“There have been a few recent PowerPC announcements that have caused the Mac Faithful to wonder anew about The Switch and the “real” reasons behind it. First, there was the 970FX announcement, which clearly showed that IBM is capable of putting out a 970 processor that compares quite well with the Pentium M in performance/watt. And then there’s the 970MP, which Apple has used to make a monster of a quad-processor 64-bit RISC workstation. To make matters even more interesting, P.A. Semi has just announced a dual-core 64-bit PowerPC processor SoC that, if the specs and numbers are to be believed, could take PowerPC to a whole new performance/watt level.”
I never agreed with switching to Intel – and I still don’t but I don’t think Steve is willing to say oops we made a mistake (especially if they have to break a contract with Intel) although now would be the time.
First I thought the switch made sense.
But then came the 970FX. Then the 970MP.
After the IBM announcement I started to question it all. Apparantly, IBM was capable of levereging the 970 platform beyond new heights. Other than that, Freescale is working on multi-core G4s too.
And now PA Semi.
It all contradicts Steve Jobs’ reasoning.
I thought the whole idea to switch was due to Intel selling the chips cheaper than IBM.
That would make it a simple question of economy. But apparently there is a lot more to it?
>”I thought the whole idea to switch was due to Intel selling the chips cheaper than IBM. ”
Actually the chips are the same price.
It costs less to produce PPC chips than it does x86. x86 is able to be produced at a similar price to PPC because of volume.
>”But apparently there is a lot more to it?”
I don’t know why its so hard for people to accept the fact that it was due to future road maps of the two companies.
I don’t know why its so hard for people to accept the fact that it was due to future road maps of the two companies.
Because IBM has been meeting their roadmaps, and Intel has not. Intel is currently a joke when speaking of roadmaps. They threw out yet another roadmap just this week.
If the switch is about roadmaps, Apple is in serious trouble. Intel hasn’t met a roadmap in more than a year, and isn’t expected to meet one for at least two more.
Care to back this up with a respectable source? I’m sure Intel’s costs are high, now, simply because of the amount of cache they put on; but AMD’s costs should be significantly lower.
In the days of Eugenia this site was interesting. One could come here for news on technology, despite all the faults in its management, which could almost always be forgiven.
In just a few months, however, it degraded to useless. Every third ‘story’ is about Why Did Apple Really Switch, repeating the same tired and uninformed arguments on PPC’s ‘promising future’. What’s more, the site’s crew, or at least one person there, apparently enjoys spending his time on the comments section, boringly defending the ‘articles’ and why they are so right.
Meanwhile, even the automatically generated http://news.google.com section on technology covers interesting news that never find their way here.
Good point. It’s important to look at which news editor is cherry picking these stories.
When you don’t have the commmunity dictating news links… you end up with a certain individual pushing a certain issue in an effort to get others to adopt his way of thinking.
Man, I’ll second that emotion. I hated to see Eugenia go. I’m sure she has her reasons, aside from the frequent flames she drew.
You forgot all the Ubuntu reviews 😉
Seriously, this is an interesting story, though I think it’s wrong and the tone is unnecessarily inflammatory. The author, Hannibal, is a man who knows a lot about microprocessor designer. Look over the CPU reviews in ArsTechnica that he wrote, they’re pretty impressive.
Personally, I think the real reason for the switch is twofold
A Guarantee of Competitiveness
By alying itself to the dominant ISA, which has two aggressive rivals producing chips for it, Apple no longer has to be worried about being stuck in the CPU doldrums the way they were with the G3 and currently are in the laptop market with the G4. This P.A. Semi thing is due late 2006, early 2007: there is no way Apple can let the Powerbook and iBook ranges limp through another 12-15 months using 1.4GHz G4s; all the screen updates in the world won’t help there (and it’s worth remembering that a big problem with the G4 is not just its clockspeed, but also its really slow FSB).
A Cheap Platform
IBM sells CPUs to Apple at a price IBM can dictate, due to the slight impact the G5 has on its bottom line. Intel sells CPUs as cheap as possible because AMD is constantly harrying them. But that’s not the whole story.
The last year Intel has being promoting its Centrino platform: a low-cost, all-in-one solution of low-power CPU and motherboard with integrated sound, wi-fi, bluetooth and optionally even graphics. This is a great solution for the Apple, which builds laptops and other small-factor PCs based on laptop parts. People should remember that Apple is not totally oblivious to price: they realise there is only so far they can get with style and marketing, before they have to justify the actual computing power.
As for the digital workplace Hannibal talks about, I’m sure Apple has its eye on it, but if you saw the “One More Thing…” announcement a few weeks ago, you saw Steve Jobs put the iMac right at the centre of it. Mac is a part of where Apple is trying to go, no doubt about it.
Will Apple save money (meaning make more money) by switching to Intel? If so then that is your reason.
It is funny how people think that companies care about anything, but money..
Actually, I do think Jobs cares about things other than money (and when you talk about what Apple “cares about”, you may as well be talking about what Jobs cares about).
I’m sure it’s worth something to Jobs to be head of a company that makes quality products and cool products. Now, I’m not claiming that’s because he’s all benevolent and cares about you, but at the very least he cares about his own ego, and wants to be able to look Bill Gates in the eyes and think, “I make better products than you.”
Being an egomaniac and control freak sounds bad, but Apple customers benefit anyhow.
“The Switch”, above all things, is not about performance, or even performance/watt. It’s about peace of mind for Steve J. It’s largely irrelevent whether competitive PowerPC CPUs exist. Apple on x86 makes one less extraneous thing for Apple to worry about. Let’s face it: Apple is good at making cases, applications, and gadgets. They’re not so good at making CPUs and chipsets. What point is there for them to be in that particular business? Apple on x86 means that CPU performance is removed as a bottleneck for Apple’s success. It let’s them get back to their core business.
