The “sky is NOT falling” OSX on x86 thread...by request

[Jason Vene]

Many thanks to eVolcre, Peace, Sun Baked and others who welcomed me to these boards, and by request I’ve started this thread as a tangent from the 100+ page thread about OSX on x86 ”Mac Moving to Intel Processors”. The first posts here are taken from my posts in that thread, and while this is in reply to eVolcre’s request, all are certainly welcome. eVolcre’s interest is to improve the “signal to noise ratio” regarding OSX on x86 and related topics. I’m breaking these initial posts into a few sections, to avoid message overflow, and group the subjects a little.

So let me introduce myself a second time: I'm a developer, mostly C++ for various OS (Unix, Linux, Apple, Windows - whatever, often cross platform) - been at it for over 25 years.

It seems some of the depression, anger and shock at the Intel for MAC announcement has subsided some. So, perhaps it’s time for a thread to discuss the merits and pitfalls of the transition.

When the choice to base OS X on Unix was made, there is no doubt that among many fine lines of reasoning, portability was at least one main point on the minds of Job and Co. The Unix inside OS X opened up options, and options is what Jobs must have considered, even 5 or 6 years ago.

The Unix interior of the OS was the best possible choice Jobs and Co could have made, in my opinion. No other core has the track record for solidity and performance. They instantly inherited years of superiority over Windows. They still have some work to do, though - plenty of room for improvement even if they stopped advancement the GUI. It’s reported that threads, for example, have a whopping 10x performance hit in server based applications (see Anandtech’s review on this), compared to other operating system implementations. At first, it may not seem pertinent to workstation or personal use, but multicore machines and threaded algorithms that take advantage of them are all about threads. This is one spot that must be “fixed” at Apple. It’s not inherent in Unix itself – but it’s not really a show stopper yet.

The implication of the x86 in all this is nearly meaningless. Jobs demonstrated an OSX x86 at the presentation Monday, and intimated that x86 builds have been made all along. As a developer of cross platform code myself, I can assure you that many if not most of the best Apple applications will port to x86 without much hassle or impact.


No matter what you may have thought about the PowerPC, it's just a bunch of switches, and so is the Intel or AMD. It's never been the sacred heart of anything. That has always been the OS, the applications and the standards of metaphors in the user interface, and the guiding hands that pull all of that together. The chip has been all but irrelevant all that time. If that were not true, Mac wouldn't have continued to be Mac as the chips transformed one to the next.

Every part of the "personality" of the operating systems, be it Unix, Windows, Mac or other, has little or nothing to do with the CPU executing it. Even the viruses that currently attack the operating system, though somewhat more intimately tied to the CPU's own language and method of operation, direct most of their attention to the OS or an application running within it.

Likewise, the range of applications which give the Mac platform guts is entirely dependent on designs which have much more to do with the minds of the people who make them, than with the language of the CPU or it’s feature set.

(continued)

[Jason Vene]

Intel chips aren’t going to corrupt or cripple Apple’s products, or change the nature of Apple. The CPU demands certain things of it's support chips, of course, but the character of the hardware - it's options, it's support for various standards, buses, memory - are all the result of designers taking advantage of options open to them, and choosing compromises in favor of one thing or another. While the CPU and it's supporting chipset provide you with options for PCIe or AGP, built in Firewire or USB, IDE, SATA or SCSI, it is the designer's choices that put those things there, aimed at pricing some product toward low end, medium, high end, workstations, servers or whatever. In this manner, Apple will remain Apple - and true to form, you'll get exactly what you've expected from previous offerings; tight integration of those components way over and above what the standard Wintel platforms have offered - all chosen as compromises among the various options available to them from the various chipsets and standards (including what Apple may have designed themselves), and implemented according the personality of the guiding hands at Apple.

There will be nothing "Windows" like about an x86 Apple, unless you boot Windows on the machine. Of that I'm absolutely certain, unless of course Jobs dies or leaves the company before it's all implemented.

Don't overlook the fact that Unix is the interior of OS X. Unix is THE ORIGINAL portable operating system. Likewise, virtually anything a COMPETENT developer makes for Unix SHOULD have some portability notion in mind. I'm not talking about OS portability here either, I'm talking about chip independence.

A primary example is Unix itself. You don't re-write Unix to move it to a new chip. You re-write C, the compiler upon which it's based - and generate a Unix kernel for the new chip with it, and all else that moves to the new chip too.

