel has. AMD on the other hand has a very nice CPU in the Athlon64/Opteron and I would be torn to choose between the G5 or Opterons in a cluster as it would come down to performance running our apps as well as price.
x86 Mac does not mean IBM PC Compatible (Score:3, Insightful)
YES, MOD PARENT UP (Score:3, Insightful)
The idea that Apple would switch to x86 simply doesn't make sense. There are no drivers, and no applications. Of course Apple would continue using their own hardware and would port their own applications, so such a machine wouldn't be a complete paperweight, but seriously, without backwards-compatibility (via PearPC etc.), why would someone want one?
Re:MacOSX on x86? (Score:4, Insightful)
Re:April Fools? Right? (Score:3, Insightful)
Yeah, they've got a lot of money to spend, but they just keep churning out crap. AMD has never had more than a fraction of Intels cash to spend, but they absoutely kick the shit out of them where effective R&D is concerned, and have for some years now.
Intel has 2 things going for them. They churn out the volume, and they have good marketing. Which is more than good enough to keep them in business, despite the fact that they come up with lemons every time they try something new.
No Joke? (was Re:April Fools? Right?) (Score:3, Insightful)
Certainly BSD is highly portable and parts of the MacOS includes bits of NetBSD which is especially portable...
But if you get over the "mine is bigger and faster than yours" basically the shift the marketing message from "Speed vs Speed" to one of technical merit about UI and so forth... because once they switch (if they switch), WinTel and Mac "computers" will have the same speed, in the same time frame.. so the Gigahertz #'s about speed/performance disapear.
It's also clear w/ the IPOD that see the value of a greater marketshare and customer acceptance. Price is a big part of that and like they have done with IDE, PCI, USB, etc. switching to Intel probably means they can lower their price without cutting profit(s), that means greater marketshare, etc.
Why? (Score:2, Insightful)
- Intel and AMD are shipping dual-core CPUs *today*. IBM's Antares is still in the design phase.
- PPC970 has not scaled as IBM anticipated. It is approximately 10-20% slower, clock for clock, than AMD64. AMD64 has scaled well with process improvements, and it easily reaches 2.6GHz with moderate aircooling. PPC970 is only clocked at 2.7GHz, and it needs watercooling to do so.
- PCI Express chipsets are available for both Intel and AMD CPUs. No such chipsets exist for PPC970. PCI-X and AGP cards are being quickly replaced by cheaper and faster PCIe solutions. The latest GPUs are often not available in AGP variants at launch, and the lag time is increasing. AGP GPUs often now cost more than PCIe GPUs.
- IBM's semiconductor fabrication business isn't doing well. Ripe from failures such as NV30 (fabrication at IBM was such a disaster that NV35 was fabricated by TSMC), IBM is realizing that being able to deliver world-class server chips in small quantities doesn't necessarily translate into being able to deliver low-end chips in high quantities. Numberous supply problems delayed the launch of the XServe G5 and new Apple desktops. Intel and AMD, in contrast, have extensive experience shipping in large volume.
- The PPC970 is expensive. Apple's new lower-end products need inexpensive CPUs to remain cost-competitive. Right now, that niche is filled by Freescale's G4. The G4 is a dog, though, and Apple needs to replace it. Inte's Pentium-M CPUs are considerably faster, similar in power draw, and are constantly falling in price. Intel's Celeron-M is even more affordable, and it still offers superior performance to the G4.
Doesn't make sense for Apple (Score:3, Insightful)
Re:New device (Score:3, Insightful)
Xserve RAID is a storage appliance. It could use a massive array of Zilog Z80's for all its relevance to the discussion at hand.
Re:MacOSX on x86? (Score:3, Insightful)
One more reason not to buy an Apple.
Re:Any Evidence At All? (Score:2, Insightful)
Re:68k emulation easy, but what about PPC emulatio (Score:3, Insightful)
Further points:
1. How do they expect to able to announce this and not kill their laptop and desktop sales for the next 15 months or so?
2. Steve Jobs is notorious for hating leaked announcements. "inside source"? This could be one colossal troll.
3. None of the typical Apple rumors sites seem to have heard anything more on this.
4. Just because IBM has had problems with designing/manufacturing faster PowerPC chips doesn't mean the architecture is at a dead end; by all accounts PowerPC offers many advantages over x86 on a very basic level.
