Apple Computer Inc.’s worries about watts are unfounded, according to IBM. The company could build PowerPC chips that satisfy the needs of the entire range of Apple’s product lines, including portables such as the PowerBook, said Rod Adkins, vice president of development for IBM’s Systems and Technology Group.
Company tries to downplay flaw in architecture, news at eleven.
Of course IBM could have done it, couldn’t they, but guess what, it’s at least 1 year late and not yet in the plans.
Why bother commenting on that ? Apple is not going back and noone else is doing desktop let alone laptop machines based on Power….
If this was the case, then why didn’t they just DO it for Apple when they were still in cohoots with one another?? I don’t understand IBM. Personally, I thought their chips were great! But it seems as though they love punishing themselves since they constaqntly shoot themselves in the foot. First with the IBM PCs not being cloned, then with them opening op COMPLETELY and being overrun by clones, OS/2, and now this.
Is it THAT hard for IBM to admit that they didn’t want to have Apple anymore what with all their interest in consoles and servers? Jeez! Makes me wonder about them sometimes…
“Our point of view is that was somewhat misleading,” said Adkins, who oversees product development for the Systems and Technology Group.
“Apple positioned it that way in the public,” Adkins said in an interview with eWEEK.com. But “Obviously I have a different point of view, because there’s nothing about power architecture that limits you in any way in terms of power management or power efficiency.”
But instead he said the nature of Apple’s relationship with IBM, including the fact that it had another processor supplier in Freescale Semiconductor Inc., is what limited the IBM chip roadmap available to it, Adkins said.
“They had Freescale primarily for the low-end and mobile solutions, and they really had IBM focus more on PowerBook, xServe and iMac. That’s where we collaborated deeply with Apple,” Adkins said.
But “There’s really nothing in the architecture that prevents having an end-to-end line that can serve [all of] their needs.”
…
“I think Apple decided…that they wanted to focus more on their software investments and that they wanted to leverage the Intel roadmap end-to-end,” he said.
Interesting to hear that.
Talk is cheap my friend. I didn’t see any 3GHz G5s come out and 2.7GHz Dual G5 is water cooled, in other words, overclocked! If the focus of IBM was on high end line of works, as stated in the article, why didn’t we see faster processors in that area?
The low-margin XBox/PS3 “cell” chips aren’t going to carry it for IBM anymore, without Apple shouldering chip-fab costs on the high-margin G5’s. I’ve said it before: IBM is screwed. And it’s too bad. Its chips right now ARE way better than Intel’s, and it would have taken minimal effort on their part to keep it that way.
It’s going to be a very long time before Intel gets anything that performs as well, or has as low power consumption, as a Power or AMD chip.
I don’t know what Steve Jobs has been sold, but he’s been sold a non-existant turkey. I would have started serious discussions with IBM and others about where they saw the Cell processor going before jumping ship to an inferior architecture. Power consumption? Really? Jobs has shown that he has no clue at all about the cip world.
So IBM is claiming they, in theory, could produce PowerPC chips to fit Apple’s energy/heat requirements. Separately, they claim they could, again, in theory, fit a G5 into a laptop. But could they do both with the same chip?
And if so, I echo what others are saying, why didn’t they? I can’t believe Apple wasn’t asking for it.
Meanwhile, it seems like a battle of roadmap vs. roadmap. What IBM claims it can do vs. what Intel claims it can do. Still lacking a G5 laptop or a 3Ghz G5, I can see why Apple might have its doubts.
Personally, it makes sense to me that Apple’s big concern might be portables, and Intel’s shown some promise with both its processors and motherboard chipsets. So maybe Apple was just going with the better total solution. Maybe the chipset has as much to do with the deal as the main processor.
I dunno. Just speculation.
amusing how one and all fell for the bait of it being a technical issue. more likely it was about $$$$$ now and in the future. i suspect the technical issue played little to no part, but made amusing press.
Yeh! Everybody seems to love IBM nowadays. However try to explain someone
how a simple 64 bit constant is loaded in this architecture and the headaches start to grow. Hint:
a = b * 16 + c * 256 + d ….
PowerPC is clearly hitting the wall performance wise.
David said:
It’s going to be a very long time before Intel gets anything that performs as well, or has as low power consumption, as a Power or AMD chip.
Go read their roadmaps to see, they are working on this. I don’t believe the chips will necessarily going to be better than AMD’s but they will be competitive. Besides, Apple are after more than just the technology itself, there’s a whole shake-hands-show-the-nipple thing going on regarding Apple being Intel’s second Tier-1 OEM, i.e. putting Apple on the same pedistal as Dell.
Who is an IBM exec told me that IBM liked the Apple relationship because it brought about the buzz, but it never made any money. Now that they are making ranks in the SuperComputer business with Power, and are proving its abilities, they don’t need Apple to keep power momentum going. Why spend 100’s millions of dollars to build a new plant to produce new apple chips (that you can’t use for anything else) that you don’t make any money off???
Because you could build an ATX form factor mobo and actually sell it if the arch is that wonder full? If would give you that much advantage over the competitive offering the marked would be just right there, since the OS could cost nothing.
I have worked for IBM before and based on that experience i would not be surprised that doing busines with them would be a headache. IBM moves at a snails pace and is wrought with business crippling bureaucracy. Considering the price vs performance Apple will be able to produce lower cost PC’s with the intel line and still be the mercedes of x86 platforms. Although those new SUN Opteron workstations are pretty slick.
I think that is the real point, that although IBM could have made the chips Apple wanted, Apple is probably IBM’s only significant customer that needs them and it wasn’t worth IBM’s effort to do it.