It doesn’t matter if the new CPUs are faster or not (though, history suggests they will be, at least most of the time over the long haul). Nobody can criticize Apple of shipping slow machines when Dell is shipping the same CPUs. That’s the biggest thing Apple gains from moving to x86.
>”They’re not so good at making CPUs and chipsets.”
Its not that they’re not good at it. Its that they don’t do it.
Yes they do. The crap-tacular G5 chipset was an Apple design, fabbed by IBM.
Actually, it was co-designed by the two.
If memory serves, Apple played a small role in the project, but it was primarily designed by IBM.
and crap-tacular? ok… now you’re trolling. Its so craptacular that Microsoft, Sony and Nintendo adopted it for their next generation gaming consoles.
The processor thus far has proven itself many times over. The reason for the switch was about how PPC and Intel’s future roadmaps compared… not by their current or past comparisons. Thus far, IBM/PPC has either maintained at least an equal or greater performance advantage to x86.
Actually the nintendo gamecube already uses a tailored PPC called “Gecko”
1) I’m talking about the G5 chipset, not the G5 itself. You know, the Northbridge. That was designed by Apple and fabbed by IBM, and was “craptacular” because it had attrociously high memory latency.
2) Microsoft, Sony, and Nintendo have had nothing to do with the G5. The PPE/Cell technology is an entirely different chip, by an entirely different design team at IBM.
Apple’s role in that was small. IBM played the senior role in the chips’s development. They didn’t just fabricate it as you suggest.
Regardless, you’re craptacular comment is way off base.
Y’all apparently type faster than you can read. My original post said:
crap-tacular G5 chipset
chipset not chip. The comment was about the core-logic chipset, the northbridge connecting the CPU to the DRAM and AGP/PCI busses, not the G5 itself. It was indeed craptacular. Anandtech’s lmbench results showed it to have a memory latency of 300ns, twice as high as that on the Nacona Xeon, even though the latter was also using an off-chip memory controller, as well as high-latency DDR2-SDRAM!
Just to make it perfectly clear, Hannibal’s interview with an IBM engineer confirms that the G5 chipset was Apple’s doing, not their own:
Peter Sandon: The chipset is an Apple design, and we’re fabbing it.
Clearly, ASIC design is not Apple’s strength. Therefore, it makes all the sense in the world for them to let Intel, which is very competent at it, handle that kind of work.
>”The chipset is an Apple design, and we’re fabbing it.”
I have read countless interviews by IBM execs that reitered that it was co-developed.
>”Clearly, ASIC design is not Apple’s strength.”[/i]
Clearly nothing… The G5 has done very well for Apple thus far.
>” it makes all the sense in the world for them to let Intel, which is very competent at it, handle that kind of work.”
Apple’s PPC macs have repeatedly gone toe to toe or excelled beyoned x86 chips. The transition is about future roadmaps and Windows compatibility…. not performace.
Even if they take a hit in performace as compared to PPC after moving to x86, they wont be any worse off than any other computer manufacturer but will have full compatibility while also retaining OS X’s advantages.
is a 150 employee startup – announcing things is one thing – shipping another – they say that first samples will ship by the end of 2006. Mobile Macs with Intel will be out for a while then.
And regarding the PPC970 – the IBM CPU is far from beeing ready for notebooks – especialy for Apples ultra thin Powerbooks. The powersupply build in the Quad G5 has a maximum powerconsumption of 1000W (!). You can only use the PPC970 in low power environments if you reduce the CPU clock dramatically (1.3GHz and below) and I bet the CPUs Intel ships next year will give more performance than such a low clocked G5 chip.
And beside this – it is all about roadmap (like Steve Jobs said) and who knows what IBM is able (or willing?) to deliver in the next years.
And regarding the PPC970 – the IBM CPU is far from beeing ready for notebooks – especialy for Apples ultra thin Powerbooks. The powersupply build in the Quad G5 has a maximum powerconsumption of 1000W (!).
It can supply a maximum of 1000W, which is something slightly different. The G5’s PSU is rather overdesigned. The single-CPU dual-cores have a 600W power supply. There is no way you need that much juice when the 970MP takes up like 100W.
You can only use the PPC970 in low power environments if you reduce the CPU clock dramatically (1.3GHz and below) and I bet the CPUs Intel ships next year will give more performance than such a low clocked G5 chip.
Intel shipped CPUs last year with better performance than such a low-clocked G5.
Indeed I agree, what companies say and what they do they do are entirely different things.
IMPO I think that a decision who affect all the Apple base of users couldn’t had been taken in the short time. Today this micro is faster (dissipate better), in a year this other manufactured present this other better…
In any case they could not rectify or the low capacity of theirs officer would be more clearly visible. So I think that they will continue defending the change in a stubborn position.
I think this case wont be over even in 2012 when people will look a hardware IBM and write : haha, I told you, steve Jobs has an hidden agenda and blabla.
Everything here was suppose to be delevered some time ago, and wasn’t. Now who cares if that it can be in the futur?
The episode with the G4 would have be the last for me. IBM is playing with Apple, and now to save the face they are trying to discredit Steve Jobs and Apple business.
It’s simply a game.
Well, maybe I should take time to write a generic message or stop reading that kind of story, because it will happen each time something new about the PPC get released.
Ah, frak that. This chip is not just a new CPU. Its a next gen bandwidth monster. It’d be a waste on a desktop. Its a server chip through and through with an entirely new class of bandwidth the world has never seen.
http://episteme.arstechnica.com/groupee/forums/a/tpc/f/77909774/m/6… for half assing the i/o wars. its definately a rant, i was a little, err, extra incap’d that night. whatever, these guys are getting it right though.
this isnt a apple/intel deal. this is a amd/pa semi deal.
a+! <chant>i/o i/o i/o!</chant>
> It’d be a waste on a desktop.