4b. The x86 instruction set would be a pretty significant step BACKWARDS, architecturally. x86-64 fixes this, more or less, but it's AMD whose proven most successful at this, and not Intel. Intel has tried to come out with a way to move beyond x86's limitations, and has consistently fallen back to favouring designs that instead manage to make x86 faster (and it is to their credit that they've managed this for so long). Nonetheless, the *only* reason to choose straight x86 is its backwards compatability -- which has no value to Apple. The possibilities for a gradual transition to 64 bits -- a transition Apple is currently halfway through -- under x86 are, at best, no better than that posed by PowerPC and the G5.
5. If Apple is sticking with PPC, but simply "shopping elsewhere" for future chips
I agree: I don't think Apple is this stupid. But it's one helluva way to build anticipation for an announcement on Monday, and if the result is egg all over CNet's face, well, that would be pretty funny.
Re:The developer scene... (Score:3, Insightful)
This is complete nonsense. Hardly any major Mac apps in 1993 were written in assembly language. In fact, the ONLY major Mac app that I can recall being written entirely in assembly language was WriteNow.
The big issues for the PPC transition were
A typical application was converted in a few days. For not yet converted apps, emulation was reasonably acceptable because even the first PPC had a considerable speed edge on the 68Ks of the time and the emulator was pretty fast because PPCs have many more registers than 68Ks.
A transition from big endian to little endian would be far more unpleasant because of all the field swapping. The speed edge of Intel processors is a lot less than the PPC edge was back then, and PPCs, especially G4s and above, have an enormous number of registers. Altivec code would be horrible in emulation.
Re:Giving up on hardware? (Score:4, Insightful)
Who cares what chip is in it? (Score:3, Insightful)
Either way, it's not as though the x86 is totally inferior to the PPC. Each has its own strengths and weaknesses compared to its rival.
Short of the zealots (who happen to gravitate to this site), I would tend to think my fellow mac users would feel the same way. The experience is what makes a Mac, not the cpu.
All I can say is that it would be kickass to have a Mac with WINE or VMWare to run 'doze in a pinch without sacrificing too much performance.
Re:Well spank my ass and call me Judy! (Score:1, Insightful)
I should type that up, actually. It's pretty funny.
What's even funnier than his paranoiac ranting is the cover article: 100mb hard drives reviewed.
Re:Well spank my ass and call me Judy! (Score:3, Insightful)
Judy!
For the record, so did Rob Enderle [appleturns.com].
Which basically gives me full confidence that Apple will NOT be announcing a switch to x86 architecture on Monday.
p
Re:The sky is falling! (Score:3, Insightful)
Apple has a history of swapping enemies and allies (Score:5, Insightful)
No, to use Apple terminology, "been there done that" . An alliance with Intel is less shocking than the alliance with IBM. How soon people forget that IBM was once the "Satan" of the Apple universe. IBM actually was a competitor unlike Intel who merely supplies competitors.
As for as MS going PowerPC, well, "been there done that" again. Windows NT 4's retail CD has x86, MIPS, Alpha, and PowerPC binaries. I remember Byte magazine comparing WinNT4 dual PPC 604s against WinNT4 dual pentiums. The verdict, dual PPC scaled better under WinNT4. The only problem was no one cared, Alpha had the performance, Intel had the price, unless you could dual boot the box into WinNT4 or MacOS there was no real point to PPC.
Re:Ooooooh! Oooooooh! (Score:1, Insightful)
Therefore, Apple's "classy" hardware requires that I slap several layers of paper and tape over my iPod's headphone jack, use "run shell script: sleep 60" instead of "delay one minute", and only connect to the internet in Mac OS X at a cost of approximately 20% CPU. Thanks, guys. Here's something that I think would be classy: instead of spending your extra engineering time making my keyboard light up automatically when it gets dark out, why not give me a modem that doesn't suck? Or maybe a wireless card that works with my other OS? Or an iPod that actually plays music through decent headphones for more than eight seconds at a time? That would be really nice.
I'm all for innovation, as long as your innovative new features ACTUALLY WORK.
Re:Any Evidence At All? (Score:3, Insightful)
Seriously. If this turns out to be bollocks--and I'd say the odds are greater than 50%--will CNet be punished at all? Will Slashdot stop running their stories? Nah. It's not going to hurt them a bit.