Then again, maybe Apple wanted to move to x86 anyway. Once Apple is fully switched to x86 and has built up a good driver base they might consider selling OSX on it’s own, or licensing it out to other manufacturers, maybe getting Dell to make cheap low margin machines to build market share and carry on with building the high end, high margin machines themselves. With Jobs in charge it’s hard to predict what Apple’s long term plan’s are.
i happen to think apple was looking for an excuse to jump ship.
It came out after the discussions that IBM had been looking for more money from Apple. I think the bit that was left off of the “We could have made low-power G5s…” statement was the “…if it would have been profitable for us to do so” addendum. Apple were a bit player for IBM, and they weren’t willing to pay more for the chips, so IBM sidelined them and concentrated on the console business where, having signed deals with Sony, Microsoft and Nintendo, they are guaranteed a solid return.
Still, the bad karma surrounding them shows the dangers of not givin Steve Jobbs what he wants!
Heh, david you realize that Apple will be using the Pentium -M based chip line and not the hot plate based Pentium 4.
The P-M is A LOT cooler and higher performing than the G4 and a lot cooler than the G5 (performance measures are not available since it is not yet available on the desktop) the P-M will no doubt be able to perform on the level of the G5 in the next year or so.
> Why bother commenting on that ? Apple is not going back and
> noone else is doing desktop let alone laptop machines based
> on Power….
It isn’t immediately clear that IBM cares whether Apple changes its mind or not. The simplest answer is that IBM wants to counter the negative publicity associated with Apple selecting the next few generations of the P4-M because of scalability problems and power consumption concerns. It’s a matter of public image, and IBM surely doesn’t wish for their architecture to be associated with the attributes of the PPC970 Apple cited as problematic for their vision of the future.
It appears that both companies are engaged in attempts to save face regarding Apple’s transition.
IBM: “We can do anything Apple wants, we just aren’t even close yet.”
I didn’t see any 3GHz G5s come out and 2.7GHz Dual G5 is water cooled, in other words, overclocked!
As far as I understand it, they are watercooled because the audio pros wanted silence, not because fans couldn’t cut it (though they do run hot).
dude… what?
MS and Apple are not joining forces and Dell is Apples direct competition why the heck would they allow Dell to sell OS X.
The main reason for the switch is the Pentium-M. IBM and AMD don’t have anything to compete with intel in this area.
IBM:
“We could have done it. We just didn’t feel it. Why? Because I said so. Yeah that’s it.”
Good PR move.
After all, there wasn’t any demand for a PowerBook G5, so why bother?
bill “how can i stop being accused of being a monopoly and hopefully put a damper on linux as well”
steve ” well you know if we swicthed to x86 chips and you could order a system that had a choice of windows or apple then there is no one that could accuse you of being a monopoly”
bill “hey and then the “alternative” OS would be apple which makes my apple holdings skyrocket as well”
micheal walking by notices the guys ” hey guys whats up?”
jake tate reporting… think about it….
For the difficulty of loading 64bit constants, this is not specific to the PPC, each RISC has the same “problem”.
But this doesn’t mean it is an important problem, does the performance of applications is limited by the loading of the 64bit constants?
No, so why throw transistors at unimportant activities (performance wise)?
That’s what RISC is about, use transistors where it matters.
I suggest that you read Hennesy&Patterson book to understand it.
> The main reason for the switch is the Pentium-M. IBM and AMD
> don’t have anything to compete with intel in this area.
This isn’t strictly true. While AMD hasn’t dedicated as many resources to a strictly mobile platform, AMD’s K8 architecture was already such that it was reasonably efficient power-wise and each revision has afforded improvements in this area. The two classes in Turion line of mobile processors, but especially the MT, will be quite compelling power-wise. There’s no clear competition for the ultralow-voltage Pentium M, though. It’s not like AMD is just sitting around with its thumb in its anus, though.
i happen to think apple was looking for an excuse to jump ship.
You mean Apple risks a lengthy transition, contradicts its own marketing and angers its hardcore fans just for the fun of it? How deluded are you?
I’ve got a PC that’s about 5 years old. It’s 1GHz Athlon. I have an iBook 933MHz G4 that’s about 2 years old. The iBook should be quicker but when I encode video with mplayer on both computers, the PC is much faster, maybe twice as fast as the iBook. I really like my iBook and all but when it comes to raw CPU power, it lags behind a computer that’s 3 years older than it. That’s pretty sad.
Go read their roadmaps to see, they are working on this.
Thats my point exactly – Apple have been sold a non-existant turkey. Did you read the part where it says all that doesn’t exist? The stuff Jobs talked about is complete fresh air.
Besides, Apple are after more than just the technology itself, there’s a whole shake-hands-show-the-nipple thing going on regarding Apple being Intel’s second Tier-1 OEM, i.e. putting Apple on the same pedistal as Dell.
Apple are playing a dangerous game putting themselves in the same bracket as Dell, and pricing themselves to the higher end isn’t going to help. They don’t have much of a business to lean on as it is.
Given the recent allegations by AMD (which I tend to think have merit), perhaps Intel made Jobs/Apple a deep sweetheart deal (bordering on predatory pricing). Intel certainly doesn’t want PowerPC to look too good outside the server room, and thought it just might cut into their other business.
I do not claim this is true, but do consider it consistent.
Heh, david you realize that Apple will be using the Pentium -M based chip line and not the hot plate based Pentium 4.