I disagree, people don’t like fan’s noise, so the small power consumption would be a hit if they managed to produce them cheap enough, which is unlikely..
> Its a server chip through and through
Well it is targeted by its maker to the embedded market, it would probably fare well as a server also but the unknown brand is be a problem for resellers..
I wish them well, but the price and the time to market will be essential for their success: it wouldn’t be the first time that a technically superior part is a failure due to high price or late delivery (or both).
Hope to see low power desktops with these chips. Maybe a mini-itx competitor to via !!!
With which OS??
I doubt that Apple would want to cause trouble in its migration to x86 by sending ‘mixed message’: the software vendors would never forgive them..
This is slightly OT, but I’ve been reading the message boards regarding “the switch”, and I think its hilarious that Mac users everywhere are running to the defense of the Pentium-M. In the late 1990s, they were being snotty about how the G4 was “faster than the fastest PII”. The PII and the P-M are almost the same damn chip! If you took a PII from 1998, shrunk it does, added a modern memory interface, and ran it at 2GHz — it would be a Dothan! Is there any reason the Mac faithful are even allowed to talk about processors?
The Pentium M is a derivative of the Pentium 3, but it’s a bit of an overstatement to state that they are basically the same. There are improvements in branch prediction, improvements in the decoder with the micro-ops fusion, and the addition of a dedicated stack operation unit, plus SpeedStep. In addition, compared to the P2 the Dothan has support for SSE and SSE2 and a longer pipeline. If you took a P2 with a P4 bus interface and added more cache and clocked it at 2GHz it wouldn’t be Dothan.
Improved future derivative architectures are essentially inappropriate for retroactive arguments. That is, disregarding the truth of any argument that the PPC7400 was superior to the Pentium 2, that the Pentium M (as a derivative of the the P6 architecture as well) is later then considered good has no effect on the past.
I think the switch has to do with Macs being able to run other operating systems via virtualization software at near native speeds as well as an eventual (5-10 years down the road) deal with someone like Dell to start selling Macs.
Looking at the last PowerBook update make obvious why the switch make sense. The machines are wonderfull except one component: the processor speed is clearly behind the PC market. Everybody was expecting a bump in processor speed: the improved G4s were suppose to be there 9 month ago! Did not hapen: Freescale probably could not deliver.
Intel can deliver!
History has shown that PPC is a losing battle. It’s a great design, but nobody can match the money Intel and AMD can throw at x86.
PPC is really fast when released, but goes way too long between revisions. Most of the time PPC isn’t very fast or very cool.
The smart money is still on Intel based on the history of PPC. One or two advances don’t mean anything in the long run.
“History has shown that PPC is a losing battle.”
Its not that PPC is a losing battle… but competing with a company has tied their OS and illegal monopoly around that processor is a losing battle. If anything, PPC has been a shining star in the processor industry.
“It’s a great design, but nobody can match the money Intel and AMD can throw at x86. “
Intel and AMD have to throw more money at the processor because its a 30 year old chip technology. It takes a lot to keep it competative. PPC has at least equaled or exceeded the performace of x86 for less money. AMD/Intel are only able to match the PPC price because of volume. The performace argument is somewhat relative but supporting price argument is a losing battle.
“The smart money is still on Intel based on the history of PPC.”
Apparently so, if Apple saw the future road maps of both and saw that x86 was finally going to have a performace advantage in the future
If anything, PPC has been a shining star in the processor industry.
In what alternate reality???
Intel and AMD have to throw more money at the processor because its a 30 year old chip technology.
Why do people have this illusion that the ISA == the chip? The fact that the Opteron uses an ISA based on a 30 year-old doesn’t do a lot to constraint the design of the CPU itself. x86 is little more than a particular compact if somewhat baroque intermediate language for a modern processor. It adds a couple of decode stages to the pipeline, and a few million transistors here and there. The vast majority of the chip is new technology.
It takes a lot to keep it competative. PPC has at least equaled or exceeded the performace of x86 for less money
Again, in what alternate reality?
AMD/Intel are only able to match the PPC price because of volume.
I have a very hard time believing that the PPC is that much cheaper to fab than a comparable x86 chip. The Opteron dual-core at 90mm has only about a 25% larger die than the 970MP at 90mm, while featuring 32KB more L1 cache, bigger TLBs, an integrated memory controller, and 3 hypertransport links. The facts simply don’t back up your claims.
>”In what alternate reality???”
[roll eyes]
>”Why do people have this illusion that the ISA == the chip? The fact that the Opteron uses an ISA based on a 30 year-old doesn’t do a lot to constraint the design of the CPU itself.”
Why do so many people think that increased financial resources automatically translates to increased performance?
>”Again, in what alternate reality?”
Something tells me you’ve been reading too many PC fanboy sites that go aout of their way to help PC users justify their computing choice.
>”I have a very hard time believing that the PPC is that much cheaper to fab than a comparable x86 chip.”[/i]
Much of it has to do with the physical size. The smaller the chip, the more you can reap from each waffer
“The facts simply don’t back up your claims.”
This has been a significant talking point for Mac users for many years as it has been repeatedly reiterated by IBM and Moto. The fact that you don’t know of it, confirms for me that you’ve only been reading the afformentioned PC news sites that do little more than help PC users justify their computing choice.
[roll eyes]
Seriously. The benchmarks would suggest that the G5 “doesn’t fall too far behind”, not that it “equals or beats x86”.
Why do so many people think that increased financial resources automatically translates to increased performance?