They've learned nothing from the Newsweek fiasco. Get two independent, on-the-record sources for a fact before publishing it as fact.
Re:Transitive Technologies (Score:2, Insightful)
Not exactly. Like others have stated, at least one side of the math has to be stuck in a register. Couple this with there being less physical registers than these virtual registers that you need, and you're talking about caching in-out registers as need be. For some algorithms, that's no big deal. But you'd be amazed at how quickly the math spills into many registers. This is why that per mHz RISC processors, with many more registers, tend to easily beat CISC processors often by a factor of 2. Besides, accessing the L1 cache is flukish. There's only so much that can be done to lock the registers into memory. And when you do, you've got to access them with extra-long instructions (32-bit addressing + opcode + other operand(s)) which further sucks down the efficiency of memory throughput/instruction cache size.
Aside from this, CPU registers are heavily renamed (remapped) to a whole host of hidden registers to improve parallelism. Essentially, the old AX, BX, CX, DX register model is is simply the public representation of the CPU architecture, while behind the scenes things are very much more RISC like in their implementation.
True. And with something like the Athlon's multiple apu/fpu pipes there's the supposed potential of 200% added performance over what the apparent clock rating would indicate (ie, such should mean that Athlons should be 50% faster than PPC). Of course, real figures are closer to 30% to 100% (ie, 65% to 100% of PPC per clock). This has a lot to do with, again, the caching-like behavior compiler code spits out. This is ignoring simd.
In any case, given the 0.5 GHz to 1.5 GHz advantage that Intel CPUs have over the current PPC chips, it's entirely likely that native PPC code written for OSX could run quite well on a JIT compiler on x86.
Yea, I don't doubt the CPU will be doable. But what about the rest of the hardware? Qemu's non-hand-optimized x86 on x86 system emulator has all sorts of performance issues when it comes to the video card emulation and the ide emulation. Of course, I'm sure that if Apple is planning on such they'll spend a lot of time getting the performance to be more reasonable.
So, I don't think the question really is whether or not it's doable or even if it the result will at all be usable. The things that make it usable are the things like fast response to keyboard, a quick display, and "fast" disk access. But, odds are that the virtual CPU will have a lower MIPs rating than a PPC CPU. You have to figure in that even if somehow you could get a 1:1 mapping for all PPC instructions and keep all the instructions roughly the same size to remove any extra bandwidth needs that the actual translation takes enough time fetching and processing on its own to greatly negate the extra Hz available.
It's in this that I think your Dynamo reference is lacking. While Dynamo might have been able to see a 20% improvement in some programs, actually trying to pull off profiling across an entire system to optimize would be near madness; only a handful of programs are likely to see a real improvement with profiling, while the majority are likely to be slower. This is mostly because profiling is extra work and won't necessarily do the best job. Even if it does, the actual profiling work might counteract this. Something like HLE, with a static list of known constructs, is probably a better trade-off. I'd be amazed that even with this that top of the line x86 32-bit would be any faster than top of the line Apple PPC.
Re:The sky is falling! (Score:3, Insightful)
Re:April Fools? Right? (Score:1, Insightful)
Dude, get a grip. First of all, it's not news, it's a rumor. Secondly, try to realize that just because MS sucks, not all software that runs on x86 machines must also suck.
Once you got NeXTSTEP installed on a 486 box, you really could forget what the CPU was. It was *just* like running on a NeXT slab, only quicker. Apple has far more resources than NeXT did (obviously) so I would expect Apple to cover far more devices than NeXT was able to.
Re:No. (Score:3, Insightful)
Nonsense (Score:1, Insightful)
Apple's online store says the highest-end Dual 2.7GHz model is "Estimated Ship: Same business day".
(I'm always amused what passes for "Insightful" on /.)
Re:April Fools? Right? (Score:3, Insightful)
Re:The 68K-PPC transition really sucked (Score:1, Insightful)
??? The PowerMac 8100 ran emulated 68k code faster than the fastest 68k mac ever made (the 840av).
Re:easy to trace (Score:5, Insightful)
Are you serious? Either you completely missed the point of that post or I'm missing the humor in yours.