The Pentium is a half-decent laptop chip and nothing more. It’s not something that Apple can use as the basis for a premium, high-end and high-priced entire range of products. Even Sun had the good sense to go with AMD because it differentiated them and gave them the actual performance they could boast about.
A relationship with Intel is never an equal one, especially for a small player like Apple. If they think that relationship will be any better than that with IBM then they’re going to be disappointed.
You know, there are other people employed at Apple other than Steve Jobs who have decision making power.
The Pentium is a half-decent laptop chip
That should read Pentium-M….
I didn’t see any 3GHz G5s come out and 2.7GHz Dual G5 is water cooled, in other words, overclocked!
Have you heard a G5 running? No, neither have I. All modern Pentiums have a fans on top that belong in a Rolls-Royce turbofan, and are just as loud.
Although I personally would have prefered Apple going the AMD 64×2/Opteron path – like one of the other posters becuase it would have differentiated them from the likes of Dell, HP and even Alienware etc. However, as a system admin its pretty hard to convince senior management to go a non Intel route – I’ve only just got the IT director at work to purchase 2 Sun Opteron workstations over Dell Precision workstations.
Moving over to the Intel bandwagon could finally give Apple a break into the non media corporates where HP and Dell dominate, although the margins for these kind of machines would be quiet small.
Anon
yes if you think apple moved to intel to compete with dell….think about it, why would they….
got to be more to it…….right…..
But can you put a fan on a G5 and assume it won’t crash and burn? you can put a watercooled system on a pentium if you want, it’s just going to cost a hell of a lot more than the standard cpu fan.
>This isn’t strictly true. While AMD hasn’t dedicated as many resources to a >strictly mobile platform, AMD’s K8 architecture was already such that it was >reasonably efficient power-wise and each revision has afforded improvements >in this area. The two classes in Turion line of mobile processors, but especially >the MT, will be quite compelling power-wise. There’s no clear competition for >the ultralow-voltage Pentium M, though. It’s not like AMD is just sitting around >with its thumb in its anus, though.
So basically you wrote that entire paragraph to say what I just said:
There’s no clear competition for the ultralow-voltage Pentium M
When Steve Jobs made his announcement he spoke of performace/watt. Basically as of right now Intel has the best performace/watt chip the pentium M. I’m sure AMD and IBM will eventually make something as competitive, but right now intel has one this one.
Thats why on my desktops I run AMD and on my laptop I run Intel (P-M).
Whether Apple is making the right decision is unknown, but I can’t see this as a stupid decision.
And that’s why IBM ditched Apple
Tell me: did you really believe in apple’s “IBM did not fulfill our needs” excuse?
Look:
IBM has been making more than 50% of its revenue in consulting and services for years now.
They sold the PC line.
Apple is a TINY market: they are irrelevant.
It is only reasonable for IBM to stop putting such an effort in building chips for a non profitable market such as the Apple one. They did not even get good PR, since close to no-one knew Apple used IBM chips, actually (outside our geek club, that is).
Pull your head out of your ass if you think the Cell processor is a desktop chip. How many articles have we had on this? Hell, I think there’s one on the front page.
From what I’ve heard of the IBM corporate culture this guy is probably doing some internal pr for his department. After all wether IBM really cared about Apple or not somebody lost that account and you know somebody somewhere is going to get canned over that. It’s good to publicly show that you and your people weren’t the ones to blame.
So basically you wrote that entire paragraph to say what I just said:
There’s no clear competition for the ultralow-voltage Pentium M
Uh, I was referring to a specific class of Pentium M chips, namely the 7W TDP ultralow-voltage processors that were sub-1GHz and most definitely not anything that will be in the future Powerbooks and Macs. Maybe you should learn what you’re talking about before writing a long post attempting to educate others on it.
>> Heh, david you realize that Apple will be using the Pentium
>> -M based chip line and not the hot plate based Pentium 4.
>
>
> The Pentium is a half-decent laptop chip and nothing more.
> It’s not something that Apple can use as the basis for a
> premium, high-end and high-priced entire range of products.
Maybe you should tell that to Intel’s platform roadmap, since the Pentium M is going to be the basis for their future processor lines. On-die memory controller, SSE3, x86-64, mutli-core Pentium Ms. It’s a shame they’re just a half-decent laptop chip! Actually the Pentium M is probably the most respectable product from Intel in years.
apple’s pattern of changing cpu architectures forces their customers and developers to upgrade sooner than they normally would. In a long term view, the revenue difference between switching architectures and sticking to one becomes significant.
It would seem the hassle of changing architectures would hurt them, but i think it has the opposite effect.
@Andrew: The watercooled G5 still has a fan, for its radiator. And its a pretty small radiator, which means that the single fan still has to run pretty fast (ie: loud) to move enough air. Certainly, the whole system isn’t much quieter than using a big heatsink/fan combo would be. I suspect the real reason watercooling is used is because it’s effective at handling a larger thermal load than a heatsink, which would indicate a processor generating quite a lot of heat. There is no way a G5 would dissipate that much power running at 2.7GHz, not unless Apple upped the voltage past spec (read: overclocked) to do it.
@David: Yes, I’ve heard the G5. Actually, the earlier models were rather loud. The newer models are quiet, but then again, so are Dells. Say what you want about them, but their ducted-intake case design results in very quiet systems.
I doubt Apple would have made the move to Intel for any one reason, chip efficiency or otherwise.
Ir was probably some combination of:
* No chip supply issues — ever.
* Keeps options open for Trusted Computing
* IBM failing to hit performance targets
* Ability to outsource motherboard design to Intel, improving Apple profitability.