Now, along the way, I'll bet there ARE a few lazy programmers that let slip into production code that's dependent on the chip in some way. But Unix and C breathe chip independence. That MUST have been on the minds of the developers working on OS X, all internal Apple products, and SHOULD have been on the mind of any developer making MAC OS X products.

I'm not saying it's effortless - I'm saying unless the product is from incompetent hands, it's not only possible, it's practicable.

There's one thing everyone should have learned when that PDP-8 was upgraded to the PDP-11 (I think those were the models, heck that was 69-70), which spawned the development of C and Unix in the first place.

If it didn't hit us then, it should have become a philosophy in a single word when the 8086 gave way to the 286, and then quickly to the 386.

If it didn't hit us then, when Mac moved from MC680X0 to the PowerPC, it should have been obvious.

However, for many who bought into the philosophy that the PowerPC was the best GDMF CPU ever designed, and always will be - the one word philosophy we should have learned in 1970, again in 1986, in 1989, in 1995, and now is:

CHANGE

...and THAT is why we must insist on portable, independent methods of construction. That is what Unix was all about, and I think Jobs really got that during the "NEXT" phase of his career.

I've skimmed past a bunch of Intel negative posts here, and most nod a positive thought toward AMD by comparison. The PowerPC chip, as monstrously wonderful as the design is (even the earlier versions), the implementation of that design never really delivered as much benefit as it seemed it should. I remember my first look at the assembler potential; all those registers, all that power, and yet - so many obstacles remained in the way toward really significant performance dividends.

Like all comparisons of performance between chips, there are things the PowerPC does better than P4, and there are things the P4 does better than the PowerPC. AMD jumps in with some serious advantages over both, with yet another set of roadblocks that keep them from overpowering either of these by generous factors in all regards. In other words, it's like comparing three engines, each with 400, 410 and 425 horsepower, different ranges of torque and efficiency. They're all capable; what translates into results depends on implementation, like the weight of the final vehicle, the stability of the thing where the rubber hits the road - even the impact of air over the contour of the outer skin.

Then....what happens if P4's future plans don't quite live up to Jobs expectations? Consider the shudder going through Intel execs to read the headline "Apple considers AMD." Even if the volume doesn't warrant it, the promotional aspect of Apple x86 will have them flogging the engineers day and night to keep that from happening. And what does that mean? The motivation for competition against AMD will be stronger than ever, and AMD's reply will be stronger than ever. Considering the recent history in that regard, this is great news for everyone.

We will, one day, not too far off I think, have the delightful option of considering a 64 core workstation. Our choice will hinge on "Apple or Non-Apple" versions, depending on whether or not we want the OS X. It's up to Apple and it's public as to just how popular the "Apple" option will become to consumers.



Considering the huge thread “MacIntosh Moving to Intel Processor” there seems to be considerable emotion behind this subject.

There's a whole group of "sky is falling" advocates. Many have the feeling the x86 chips are awful, and have the view that PowerPC chips were technologically orders of magnitude better. Mix in with that the view that Intel has no innovative brains left, and only PowerPC had the future of computing in hand.

CPU's are going to evolve. By the time Apple puts an Intel chip in a machine for consumer purchase, it will not be the P4 you know today. Take that in with a deep breath if you're among those with the belief that all things Intel are junk.

If you developed a fondness for the PowerPC, under the assumption it was so much more powerful, then this move probably irritates you. Well, first, most of the performance gains you’ve been told to believe are actually myth and marketing hype. Oh there certainly are “whiz-bang hyper” this, and “hi-speed super” that, and “multi-ganged” stuff here, and “vector-whatnot’s” plastered all over these chips. They do serve genuine purposes, but don’t be deceived by pointless arguments of performance benefits. Opterons, Xeons and PowerPC chips are all much more powerful than what we used a few years ago, and each of them have leapfrogged the other in terms of features, performance and flexibility. If you ever thought the PowerPC chip was the heart of Apple’s superiority, or that the PowerPC was the best CPU ever built and the no other would ever catch up to it, you were mislead by marketing. It’s a “group identity” concept used to take advantage of humans’ tendency to think in terms of “us” versus “them”. Any moment along the timeline that one of these processors legitimately had the “fasted CPU” claim to fame, it was lost within a matter of weeks to another. Every favored feature you’ve known that gives PowerPC and advantage has a thousand technical “gotchas” you’ve never heard about, and so, to, it is for Hyper threading in the P4, or the “Quantispeed” of the AMD, or any other feature you can name.