Intel != Pentium. The poster was just saying that because Apple now owns a big chunk of PPC IP, it will simply have another manufacturer (i.e. Intel) supply them with PPC chips. Nothing more, nothing less.
Re:Overlooked points... (Score:3, Insightful)
And monkeys will fly out of my ass.
Why would Intel devote an entire design team to build something for Apple with it's meagre 2% PC market share? It makes no financial sense.
Re:Hmmmm (Score:3, Insightful)
Windows only installs on all modern x86 for two reasons:
1) Most hardware strives to be backward compatible to an IBM PC (BIOS, CGA/VGA/VESA, PCI/ISA, IRQs, etc)
2) Windows comes bundled with a lot of drivers
An Apple x86 won't be:
1) Backwards compatible with an IBM PC and it's legacy hardware
2) driver compatible with Windows
Re:Architecture change performance hit? (Score:3, Insightful)
The answer is clear... it's not x86 anything. They're not switching architectures again. Not this quickly. (they haven't even fully exploited the 64-bit-ness of the damn G5 for shit's sake.) They just need more capacity. IBM is having troubles with capacity (wonder if this will hurt XBox 360? It might).... Intel's got fabs out the WAZOO... and could make TONS of chips for Apple if they needed it.
In some people's eyes, Apple may be dumb, but certainly not that dumb. And Jobs isn't about to shit all over the momentum he's got with the Mac line by switching the damn CPU.
My guess is this is much ado about nothing... and Intel's fab power is what Apple is tapping into for it's own PPC chips.
The switch to x86, even if it occurred, would be more complex than just "keeping in step" with the OS. There's much more to it than just a "recompile"...
Re:easy to trace (Score:3, Insightful)
You're right. Intel probably wouldn't design a PowerPC chip, and the GP isn't suggesting that. In fact, if Apple wanted Intel to design a PowerPC chip for them, they already have more than enough IP and cross-licensing to do that without invoking a "Moto" clause. No, the "Moto" clause, if one exists, would be one that gives Apple the right to tell IBM "You didn't meet your numbers. Hand over the dies," and have them comply without penalty.
I have no idea if this is true or not, but it seems a hell of a lot more likely than Apple moving to x86, except maybe Centrino for laptops or tablets or something.... If Apple does move to x86, though, that's the day I dump all my stock and start looking for another computer vendor. If I want a shitty PC, I can buy one for a whole lot less from other companies.
As for me, I pay extra money for Macs because they aren't x86, because there aren't hundreds of thousands of script kiddies who know the chip architecture and have pre-written shell code floating around in their heads. It's not a big improvement in security, but it does make some difference. If you could write one exploit for Mac and PC versions of the vulnerability with only subtle changes in the system calls, you can bet your ass the Mac platform would be as virus/worm-ridden as Windows faster than you can say "Thank you, Steve."
More than that, since Darwin is open source, if the Mac moves to x86, I guarantee you could make Mac OS X run on any cheap $200 x86 PC in a matter of hours just by doing simple ports of a handful of drivers (and by substituting the boot loader and kernel from Darwin if they didn't use a PC BIOS)... which is exactly the reason that rumors of Apple switching to x86 are without any doubt, utter bullshit.
There is NO WAY WHATSOEVER for Mac hardware to move to x86 without it decimating Apple as a hardware company. A separate Mac OS X for x86 MIGHT be possible (at high enough cost to make up for the loss in hardware sales), but imagine being able to pick up a $129 copy of Mac OS X and run it on a $200 PC just by duplicating the DVD and adding a handful of additional chipset drivers.... Possibly as simple as running the x86 equivalent of XPostFacto and sitting back....
So raise your hands if you would still buy a Mac if you could just install Mac OS X on any $200 PC and get the same perforrmance and the same user experience as a Mac costing significantly more. Anybody? Anybody? I didn't think so. No, this story is about using Intel's fab to make the 970. It has to be. Anything else would be suicide for Apple, and I KNOW Steve isn't that stupid.
Re:OpenVMS: VAX to Alpha to Itanium (Score:3, Insightful)
You can buy an Opteron machine for about a tenth to a fifth of the price, get better performance, use less power, and less air-conditioning. Have you used Itanium ?
No, but I know people who have.
I also know that intel and SGI are having to give them away for free because they're so bad, no one will pay money for them.