* Vastly improved mobility products in very short order
* End of the megahertz marketing problem
* Future Apple/Intel projects we won’t know about for a while
* Cool new “MacTel” trademark
i prefer PowerPC, but as Steve Jobs said way back in his 1997 MacWorld keynote, the soul of the Mac is the operating system. It would seem that while OS 8 has yielded to Tiger, Job’s opinion that user experience is primary has remained consistent over the years.
My powerbook G4 1.5GHZ Laptop has a battery life of less than an three hours. Compile a large project…it becames a hotplate. I’ll welcome an x86 laptop that runs longer and cooler anyday. The mhz myth is over. For me it’s about the OS. ibm doesn’t give two shits about apple. They are in bed with sony and the cell processer now, which btw will never make it in a laptop. Don’t worry about amd…if they make anything better than intel….apple will use it. I’m happy about processor change….no more overclocked nonsense…tricking users to think that processor speed was getting faster. If apple was selling a dual xeon 3ghz box right now…I would buy one in a second. I’m looking forward to the day where I will be able to buy an upgrade plain pc video card and putting it into my mac. Fsck ibm. Intel will deliver…worst case…a mac will be on par with the best pc/laptop.
Out goes PowerPC and Out Goes Open Firmware.
in comes x86, and PC Bios.
Well there goes Target disk mode. PC BIOS can’t support it.
There goes pure Plug and play. As PC BIOS can’t deal with it well.
out goes working suspend/sleep as MSFT has fskced it up so much to stop Linux from using it.
Apple is switching to the crappy carrying baggage from 1979 x86 line.
And before you PC fanboys start There is a reason why target disk mode was never copied, there is a reason why your brand new dell only has 15 IRQ and needs 30, There is a reason why Windows crashes randomly on suspend/sleep.
PC Bios where never designed to handle the amount of hardware we have today. In order to make it work, they tack on crutch after crutch. When the new Intel Powerbooks ship with features not working or causing kernel Panics. Blame Steve Jobs. he is the one who chose faulty hardware. As for comparing a pentium M Chip that is on the drawing board vs chips on hand. yea right. If you can’t compare them because they don’t have benchmarks yet, then they don’t exist and don’t perform. Theoretical performance doesn’t work.
As for Cell chips, they may not be the fastest, of course if you make a Dual Processor machine with one cell, and one regular power chip, how much performance would you get?
Quote: “If this was the case, then why didn’t they just DO it for Apple when they were still in cohoots with one another??”
Well duh! Anything is possible in theory. The cost of doing what Apple wanted? That’s a different story. Was Apple going to foot the bill for the re-tooling, etc etc? I bet not…
Apple wanted a particular chip, for a particular purpose, and then wanted to modify it to do other things. That’s fine, but they have to be realistic that altering things to suit other areas can be costly in terms of development and testing.
Apple wanted to the cake, but didn’t want to help bake it. Steve Jobs tried to pressure IBM, IBM told him to fuck off, he had a dummy spit and decided to side with Intel. There’s not much more to it than that.
Dave
Apple needs a consistent supply of cooler, faster and cheaper general purpose chips than the competition across it’s entire product line to gain market share. IBM knows this, they were in the PC business themselves once.
IBM promised and FAILED to deliver. They QUIT the PC market altogether, even selling off their PC division. Apple has no choice but to go to Intel, they need to sell hardware, like PowerBook G5’s a year ago.
IMO IBM is a bunch of losers, they had their opportunity and blew it. It’s obvious by this article their ego is severely bruised and they deserve it.
The new PowerMac G5’s with 970MP dual cores are going to be so hot as to require liquid cooling across the entire line. And the fan revs of the Dual 2.7 is nothing like the quiet Dual 2 Ghz’s that were very nice, I can only imagine what the Dual cores will sound like, having two hot chips so close to each other. Make a nice space heater.
The Cell chip is WEAK on CPU power, but high on graphics power, it makes a good game chip and that’s about all. And this is coming from early developers working with it right now.
I suspect the BIOS on a MacTel will be nothing like what you see on a PC, I’m guessing Apple and Moto/Freescale has something up their sleeve. Given the iTunes phone that Motorola is coming out with show they are continuing their relationship.
I think this chip (?) will lock Mac OS X to Apple hardware quite nicely and still keep all the functionality we have now. Then we can run ‘x86 based software natively, perhaps in a shell or something so it’s safe.
So many third party programs are riddled with faults that place the entire OS in jeopardy. Adobe and Real Networks are the latest culprits.
Mac OS X run on a separate processor might do the trick very well.
I’m not sure apple will go with the PC bios, intel have been wanting to push EFI (Extensible Firmware Interface) onto x86 for a while. But no one wants to take the leap. So I think there is a reasonable chance that the Macs will not have a pc bios… However It wouldn’t surprise me if it could emulate one (similer to how Macs load new world roms).
Making MacOS dependant on EFI would limit the ability to run MacOS on non apple hardware (you could still emulate / virtulize it).
You can bet openfirmware will be in the Intel Macs. No way around it. It is not inherent to the PPC platform so why would someone think Apple would be forced to use a PC BIOS??? Get a clue.
No Apple’s Intel Macs will run OS X only. You may be able to jam a PC BIOS in there, but the functionality (read: working) will not be there, no matter how hard someone tries. It’s all about BootX and BootX will not work on a PC BIOS.