I have no such illusions. However, like most people, I believe benchmarks are indicators of increased performance. The benchmarks show the Opteron kicking ass and taking names. The new SPEC figures for the 970MP are indeed quite respectable, but its integer performance is still in 2.0GHz Opteron territory, and if the history of the 970 is anything to go by, these 2.5GHz MPs will likely not get above 2.7GHz before the Intel PowerMac refresh. The 2.0GHz G5 was also a good performer for its time, but that was 2003, when the 1.8GHz Opteron was brand spanking new.
Something tells me you’ve been reading too many PC fanboy sites that go aout of their way to help PC users justify their computing choice.
Oh gag me. Yes, spec.org is such a “fanboy” site. Seriously, I’m a Linux user. What do I care what CPU is under the hood? Ubuntu looks the same on all of them.
Much of it has to do with the physical size. The smaller the chip, the more you can reap from each waffer
And the G5 is only marginally smaller than a comparable 90nm Athlon64.
This has been a significant talking point for Mac users for many years as it has been repeatedly reiterated by IBM and Moto.
You’re not serious, are you? Kind of like how “the G4 is faster than the fastest PII” was a “significant talking point for Mac users”? Please! As if “talking points” have jack-shit to do with reality.
The fact that you don’t know of it, confirms for me that you’ve only been reading the afformentioned PC news sites that do little more than help PC users justify their computing choice.
My 970MP PowerMac is arriving tomorrow. Pyschoanalyze me now Dr. Lecter…
Something tells me you’ve been reading too many PC fanboy sites that go aout of their way to help PC users justify their computing choice.
No justification needed. For years we were saying “x86 is the future” and Apple’s choice only confirms what most of us have already known.
Think about this for a moment, you are on an internet forum debating about a processor that Apple has decided to replace.
Who is trying to justify something here ?
This has been a significant talking point for Mac users for many years as it has been repeatedly reiterated by IBM and Moto. The fact that you don’t know of it, confirms for me that you’ve only been reading the afformentioned PC news sites that do little more than help PC users justify their computing choice.
PPC is close to dead on the desktop and apple’s switch confirms it. In a couple of years it will be nothing more than a memory for fanboys to cry about ‘remember how GREAT it was! waaa waaa waaa!
The only person trying to justify anything is you my friend.
Not only are you completely divorced from reality, you or your ilk are moderating down posts that are not off-topic and are not ‘trolls.’
Instead of baselessly claiming that the 970FX or 970MP are superior in terms of performance than unnamed x86 processors, why don’t you provide benchmarks that are not from Apple’s website?
Oh and just to prevent any PeE SeE fanboy comments, I have to suffix this post with “Among the many computers I own, I have a G5 iMac.”
I don’t see why everyone makes a big deal over the “switch” to Intel. The entire line will not be converted overnight and, with Universal Binaries and Rosetta, users won’t particularly care what CPU they are running.
I see the “switch” to Intel as just a way for Apple to gain flexibility. Now they have the freedom to build new machines using whatever CPU is most attrative at a given time — be it PowerPC or x86.
If the new PowerPC chips from IBM or this startup materialize with superior characteristics, Apple will use them, there’s no reason not to.
I really can’t friggin’ believe these stories with multiple links in them. How am I supposed to know which is the original article? I may or may not choose to click every link in an article so why are you torturing me with this guessing-game crap, Thom?
http://www.advogato.org/article/860.html
He’s damaged his and arstechnica’s good reputations with that illogical and ill-tempered article.
While Hannibal’s technical expertise is unparalleled in the journalistic profession, his opinion columns have repeatedly taken a slant against Apple in recent history. Its very disapointing because the attitude he takes is one of a person who took a bold stance on the platform wars years ago, has made several bold proclimations along the way and now that many of his predictions weren’t even close, he has become very bitter.
All too often people defend his articles because of his past exelence in being ballance and well-thought-out with regard to technology, but its surprising that they’re so quick to look past his obvious agenda.
“He’s damaged his and arstechnica’s good reputations”
Haha, good one.
I don’t know if that was meant to be a slight against Hannibal or Ars, but if it is, I’d like to point out that Ars is pretty much a beacon of hope for internet journalism (or journalism in general). Unlike the vast majority of places today (both in the “real” media and the online media), the people at Ars generally (gasp) know what the heck they are talking about.
I’m not particularly interested in whether Hannibal has a particular agenda. Pretty much everyone does these days, and most people are smart enough to read between the lines. What matters is that these guys have good information, which is the ultimately what you look for in a technology site.
Believe it or not, there are business situations where boredom is highly desired: one of those is price, another is performance, and then there’s delivery.
What does it matter if you have a great price and great performance but have an unpredictable delivery? (refer to the many manufacturing delays and shortages Apple has had to deal with in regards to their processor suppliers)
What good does it do to have great performance and great delivery but too high of a price? (fill this in as you wish, I’m sure you can find something)
What good does it do to have substandard performance, great price, and reliable delivery? (think Apple G4’s of late: I don’t recall them having much of a supply issue, but they’re a bit underwhelming for today’s expectations)
There are also the added issues of engineering time and effort for making customized chipsets for a much more limited market (Macs haven’t had nearly the economy of scale to play with to entice others to voluntarily do their engineering for them) to support the basic hardware.
It’s much cheaper to be a basic integrator of existing hardware components, with a minimal of tooling overhead, and is much faster for making changes, too: look at Dell.
By going the x86 route, they have all the economies of scale to play with, that puts them on an equal hardware engineering and manufacturing footing of all their other competitors, at least when it comes to the stuff under the case. This allows them to specialize on the form factor (where appropriate) and the exterior appearances, as well as creating a whole integrated package that combines software, etc. that truly delineates them from their competitors: basically, they’ll be improving their boredom in the beneficial way businesses like (reliability in expectations and logistics) while still being able to innovate in the things that people really care about, the total user experience.