Jobs is no idiot. He's not about to select a chip from TODAY'S lineup for machines to be released in 12 months or more. He's got engineering samples of chips no P4 customer will see for a year, and may just know a thing or two about what Intel performance will be compared to PowerPC's roadmap for a computer to be released in 12 months. Consider that before you spend too much energy concerned about Intel's current offerings. Have a little faith in Job's judgment here. He's had a good record about "looking ahead" for a while now. While some might think choosing Intel is the worst selection in history, don’t assume Jobs isn’t at least as aware of all those things his company printed about G3, G4 or G5 performance comparison to Intel chips. Is it that he didn’t mean it? It’s more like that exercise we had in grade school, where they made us create a fake menu, trying to make simple dishes sound wonderful. What would you expect them to have said, “PowerPC is just another wafer, it’s what I put into that counts.” – Even though that’s essentially true, it doesn’t build any kind of mental image that gives motivation to the fans. Depicting the Intel Blue dancers on fire; now that’s imagery. It wasn’t about honesty, it was about creating desire. The fact is, the PowerPC, as good as it is, is just another way to do the same thing.

[Jason Vene]

So, bring in your best thoughts, pro, con, whatever.

[wrldwzrd89]

I agree with you 100%. I haven't bothered to read the thread from which this one was spawned, but I still think that the transition to Intel chips will help Apple in the long run.

[Applespider]

I agree that the sky isn't falling in... and it's a fine rational post which I agree with.

But I hate to say that this thread will fill up with just as much FUD as has strangled every other Mac/Intel thread that has started...

Good job for trying though! ;)

[biohazard6969]

change is good, go with the flow, i wouldn't worry, steve prbly has something up his sleeve and this can't be as bad as we all make it out to be

[iSaint]

As someone who doesn't approach programming, and understands little about the inner workings of the Apple computer, I appreciate this thread. I understand a bit more. However, I also didn't think 'the sky was falling' due to Apple's dedication to quality systems and superior OS.

Thanks

[QCassidy352]

Thanks for the in depth reply in language that we all can grasp. I agree with you except about one paragraph:

Then....what happens if P4's future plans don't quite live up to Jobs expectations? Consider the shudder going through Intel execs to read the headline "Apple considers AMD." Even if the volume doesn't warrant it, the promotional aspect of Apple x86 will have them flogging the engineers day and night to keep that from happening. And what does that mean? The motivation for competition against AMD will be stronger than ever, and AMD's reply will be stronger than ever. Considering the recent history in that regard, this is great news for everyone.

The reason this logic strikes me as suspect is that it's very much what was said when Apple announced the G5 two years ago. "IBM has the resources and the desire to beat intel and AMD; now it will be a real race instead of the joke it was with Motorola, which just gave up on the G4."

In all honesty, I never would have expected IBM to falter as badly as they did in terms of both yields and speed bumps with the 970. It's easy to say now that the same won't happen to intel, but I'll believe it when I see it.

[Jason Vene]

QCassidy352:

You make a good point, and one I was hinting toward, yet didn't quite bring to fruit in that form.

I'm not trying to say Apple will depend on Intel entirely. Indeed, the switch to x86 invokes the option of using AMD chips, and I for one would be very happy if they did.

I read one post that referred to Intel as a "loose wheel on the wagon" or words to that effect, and indeed, Intel has been on defense since AMD took the lead on 64bit x86 AND dual core offerings.

Actually, though, Intel was on the defense a bit earlier than that. Intel put it's thrust behind less circuit complexity in favor of raw speed, while AMD recognized that raw speed increases become more meaningful when the circuit complexity returns higher IPC.

I think, though, that Intel's current "copy" of the x86-64 and response to dual core AMD isn't quite up to par. I read that virtually all of the engineering staff previously devoted to Itanium have been moved, which might bring Itanium like design concepts into the Pentium lineup, making their response against AMD's next set of moves a bit stronger (if not, they'll fall even further behind and have to scramble). If Jobs does have an insight into future Intel plans, and assuming those plans aren't pure hot air, the move is quite correct. If not, then Apple's machine, at first, won't have the guts of an AMD box, yet AMD will be an option (depending on the exact details in his current 'new' contract with Intel).

One thing Intel can deliver is chips. Maybe not the best designs, but they can deliver. AMD has gotten better, and Apple is about the size AMD can supply, but still AMD is likely no better at yields than IBM's PowerPC division. If Apple has any hope of making 4, 8 or more core workstations and/or servers, AMD is already set.