As for Adkins, yeah whatever. We can do this and that and the 970fx is just fine for laptops. PowerTune baby! Get real. That sucker produces more heat and uses more power faster than they will ever claim. Hell FreeScale’s docs on the 7448 say that is consumes 21W at 1.5GHz. What is the 970fx at that speed, somewhere around 50W! Goodbye balls.
It’s specifically stated in the development documentation for porting to Mac/x86 that OpenFirmware is not being used. Just stop embarrassing yourself, if not for your own self-respect than to spare others from your misinformation.
No G5 3.0 even today, 2 years later, no dual core yet, no laptop chip yet…
Intel would love this relationship because, unlike microsoft, Apple loves to jump on new features, and Intel hasn’t’ had a partner that would implement what’s available on the chip all the way back to the 286. So, you can see why Intel would love Apple on board.
IBM: Money and Commitment seem to be the problem.
IBM lost focus on Apple’s business.
IBM didn’t see the good pr Apple brought to the Power line.
IBM forgot the benefits of the relationship.
– More developer experience with Power.
– More support from Linux on Power.
Looks like the type of decision a Stock Market Analyst would make.
The business isn’t very profitable today: Drop It.
The Potential was hugh, but it won’t help your 3 month projections.
IBM is now in damage control; they’re now viewed as a stingy CPU producer, unwilling to sink the millions if not billion to R&D to make their chip work.
If they had a G5 ready for powerbook, why didn’t they make it available? to whose benefit would it be if Apple refused the G5 chip? IBM also failed to reach 3Ghz – IBM is the sole contributing factor for the failure to keep to their road map; Apple is the customer, IBM is the producer; they failed to meet Apples demands, they also failed to meet volume when required.
Anyone remember the Xserve fiasco? the PowerMac fiasco? these are fiascos all caused by IBMs lack of capacity at their chip plant, either that, or they’re simply playing Apple for a chump – well, thats their problem; Intel now as a show pony they need to show off all their latest cpus, chipsets and wireless technologies – what has IBM got? selling a few blades which haven’t exactly set the world alight in regards to volume, and let us not forget the over priced, third rate, ‘build your own’ kit from Pegasos with an out of date processor, slow expansion buses and a lack of software from third party vendors.
And won Microsoft and Sony.
IBM will be selling more chips to Microsoft than Intel will to Apple. IBM are going after their big bucks. Sad to see them leave though
Apple’s reasons for moving to Intel had nothing to do with IBM not providing a fast PPC at low power. IBM in fact DID deliver the goods: a 64bit PowerPC that runs at up to 4GHz and at low power – simultaneously! Only Microsoft, Sony, and Nintendo are buying it, not Apple.
People ignore that by simplifying the PowerPC, IBM has gotten it running at up to 4GHz. It’s not as fast as the G5, but it’s not THAT slow, plus it makes it easier to make multicore CPUs. Microsoft is getting three cores, each with dual SMT, for a total of six logical processors with a BIG level 2 cache – ALL ON THE SAME CHIP RUNNING AT OVER 3GHz!
Now remember that these are settop boxes/game consoles. They can’t be power hogging monsters with water cooling or huge fans. Look at the Nintendo Revolution – it’s positively tiny. These chips were PERFECT for laptops. I imagine IBM is puzzled why Apple didn’t want a nice 3.5 GHz 6 core laptop… and so am I.
The Cell proc @ 4 ghz is not the same as a typical PPC chip found in Macs. I do not know however the proc MS is getting…so I cannot comment on that.
It is more that IBM did not…
anyway, after the usual corporate delay, now we get the corporate moaning of the people who probably got the beating for loosing a customer.
YOU need to get a clue. IBM is making a server blade using two Cells. The processing power lost due to the simplifications is NOT that much, and can be mostly offset by good scheduling by the compiler. Get your facts right before replying.
ofcourse ibm could create those cpu’s. they have the cpu architecture and technology that far exceeds that of intel.
but ibm doesnt do charity – everyone can get ibm to make them powerpc chips, but if apple doesnt want to pay up for new and better powerpc chips, then ibm wont have reason to manufacture them – hence, since apple doesnt want to be inovative, they will have to find a cheaper alternative instead. in comes intel.
it really boils down to ibm having focus elsewhere, on much larger markeds, and apple whining doesnt make their markedshare any bigger or bring income to ibm.
pathetic by apple really. and all the zealots are quick to blame ibm – but atleast it makes for a good laugh reading this thread.
Switching to Cell would have been just as hard as switching to x86. The Cell is _not_ a PowerPC. It has a very simplified PowerPC unit as it’s base processor but it wouldn’t run OS X out of the box or any applications compiled for a PowerPC. So Apple’s choice was either a processor designed for game consoles that is difficult to program for and will be mostly underused in a desktop machine or Intel’s Pentium M based x86 chips that use very little power and give top of the line Pentium 4 and Athlon 64 systems a run for their money.
Oh, and Cell doesn’t have Altivec either so that isn’t any different from switching to x86.
all your points are invalid since cell isnt a set technologi – if you want a beefier powerpc part, insert one. if you want altivec, insert an altivec unit. if you want more or fewer spu’s, ask ibm to make you one that suits your needs.
its just the current cell showcased which is ment for a gameconsole.
I think IBM made a big mistake by not convincing MS, Sony, and Nintendo to use an improved version of the PPC. It almost seems that MS might have tempted IBM into destroying the PPC line, in order to kill two birds with one stone. In addition, Apple was planning on plan “B” for some time. They probably wanted to do an Intel move anyway as soon as “Classic” was killed. Apple will also have a complete motherboard from Intel. It is not just the processor. They will be on equal footing with DELL, and have their own OS! Sounds like a good move regardless.