So, in the final analysis, the x86 switch by Apple can be reduced down to the desire to have boredom when it comes to logistics as mentioned above, since at least if they have a problem with getting their needed hardware (other than any cases where they have special cases made that aren’t available by the typical PC makers’ vendors) then all of their competitors are likely to suffer equally.
So, in reality, it means zero that this new company has promises of providing souped-up PPC processors, and that IBM can make the processors Apple wants now, if it takes too long and costs too much for Apple with their limited scale to afford to pay IBM to do so, and if IBM doesn’t deliver what they promised for performance at a time they promised, with reliable production that Apple can buy. IBM, Motorola/FreeScale and this other company just provide too much excitement for stockholders and the management.
Really, isn’t boredom sometimes a great thing?
“if it takes too long and costs too much for Apple with their limited scale to afford to pay IBM to do so”
Price has little to do with it. The chips have been equally priced to x86 chips.
Apparently, I must have forgotten to add in the factor that “current chip price Apple is willing to pay” combined with “expected number of chips that will be developed and made in the future by IBM or partners” are not the same animal.
While the current chips are priced competitively for either one to be dropped onto a motherboard now, the price Apple would have to pay to convince IBM that it’s worth their while to go through the effort to design and manufacture the relatively few (think about economies of scale and how expensive it is to design and manufacture microprocessors, especially compared to their anticipated useful resale life as non-embedded processors: very short cycle!) chips they’ll buy, and it comes down to boredom, once again, and it works for both Apple AND IBM wanting boredom now and in the future.
The x86 chips have a much larger pool of buyers in terms of scale to spread the investment in the manufacturing lines and development of the next versions of the chips compared to advancing the PPC processors, so it’s no big deal to keep pushing things forward, at least currently.
Real benchmarks (not microbenchmarks hand-chosen by advocates) have never shown any compelling advantage to PowerPC or 970 derivatives. They’re at best roughly equal to the equivalent x86 (usually at launch time) and tend to fall behind over time until the next launch.
There’s always an miraculous G(N+1) or 970Z right around the corner with amazing performance that will blow away x86 performance with some combination of { instruction-level parallelism, giant caches, multiple cores, … } but somehow, this never seems to happen. This has been true for years, and not just for Apple: the economics of trying to beat x86 on the desktop have essentially destroyed most of the other RISC competitors or driven them into niche markets (a fate also shared by Intel’s own Itanium).
Why should Apple – which innovates in a completely different area – tie its fate to this miracle finally coming true?
>”[PPC is] at best roughly equal to the equivalent x86 (usually at launch time) and tend to fall behind over time until the next launch. ”
You have that a bit twisted. At launch, the speeds are usually significantly faster, sometime in the middle, x86 and PPC level out and towards the end x86 gets a performace advantage… then PPC gets updated again and the process starts all over.
Jobs sees something in performance down the road that indicates that Intel will best PPC consistantly.
If by “middle” you mean “beginning”, then yeah, it’s somewhere in the middle. The G5 hasn’t been cutting edge since 2004. The dual cores are again competitive, but only because all the existing dual-core CPUs are also clocked rather low.
When the G5 launched, the Opteron already had superior SPEC scores. With the PPC 7400, AltiVec was introduced and provided better performance in vector computations than Katmai’s SSE implementation, but general computation was never significantly faster. If you want we can travel back in time to 1997 for the PPC 750 vs. Klamath or its derivatives–especially the Mendocino Celeron. The thing is, and this was especially true of the PPC 7400, that Apple ends up riding a processor line too long until the prices for some of its machines are completely unjustifiable in terms of performance. This is quite possibly a means of avoiding this problem in the future.
If you inappropriately moderate my comments or the comments of any other members in the future with your shill accounts, I will specifically report you. You’ve tested my patience with your intellectual dishonesty as much as I will tolerate. Make no mistake that when they purged your half-dozen kellyN accounts that was a clear message to stop abusing their forum.
I don’t care if Apple puts a potato chip in my Mac, so long as its fast enough. I’m not a PPC assembly language coder, so why should I care? I can’t imagine a reason why any other sensible person would care, either.
Remember, the probability that IBM is incapable of making a fast, cool and cheap PPC is much, much less that the probabiity that IBM wasn’t going to give Apple a better deal than they could get with Intel.
Besides, would you want to bet your company on the likelihood that IBM — a very, very big company — would continue to make PPC chips — a small contributor it its revenue — in perpetuity?
“would you want to bet your company on the likelihood that IBM — a very, very big company — would continue to make PPC chips — a small contributor it its revenue — in perpetuity?”
You gues continually amaze me. You actually think that PPC is something that IBM develoeps exclusively for Apple.
Lets be honest 90% of home users never use their cpu to the full anyway, generally its professionals that crave the power.
Also I have learnt over the years not to second guess apple too much, especially when jobs is about. For all we know they want to licence the os to others, or utalise it in a differnt KIND of device.
buisness decisions like this are not made overnight, infact it had its foundations laid when the OSX was first written.. circa 5 YEARS ago
I’m not anti or pro this switch, but I just cant understand why people are so upset about it.
After all I dont think ppc is that great (its not bad either) that mac useres keep complaining.
Macs will be Macs nothing will change (for the useres)
“Im talking about the G5 chipset, not the G5 itself. You know, the Northbridge. That was designed by Apple and fabbed by IBM, and was “craptacular” because it had attrociously high memory latency. ”
“chipset not chip. The comment was about the core-logic chipset, the northbridge connecting the CPU to the DRAM and AGP/PCI busses, not the G5 itself. It was indeed craptacular. Anandtech’s lmbench results showed it to have a memory latency of 300ns, twice as high as that on the Nacona Xeon, even though the latter was also using an off-chip memory controller, as well as high-latency DDR2-SDRAM!”