While I agree with your view of "when I see it, it becomes believable" with regard to Intel products, they're already in "surprise" mode with response, and they have the cash to devote to the problem, possibly new brains on the project, and more motivation than ever to regain position. Even if they don't succeed, they will be releasing better chips than today, and AMD simply MUST be thinking about ways to get Jobs attention.

[eVolcre]

I agree that the sky isn't falling in... and it's a fine rational post which I agree with.

But I hate to say that this thread will fill up with just as much FUD as has strangled every other Mac/Intel thread that has started...

Good job for trying though! ;)

Awesome, Thanks Jason. I just saw this. And be forewarned, I'm making a top 10 list of questions to educate myself. :)

Applespider, yeah but I think if a few of us stay involved we can prevent that from happening. Either by not responding to the FUD and not rehashing old info or by reporting a post to the moderator. I really want to see THIS thread reach a 100 pages - but with mature and intelligent posts.

OK - back in a bit.

eV

[paulypants]

At this point there are only 2 things I'm personally worried about:

1) Being absolutely positive that Apple has a way to lock OS X to Macs

2) If you can boot Windows on a Mac or run WINE, then why would many developers spend the extra $$ to port Mac OS versions of their software? Personally I hope that, in the end, Apple will make it difficult if not impossible to install Windows on Macs, but I doubt it.

[Tilmitt]

Omg you people are ricidulous! You've all convinced yourselves that this backward architecture which most of you ridiculed is somehow a progression! According to Macpolls.com 74.9% of you did not want this to happen before the announcement, and even then most of you thought that as bad as it would be to go to Intel PPC, x86 would be utterly unthinkable. Yet after the announcement again according to Macpolls only 20.78% of you thought it was a bad thing. This is crazy! And it's sad to see people just go with what Steve says despite it being completely different to what they believe. Furthermore you all harp on saying "oh actually this is really good and it was obvious all along that we should have gone to x86!" Principles go down the drain with you people.

To quote a friend "So this is how Power PC dies, with thunderous applause."

You guys are the clapping senators as Palpatine leads you into darkness!

[Jason Vene]

PaulPants:

Point 1 is quite valid. For every possible means I know that CAN tie the OS to some hardware, I also know of a means to defeat it. It was reported that the development x86 version of OS X is already in the underground pipeline (available for download at this moment), and is NOT locked up.

As a counterpoint, though, I'm not sure what market force pirated operating systems has on the bottom line. Microsoft's Windows is probably the most pirated OS in the world, and while they claim it's robbed them of billions, they're not hurting much.



Point 2, I'm not quite so certain about. Since you can boot Linux and FreeBSD (and others) on a Windows machine, why would developers continue to develop for Windows? Or the reverse?

Alternately, consider the point that development for Apple just got easier. Until now, I'd have to purchase an Apple PowerPC to develop Apple counterparts to my Linux / Windows builds. Now, when I upgrade my development machine, I can consider an Apple development system upon which I can develop all 3. I may even be able to unify my development toolset in the process, a little.

[paulypants]

Omg you people are ricidulous! You've all convinced yourselves that this backward architecture which most of you ridiculed is somehow a progression! According to Macpolls.com 74.9% of you did not want this to happen before the announcement, and even then most of you thought that as bad as it would be to go to Intel PPC, x86 would be utterly unthinkable. Yet after the announcement again according to Macpolls only 20.78% of you thought it was a bad thing. This is crazy! And it's sad to see people just go with what Steve says despite it being completely different to what they believe. Furthermore you all harp on saying "oh actually this is really good and it was obvious all along that we should have gone to x86!" Principles go down the drain with you people.

To quote a friend "So this is how Power PC dies, with thunderous applause."

You guys are the clapping senators as Palpatine leads you into darkness!

Palpatine? How old are you? What would you like us to do, arm ourselves and start a rebellion? Oh, how about picket outside of 1 Infinite Loop! Picketing is so "in" these days, an utter waste of time, but people think it's cool. C'mon, the best thing to do is remain positive and look at the bright side instead of solely focusing on the negatives. Roll with the punches...

[Jason Vene]

Tilmitt:

I for one have never been among those saying it was a bad idea, though I have crossed many posts with the theme: at first I was horrified, then after I thought about it, now I'm jazzed.

Exactly what principles are you talking about here, too?

What are you expecting is going to be problematic?

What hardware oriented character traits do you think will be lost, and what will be adopted (good and/or bad)?

What about Intel the architecture is backward?

Who actually killed the PowerPC?

Why would the PowerPC be better?