IBM is making a server blade using two Cells.
I’m looking forward to see the benchmarks.
The processing power lost due to the simplifications is NOT that much
How do you know? The G5 has two each of integer, floating-point, and load/store units. The Cell’s PPE only has one each. Furthermore, the lack of hardware branch prediction and out-of-order execution, means lots more pipeline stalls. And with a 20 stage pipeline, those stalls will hurt badly.
Therefore I think the Cell will struggle to reach even a third of the integer performance of the G5.
and can be mostly offset by good scheduling by the compiler
So performance for existing software that hasn’t been scheduled for it will be atrocious.
But your assertion isn’t true anyway, because memory latencies are rather unpredictable at compile time.
There are two basic strategies for the compiler to deal with that.
Assume everything is in L1 cache and schedule memory fetches close to their uses, but take the full performance hit when the operand isn’t in L1.
Or schedule memory fetches as early as possible in order to hide latencies as much as possible. But this requires more registers, thereby increasing memory traffic due to register spills/fills.
Finding the best compromise is difficult and in any case suboptimal.
Compilers for out-of-order processors simply follow the first strategy and rely on register renaming and out-of-order execution to hide latencies.
ofcourse ibm could create those cpu’s. they have the cpu architecture and technology that far exceeds that of intel.
but ibm doesnt do charity – everyone can get ibm to make them powerpc chips, but if apple doesnt want to pay up for new and better powerpc chips, then ibm wont have reason to manufacture them – hence, since apple doesnt want to be inovative, they will have to find a cheaper alternative instead. in comes intel.
it really boils down to ibm having focus elsewhere, on much larger markeds, and apple whining doesnt make their markedshare any bigger or bring income to ibm.
pathetic by apple really. and all the zealots are quick to blame ibm – but atleast it makes for a good laugh reading this thread.
Bullcrap; its IBM job to sell PowerPC processors, return part of the profit to share holders, and invest the rest into their next version of their processor – it isn’t Apples responsibility to develop the processor, just as it isn’t HP or Dells responsibility to develop processors with AMD or Intel.
AMD and Intel sell products to customers and OEMs, and quite frankly, it isn’t the OEM’s responsibility to pay for ‘extra development’, one would expect the company to improve their own products! they are in the busines to make money off selling competitively priced/performance processors!
IBM failed to live up to the promises they gave Apple; rather than Apple going down the old path of broken promises, as with the case of Freescale, they chose not to risk another fiasco – the x86 processor was the next best option.
Want a million units? sure, Intel will have no difficulties fullfilling that order compared to the fiasco surrounding the 970FX release, and the delays in supply to Apple so they could ship faster Xserves and PowerMacs.
Intel has a road map, and they have been sticking to it; coupled that with the fact that the WHOLE motherboard, processor, firware, wireless networking will all be supplied by Intel, the only thing Apple will be left to do is design a box around the motherboard, create the drivers, and dump the whole thing in a nice case for the end user.
Intel was willing to take more of the hardware development on, and as a result, if Apples computer growth keeps up, which is above the industry average, Intel is in a perfect spot to benefit from the rise in demand for Apple products; coupled that with the possibility of Apple either being able to offer faster products at lower prices, since they no longer have to carry the large burden like they used to, be prepared to see computers a few hundred dollars cheaper, and possibly some new software titles as Apple prepares to expand its portfolio.
we all see it when the cpu’s are released. it’s not very good to speculate about dream cpu’s. i mean noone here has seen cell or the 3core ppc in action.
i heard that cell has vmx (its ppc core).
of course, the ppc cores in cell/3core ppc ms cpu are much simpler than the g5 where, but i don’t believe that article, that the 3 core ppc ms cpu is only twice as fast as the celeron 733mhz of the xbox 1. we will all see, but for now, i don’t believe it.
the only thing i know for sure is, that if nintendo uses a dual g5 1.8ghz system for their revolution, this cpus are much slower than x86 cpus.
i can’t speak much about it because i never used a mac g5 before, but i saw a dmesg of a linux system running on a g5 1.8ghz and it says it has something around 1222 BogoMIPS. (not that this is a reliable benchmark or something, i wonder if it has some meaning).
of course, the software on revolution/cell and xbox360 will be well optimized for the VMX and/or SPU units of the PPC’s or cell cpus in that consoles. and that will give a really good performance. i have seen a benchmark on linux ppc where a g5 1.8 runs a vmx optimized rendering or something..programm and is as twice as fast as a x86 3ghz p4.
i repeat: we will all see, when the consoles are released
calm down my ppc jedis
IBM: “We could’ve made a low-power G5.”
Then, why didn’t you?
Answer: “We don’t want to fool with desktop CPUs anymore. We want to make special purpose ASIC’s for game consoles and embedded applications. This leaves Apple high-and-dry, but that’s just tough for them.”
This makes the second time a chip supplier has done a similar thing to Apple. If the truth were told, Apple should have moved to Intel when Motorola strung them along waiting for high-speed G5s.
Isn’t it about time Steve Jobs steps aside for a new kid on the block that takes charge while our genius Jobs advises and keeps on thinking about revolutionairy inventions?This is the second time a major chips suplier backs of.My guess would be Steve has overvalued his hand of cards again.some people are outstanding scientists and/or geniuses but lousy leaders.The same is apparent for Gates and Balmer,about time they retire to.
Switching to Cell would have been just as hard as switching to x86. The Cell is _not_ a PowerPC. It has a very simplified PowerPC unit as it’s base processor but it wouldn’t run OS X out of the box or any applications compiled for a PowerPC.