I dont understand really where you got your information from, you told about Abandtech, i would not believe them, i mean i dont believe people that are supposed to make tests but they dont know what is the difference between threads and process. You told about lmbench, but i never had such memory latencies with a powermac G5 by profiling the performances with Monster (a CHUD tool) which gives you a complete analysis of the processor and controleur system (northbridge) performances. So please dont only believe test that are done by people who are not really professional, i would suggets you to use Monster and check by yourself.
And even if the Apple ASIC has a higher latency than the Naconna, well it still manages to coumminicate at 21,6 GBps with two procerssors, the Nacona, tell me ……… communicates at 6.4 GBps with two xeons. Apple ASIC has 3.4 times more bandwidth than the Naconna platform. And Apple ASiC use Hypertransport to communicate with the rest of the system at 4.8 GBps, something that Naconna surely does not do. More over Apple ASIC is a point to point architecture which gives a dedicated bandwidth to each request from somewhere else in the systeme. Naconna is not a point point ASIC. So tell us now which one is “craptacular” ?
So instead of trolling and speaking on only one component of the system, you should first look at the system overall, and try to understand it.
But i dont want to compare both systems, Intel will manage to get a better ASIC with the help of apple (i dont think that apple is commited to use their old designed Xeon northbridge that Intel use now for their dual core systems. Can you imagine 4 cores sharing 6.4 GBps to access the northbridge, thats a huge bottleneck and it even worse when you think that two core per processor do not comminicate directly), i guess Intel has a lot to learn from Apple in this regard.
I believe that Intel will be able to propose great platform next year, so the point of view of Arstechnica is somehow strange. They are going to a story that apple do not want to design great computers, they are more commited to the iPod, ….bla, bla…. That’s very wrong. The guy who wrote it forget that Mac OS X, and the huge work that apple did so far on this system. Apple does not build osx for the ipod, but for the mac. Apple does not build so much great pro apps for the ipod, they do it for the mac platform. So apple is more than ever comitted to the mac, regardless of the processor.
Arstechnica like to tell about the G5 dual core, the low power powerpc 970fx, but should we remind them that IBM had big production problems, the G5 dual core has been late for months. This is somethiong that apple cant continue to deal with. They need someone to provide stong processors production, something that Motorola and IBM failed to propose. The 970fx, yes its power efficient, but it runs only at 1.6 max, that’s desapointing considering that intel runs their pentium M at 2Ghz+ with roughly the same power consumption. And where is the G5 at 3 ghz that IBM promised?
IBM is now providing a low power G5, but intel is ready to propose dual core processors for laptops, ther is no way to put a dual core G5 in a powerbook, i guaranty you, and so far IBM has told anything about to develop a low power dual core G5. Considering the small success that they had with the 970fx in this task, i doubt that IBM has anything to propose against intel. And what about the 65 nm process at IBM?
Yes there is this pa semi processor, its power based, it is very low power. But thats not gonna to be available before the end of 2006 or even 2007 in production, so thats not good for apple. Apple has to be competitive against the pcs, and having to wait until that time to get a good dual core chip for their laptop is too long. Intel is planning to introduce Yonah at the beginning of next year. And nothig tell us if pa semi can supply apple in processors, nothing tell us that they can have the production suitable to the demand of apple. IBM did not succeed, so i doubt that Pa semi can, even if they got their processors manufactured in several fabs. And more over they target with their new chip embeded market, and it design for this, something quite different to the desktop computer market.
The dual core G4. No way, that a processor for embeded market, it support too many technologies that are not spread in the computer world (the use of RapidIO for example) amd Freescale has delayed it for the end of 2006. And anyway we khow that Freescale has many problemes of production to satisfy apple, look they are still not able to provide in quantity 7448 G4s to apple. It could have brought some improvements to the powerbook line (bigger L2 cache, optimisation in Altivec, slightly faster bus), no!!!!! apple has still to be stuck with the G4 7447. So i relly would not bet on them to bring me some dual core chips.
I an the first lover of the powerpc, the G5 dual core is an awesome chip for servers and worstastation, thats a very powerful chip, i would say that it gives to high end x86 chips money to run. Powermac have a great architecture better than any Xeon bas machine and in match with Opteron based machines.. But IBM has too many production problems, their roadmap is sure not as agressive as the intel one at least in the chips for laptops. It seems that Intel is coming with nice chips in the midle of the next year, and i believe that apple will collaborate with intel a lot.
But now maybe the main point is that Apple has an OS that basicly runs everywhere. So thats Apple who has the possibility to choose whatever chips that they think its good at the end. Apple may first choose intel for the laptops and keep he power for high end performance, until Intel is ready. Thats not difficult, Xcode allows developper to manage both platforms very easilly, anyway they still have to suport the powerpc platform for a while, and for the users, well…..they wont notice what it is happening.
“And even if the Apple ASIC has a higher latency than the Naconna, well it still manages to coumminicate at 21,6 GBps with two procerssors, the Nacona, tell me ……… communicates at 6.4 GBps with two xeons. Apple ASIC has 3.4 times more bandwidth than the Naconna platform. And Apple ASiC use Hypertransport to communicate with the rest of the system at 4.8 GBps, something that Naconna surely does not do. More over Apple ASIC is a point to point architecture which gives a dedicated bandwidth to each request from somewhere else in the systeme. Naconna is not a point point ASIC. So tell us now which one is “craptacular” ?”