The PPE is binary compatible with existing PPC cores, changes are needed in the OS but getting OS X running would be relatively easy.
Oh, and Cell doesn’t have Altivec.
Yes it does.
The last time Steve Jobs ‘retired’ from Apple the company was nearly driven into bankruptcy. They’ve been doing well enough since he came back. ‘It it ain’t broke, don’t fix it’.
apple was profitable under john scully for many years
> For the difficulty of loading 64bit constants, this is not specific to the PPC,
> each RISC has the same “problem”.
>
> But this doesn’t mean it is an important problem, does the performance of
> applications is limited by the loading of the 64bit constants?
> No, so why throw transistors at unimportant activities (performance wise)?
> That’s what RISC is about, use transistors where it matters.
>
> I suggest that you read Hennesy&Patterson book to understand it.
I suggest that you take a look at what modern compilers generate for languages like C++ or Java and realize that Hennesy&Patterson doesn’t take this in to account. On the other hand you may answer the question why those wonderfull RISCs are then so much behind in terms of performance.
If you listened to the software companies reaction (Adobe, Quark, etc…), they were the pusher. They did not like developing on two different platforms.
Personally I think Cell is going to be huge (granted I am an IBMER) because it will not only be in the PS3, Sony is going to put it in everything from the PS3 to HDTVs.
My opnion is Apple needed more software and moving to Intel would give them that.
I love the PowerPC chip since studying it in school at UT back when it was under development. Its a wonderful architecture, much cleaner than x86 and there is no doubt that IBM could deliver any chip variation needed. The problem is IBM, YOU CHOSE NOT TOO DELIVER YOU IDIOTS! In one of the greatest blunders of the computing industry they let Apple slip away. Apple may not have been their biggest chip consumer but they where by far their biggest chip advertiser. Without Apple constently putting PowerPC out there who is going to see it now? The people that buy game consoles sure don’t care whats inside the box they just care it works. The bottom line it was a very bad move by IBM.
ok.. so where’s that powerbook g5?
why’s the powermac water cooled?
where’s the 3.0ghz by summer 2004?
i’m sorry, but IBM dropped the ball. end of story
1. I want XScale in the next iPod. It’s t3h best ARM derived architecture ev4r. Intel would jump at the chance to show off to its hardware partners that the almighty iPod is powered by xScale, and you KNOW they would give them a sweeter deal than even Dell or HPaq just for the buzz (the iPod is hotter than the sun).
2. Intel does motherboard chipset design, saving Apple the necessity of designing Hypertransport Northbridges and southbridges themselves. They can just use whatever newest thing and slap an ICH in there and *poof* new Mac revision in half the time.
3. Like many said, Intel HAS NO supply issues. Even AMD is troubled by these, and IBM wouldn’t, but for their three next-gen console commitments.
4. Netburst (traditional Pentium 4 processor) is dead. It’s too big and too hot and keeps getting spanked by the latest Athlons in power/performance AND price/performance comparisons. Banias’ (Pentium-M) descendants are the future. They will be the new desktops, the new Centrinos, and the new Xeons and this is what Apple wants.
5. Apple will eventually go EMT64 exclusive. I personally wish they would do it off the bat, but they are chained to Intel’s processor plans, and Intel is trying to stall somewhat on 64-bit laptops (just like they stalled on 64 bit desktops and non Itanium 64 bit servers until AMD forced them to play catch up). Appe will also embrace EFI so they can (hopefully) offer most of the Open firmware features on MacTel. This isn’t speculation, as they will follow Intel’s
lead on everything where it benefits them. Theyu are also planning on implementing an extra feature chip to identify Macintoshes as not PC’s. MPAA and RIAA will MAKE them implement the DRM silliness. iTunes depends muchly on the willingness of the copyright holders to play nice, and they <3 the idea of hardware DRM.
6. Earlier posters are also right, there is no ONE reason Apple left the fold. It was like a snowball rolling down a mountain in a cartoon — it became an avalanche and they had to break up. Reasons include: Apple is a niche player. IBM’s attention is elsewhere (MS, Sony, Nintendo, supercomputers). Profits from Apple are steady, but slim. IBM failed performance and supply expectations more than once. IBM could have but did not commit as much of its CONSIDERABLE engineering resources to Apple’s wide range of needs even though APPLE ASKED FIRST. Apple was possibly not willing to pony up the cash and its own resources to BUY IBM’s considerable engineering resources. Intel added as much honey to the deal as possible (putting them equal to the almighty Dell? who knew?), including some things we STILL don’t know about. IBM is stubborn and stodgy and tries not to take risks (though it is better than it USED to be). Apple is stubborn and prissy and prone to tantrums.
Apple was possibly not willing to pony up the cash and its own resources to BUY IBM’s considerable engineering resources. Intel added as much honey to the deal as possible (putting them equal to the almighty Dell? who knew?), including some things we STILL don’t know about.
I think that pretty much covers it, the reason from reading IBM and Apple’s statements is financial. IBM can give them the chips but Apple need to finance the development, Intel on the other hand probably paid Apple to come to them.
ok.. so where’s that powerbook g5?
The problem is the NorthBridge, the CPU is already cool enough.
why’s the powermac water cooled?
Look up “power density”.
where’s the 3.0ghz by summer 2004?
Same place as the 4GHz P4 and the 3GHz AMD64.
Even IBM cannot break the laws of physics.
Watch the keynote carefully, the reason was *not* performance.
Why can’t Apple use two different types of processors?