I don’t care what part of the system were talking about…It is significantly harder to improve latency! I can’t stress this enough.
First, I wouldn’t trust Ananadtech further than I could throw them, but I trust lmbench. It’s very mature code and has proven to be a reliable indicator of what it measures.
And even if the Apple ASIC has a higher latency than the Naconna, well it still manages to coumminicate at 21,6 GBps with two procerssors
That’s merely a function of the processor bus, which was an IBM designed component. Yeah, Apple managed to implement IBM’s processor bus correctly. Good for them. However, we don’t know how well they implemented it. You’re just citing theoretical peak bandwidths — we don’t know if the chipset can actually live up to that.
Apple ASIC has 3.4 times more bandwidth than the Naconna platform.
IBM’s CPU bus has 3.4 times more bandwidth than Nacona. The stuff Apple designed, the DDR controller, has twice the latency of Nacona.
More over Apple ASIC is a point to point architecture which gives a dedicated bandwidth to each request from somewhere else in the systeme.
Again, there is no way you can give Apple credit for making the ASIC point to point when it wasn’t their decision to do so! The PPC970 has a point to point processor bus. It’s FSB has no mechanisms for resolving bus contention. The only way to make a chipset for the PPC970 is to make it point to point. That doesn’t change the fact that the implementation of the parts Apple had control over sucked.
As for trying to understand the whole system — I do understand it. You spout out all this info about theoretical peaks, but fail to realize that those theoretical peaks are simply the result of the nature of the PPC970. The chipset itself, the parts Apple was responsible for, clearly has shortcomings. Let’s put this another way. The early VIA K8 chipsets were pretty impressive too. Point-to-point processor bus, etc. That doesn’t mean that they didn’t suck. VIA doesn’t get any credit for point-to-point processor busses — its mandated by the Athlon 64. They do get credit for the stuff that’s under their control, and blame when that stuff fails to be up to par.
The 970FX was announced as a perfectly capable laptop chip.
It’s still not available, and even on paper it hardly keeps up with the Pentium M, never mind Yohah and Merom.
The 970MP was announced as a solid, 64-bit RISC dual-core server/workstation chip.
Yes, solid, nothing more. Actually somewhat slow compared to AMD’s current offerings and what’s coming up from Intel.
Intel is taking a performance/watt beating everywhere except for the Pentium M.
What’s that got to do with anything? He’s said in his own articles that the IntelMac lineup will be based on Yonah and Merom.
Intel has just shaken up their server lineup for 2007.
The shakeup concerned Itanium, which is of no interest to Apple, and the 4-core Whitefield, which would be nice but is not critical to Apple with its rather limited server business.
PA Semi is set for a major performance/watt breakthrough in 2007 with a dual-core, 64-bit, Altivec-enabled processor that looks perfect for the Mac line.
And he would have expected Apple to rely on that, a small unproven company? Their integer performance numbers don’t look great even now, never mind in 2007.
Oh well, at least he didn’t mention Cell, that would really have destroyed his credibility.
Yes, Apple is taking a bit of a risk relying so much on Intel’s upcoming architecture, and a delay for Yonah or Merom would be very bad news for them. But presumably they’ve been given cast-iron guarantees by Intel (read: big damages in the unlikely case things do go wrong). And of course there’s always AMD as a backup.
P.A. Semi doesn’t have the ability to support a company as big as Apple at this point with the range in products Apple has now.
Intel does right now.
Apple switched to X86 because they always wanted to. It’s as simple as that. They probably thought x86 would eventually die and be replaced by powerpc or something when power first came out. Why they waited so long to go Intel is easily explained as well. Making all of the Apple developers move from powerpc to intel and Mac OSX at the same time would be asking too much, instead they waited five years for osX to stabilize and be adopted, then they simply changed the CPU architecture. It’s as simple as that. Apple always wanted to go X86. Noone forced them to.
If you say that
“I don’t care what part of the system were talking about…It is significantly harder to improve latency! I can’t stress this enough.”
it means you dont really understand how each element contibute to the performance of a system as a whole. And even your comment does not make so much sense. Sorry!!!
Ok, I’m going to say this once… Steve Jobs is not a hardware guy, he is a software guy. The Woz was hardware. If you noticed at the iMac presentation like one slide was about performance. Maybe you might have also noticed that CPU’s haven’t really got much faster. In fact, it’s not faster processors, it’s more processors. Meaning more software running at the same time. Apple doesn’t need to be spending money designing hardware chipsets and whatnot when they could have Intel do it for them cheaper and better than they could. Then Apple can focus on something that will make them more money, like cool new programs or gadgets. People don’t buy machines because they are 10% faster than what they bought 2 years ago. They buy new machines when they are 200% faster than their last machine. If machines don’t get much faster, then it’s going to be a software game. Apple is making great software and it doesn’t matter what machine Apple uses. Also, Intel has Trusted Computing, which Apple already uses. This will allow Apple to get HUGE movie distribution deals when the REAL Video iPod comes out. Call me crazy, but next summer when an Intel based Mac Mini comes out with a new iTunes MOVIE STORE capability and Front Row… Then the Intel move will make sense. Performance Per Watt in a small gadget is where Apple really wants to be at. The PC is dead, the Media Center Computer is just being born. Welcome to PC 2.0. Where do you want to go today?
fight on!
ppc rulez!
and buy an amigaone too
The reason probably was only marginally, if at all, about processor performance. It is much more likely to have been about the total costs of supply of the system.
If you look at this from the outside, the prices for PPC processors and comparable X86 ones are not too crazily different. But if you look at prices for PPC main boards and X86 main boards, its factors of 5-10. Then there is the driver problem, integrating them with the graphics cards and so on.
I know that, in defiance of the observable facts, there will be people who say that the price of comparable Mac and Intel based machines is the same, but it obviously isn’t, the Macs are around double the price, and this is why, and this is why they had to change. It had nothing to do with the processor itself or its performance, the processor only had to be good enough. It had to do with the total costs of the system.
So. if there is anyone who really wants to make an informed contribution, and has the industry knowledge to do so, what it would be nice to see is a real analysis of the full costs of the different alternatives.
This being OSNews in its present incarnation, I am sure someone will post saying that the two sorts of boards use the same chips so the costs must be the same. That isn’t the problem, component cost. The problem is the costs of the whole supply chain. It is not because some chips are so expensive that PPC main boards are so expensive. Its because of how many are made, where they are made, what for, how long the runs are…. This sort of stuff. There may be a componenet difference as well, but I doubt its the main driver for the differences we see in the market.
If you think this is wrong – just give a link where one can buy a PPC main board with the usual features, built in graphics, networking and so on, for under $100. I’ll buy it at once.
“I know that, in defiance of the observable facts, there will be people who say that the price of comparable Mac and Intel based machines is the same”
The numbers don’t lie. Mac hardware DOES cost the same if not less than comperably equipped hardware. When it doesn’t in some people’s comparisons… its ONLY when the PC is not matched equally or as close as possible in hardware, software and OS.
“Macs are around double the price”
Actually, more often than not, PCs end to be 30-50% more.
“It had nothing to do with the processor itself or its performance”
It was about future performace… not current perforamce. The G5 has either equalled the speeds of x86 chips or exceeded them. Jobs said he found benefits in x86’s future roadmap as it compared to PPC… In the future.
“the processor only had to be good enough.”
Agreed. It just had to be at least equal or slightly better. That’s “good enough” as opposed to being the necessary 50-100% better.
“It had to do with the total costs of the system.”
The price of PPCs cost no more than x96 chips.
There’s not too many fiercely burning light bulbs here, is there?
Strategy 101:
1) Announce the jump to Intel
2) Make IBM shift to high gear and deliver new chips
3) build your software for Intel machines
4) sell in both spaces
5) rake in the big bucks as droves of Wintel users make the smooth, easy, safe and wonderful transition to Mactel
It’s easy.
Did I say Apple is IBM’s only PPC customer? No.
I said the PPC is a minor component of IBM’s revenue stream. IBM needs to make PPC chips a lot less than Intel needs to make x86 chips.
A better queston is this Why should Apple stay with the PPC?
Semi’s chip’s a f–king POWER, not a PowerPC. And it’ll ship at the end of 2006, if ever. The new G5s are severely underdelivering (2.5 GHz max?).
But don’t let facts get into your way.
POWER and PowerPC are very close.
“PowerPC is largely based on IBM’s earlier POWER architecture, and retains a high level of compatibility with it; the architectures have remained close enough that the same programs and operating systems will run on both if some care is taken in preparation.”
http://en.wikipedia.org/wiki/Powerpc
http://en.wikipedia.org/wiki/IBM_POWER
The next-gen Amiga (admittedly a fringe piece of hardware) has been living with G3 and G4 CPUs for far too long. Maybe these uber-performing 64-bit PPC chips will finally convince someone to get cracking and design a 64-bit AmigaOne.
What I find so sad about that PPC vs x86 competition is that the x86 has a 35+ years old programming model, the PPC a 15+ years old programming model based on even older RISC ideas.
The impressives architectures of current CPUs ( superscalar, OOO scheduling, gigantic caches… ) shall not obliterate the fact that it is a very conservative way of computing, geared towards the effective compilation of low level languages like C.
– Where are the features for helping dynamic languages with dynamic dispatch in CPUs where the pipeline forbids effective indirect JSR/CALL ?
– Where is the hardware GC ?
– Where is the sensible hardware bound checking of calculations for seamless exceptions and data integrity protection ( whithout DRM, please ) ?
– Where are the hardware communication channels for parallel programming ( remember Occam ? )
– Where are the neural networks ?
There is so much money at stake that no one can offer to bet on radical new concepts which also clearly implies radical changes to programming languages and operating systems. So I’m pretty pessimistic, future processors will pretty much look like 20 years old systems.
[Treza]
All very good points. (Although I’m not sure whether fixing the GC model in hardware would really be a good idea, because there’s no optimum model for all applications and programming languages.)
I’d also like to see register stacks with automatic spilling (saving instructions and actual spills) and more compact code (reducing bandwidth and cache requirements).
More compact code will mean more code looking like CISC code. Which is x86 territory.
WHO CARES ANYMORE?
Apple will continue to make Apple products. Nice fit and finish, nice performance, gaining more and more consumer buy-in…
Personally, I could care less if they used some previously unknown chip, as long as everything continues to work and performance continues to improve and not degrade.
“Personally, I could care less if they used some previously unknown chip, as long as everything continues to work and performance continues to improve and not degrade.”
Agreed and Intel will make the Mac better. We won’t have short comings and pitfalls with IBM/Moto in performance and availibility.
>”We won’t have short comings and pitfalls with IBM/Moto in performance and availibility.”
While I will agree with you with regard to Moto… IBM thus far has not had performace putfalls in comparison to x86.
It wasn’t about supply, it was about demand. And it was about time. Apple promised a 3 GHz PowerPC Mac and wasn’t able to deliver because of IBM. IBM messed it up. Intel is always developing and putting out newer things all the time. This isn’t the same for IBM’s roadmap at all it seems. The switch is fine, and the consumers don’t care. The ones who DO care are people who don’t have Macintosh systems. The Intel switch will mean lower prices (which we see already lower prices with most systems from Apple) and everyone likes lower prices.