If you had a PPC and a x86. One to run old code one to run the new stuff? Then the transition would be much cleaner and one would get the benefit of each CPU. And as we go more x86 then you could pop in a Cell instead of a G5 and then you have a vector CPU and a normal CPU. I think that if you were to use them together then that could be really good.
I think that the whole idea of computer architecture seems to be a saccade thing that people don’t want to add something new. Add a vector math unit on the board that developers can choose to use. Just like apple has made all of the Core technology (Core image, Core audio, Core video, Core data……..) They could make a Core Vector which would give developers the ability to use the vector processor.
I am sure that one could use it in normal computing if the Vector processor isn’t being used.
This would also give a performance boost to Apple computers over there counterparts. and guaranty that OS X to stay on Apple computers (that is what apple wants).
If you want Apple to choose a non-x86 processor, why choose the Cell? It only sounds good on paper because it’s a PPC instruction set. For the future roadmap it brings Apple, they might as well have chosen the MIPS chip that’s used in PS2. Or Ultrasparc.
Guys, whether or not Cell is good on paper, it shows little knowledge of the processor market to recommend Apple use it. The reason Apple went with X86 is scale of supply and resources being thrown at it for desktop applications. They won’t have to worry again in 3 years when Cell is completely outdated. X86 has a future today, it has a future 5 years from now, it probably has a future 10-15 years from now and likely beyond. Not even Intel themselves could kill x86, and they badly wanted to with Itanic.
I bet people wouldn’t be complaining so much if Apple originally declared they were going with AMD. (Which they actually CAN if they switch to x86).
If you listened to the software companies reaction (Adobe, Quark, etc…), they were the pusher. They did not like developing on two different platforms.
Finally – someone speaking some sense. If you look at the software market, Window software always comes first. It is extremely optimized for the x86, using features like SSE/SSE2. Then the company has to find someone who can not only port it to Mac OS, but can optimize the software all over again for PPC and AltiVec. Moving the Mac to x86 cuts the biggest part of that port – optimizing for PPC/AltiVec. They can keep the x86 specific optimizations unchanged for OSX86 programs.
Where is their leverage against Intel? If OSX-86 starts to make a serious inroad on Windows, Microsoft and their faithful, will put pressure on Intel that Apple can’t counteract.
Apple should have brokered a deal with IBM and AMD for a new hardware platform instead of pitching in with Intel. If they were too tiny to affect IBM, they won’t be any happier with Intel.
> Window software always comes first. It is extremely optimized for the x86,
> using features like SSE/SSE2.
This simple isn’t true. About 99.99% if not even more of the software out there is not optimized at all. People take what the compiler spits at them. That’s all about it. If you optimize you do it on the algorithmic level most of the time. The few very few areas where micro optimizations matter can and are already handled and don’t even surface as a blip on the radar.
The PPE is binary compatible with existing PPC cores, changes are needed in the OS but getting OS X running would be relatively easy.
True. But performance would be terrible, unless applications are recompiled and scheduled for the in-order PPE. And even then, performance wouldn’t compare well with the G5.
Why Apple doesn’t want to use IBM ready to go Power4, Power4+ or Power 5 chips instead of asking IBM for a custom made G4, G5?
Power5 chip is an excellent workmanship piece. You can order it in 2-ways up to 64-ways configuration.
I kept reading the article looking for the phrase “except a lack of actual order numbers to make it worth our effort.”
Dissapointed not to find it. Intel sells more x86 than IBM did Power… Hell, AMD sells more x86 than IBM did power… Hell, I wouldn’t be surprised if you added the Cyrix M2’s and newer VIA M3 chips together you came up with more sales than actual PPC chips sold…
Damnation, I’d not be surprised if more 486 chips were sold than powerpc chips (all models) were made.
This simple isn’t true. About 99.99% if not even more of the software out there is not optimized at all. People take what the compiler spits at them.
You pulled that figure out of your butt, that’s for sure. Every open source program I’ve looked over has optimizations for MMX, 3DNow, SSE, and in a number of cases, AltiVec. Although you can’t check the code of closed source programs, if you look at the feature lists and such, they do usually claim to have optimized the program for MMX and SSE. In fact, many programs REQUIRE MMX and/or SSE because of the optimizations. That will be in the minimum required specs list.
But performance would be terrible, unless applications are recompiled and scheduled for the in-order PPE. And even then, performance wouldn’t compare well with the G5.
The performance would be far better than in Rosetta on an x86, that’s for certain. Also, companies would simply recompile and put up an update for registered customers. That’s easier than changing the program to work on x86.
No, there’s something else going on that people aren’t privy to (yet). Eventually, some employee will leak the real reason Apple went to Intel.
The performance would be far better than in Rosetta on an x86, that’s for certain
I wouldn’t be so sure about that. Existing PPC software might well run faster on Rosetta than on the Cell. Remember it does binary translation, not emulation, so it will take full advantage of out-of-order execution.
Also, companies would simply recompile and put up an update for registered customers.
Perhaps, perhaps not. In any case they can just as well put up x86 versions. And they will definitely run faster than even the recompiled Cell versions.
That’s easier than changing the program to work on x86.
Only if you’re using hand-written assembler, in particular Altivec. But even then you have to reschedule it for the in-order PPE if you want good performance. That’s not all that much easier than translating it to out-of-order x86 code.
No, there’s something else going on that people aren’t privy to (yet). Eventually, some employee will leak the real reason Apple went to Intel.
If you want to start a conspiracy theory, at least put up one or two suspected reasons. It’s a bit boring otherwise
.