The truth behind these and other points is the reason for this thread...so come back for more, will you?

[paulypants]

PaulPants:

Point 1 is quite valid. For every possible means I know that CAN tie the OS to some hardware, I also know of a means to defeat it. It was reported that the development x86 version of OS X is already in the underground pipeline (available for download at this moment), and is NOT locked up.

As a counterpoint, though, I'm not sure what market force pirated operating systems has on the bottom line. Microsoft's Windows is probably the most pirated OS in the world, and while they claim it's robbed them of billions, they're not hurting much.



Point 2, I'm not quite so certain about. Since you can boot Linux and FreeBSD (and others) on a Windows machine, why would developers continue to develop for Windows? Or the reverse?

Alternately, consider the point that development for Apple just got easier. Until now, I'd have to purchase an Apple PowerPC to develop Apple counterparts to my Linux / Windows builds. Now, when I upgrade my development machine, I can consider an Apple development system upon which I can develop all 3. I may even be able to unify my development toolset in the process, a little.

Thanks for that. I agree, MS is definitely not hurting too much, however the y have a much higher volume because they've licensed their OS and aren't tied to specific hardware. I have heard that the dev build is out there, but it is a 1-off build and would be pretty useless as future updates would not be available for it. Hopefully the piracy option will just be a headache for end users in the long run and won't run rampant.

Indeed, but say game developers still have to port many games from DirectX to OpenGL, which entails much work. They don't port many as it is now, if the dynamics change it may get worse (conversely it may get better :))

Here's to hoping, I'm positive Apple is way ahead of me with these concerns... :)

[Tilmitt]

Palpatine? How old are you? What would you like us to do, arm ourselves and start a rebellion? Oh, how about picket outside of 1 Infinite Loop! Picketing is so "in" these days, an utter waste of time, but people think it's cool. C'mon, the best thing to do is remain positive and look at the bright side instead of solely focusing on the negatives. Roll with the punches...

Well i thought most people on this site could relate to the Palpatine reference. But it was only a bit of fun, I love to equate things with Star Wars these days! Ideally I would like everyone to boycott Intel Macs. I will be. I'm going to load up a powermac with 8GB of ram and best graphics card and whatever just as they're about to be discontinued and I shall never buy another computer, mac or not, with an x86 CPU in it. I'll last at least 5 years and I have nowhere to go for the foreseeable future so it's likely that i will run that system into the ground before having to buy something new. Hopefully by then there'll be more elegent options than x86. Few people seem to have any fighting spirit. It's sad.

In any case I'll be odering a 15" Powerbook tomorrow, and i'll be proud to be running an elegent powerpc!

[Tilmitt]

Tilmitt:

I for one have never been among those saying it was a bad idea, though I have crossed many posts with the theme: at first I was horrified, then after I thought about it, now I'm jazzed.

Exactly what principles are you talking about here, too?

What are you expecting is going to be problematic?

What hardware oriented character traits do you think will be lost, and what will be adopted (good and/or bad)?

What about Intel the architecture is backward?

Who actually killed the PowerPC?

Why would the PowerPC be better?

The truth behind these and other points is the reason for this thread...so come back for more, will you?

I don't expect anything to be problematic. Apple will likely pull this switch off very well. I disagree strongely with it on ideological grounds. We will loose an architecture that is elegent and gain a monstrosity of a relic from the 1970's. x86 manages things really dirty deep down, it's ugly. It's just wrong and so what if we gain praticalities! Apple isn't about trading off elegence to get things cheaper, it's about doing things right. I really thought Apple were something special, I don't know what's gotten into them.

In truth IBM and Freescale haven't been trying their best to agressively ramp the PPC. I admit that and perhaps Apple were in a less than desireable situation. But we weren't getting tr0wned by PC's or anything. And I think there was hope on the horizon. Just as Intel has all these wow magical products planned etc (the exact opposite of their very own net-burst pos architecture) freescale and IBM had dualcore and whatnot well on the way.

The PowerPC would be better from a elegent engineering viewpoint. Now we're left with a pratical x86 monoculture with no desktop RISC chips being produced and enhanced. I believe this will stifle innovation and it's frankly a disgrace that so much Research and Development is put into a horribly backward architecture that is x86. Humanity would be much better served if that effort was put into newer designs. Intel and AMD have covered their ugly daughters in an inch of makeup, but they're still ugly.

[paulypants]

Well i thought most people on this site could relate to the Palpatine reference. But it was only a bit of fun, I love to equate things with Star Wars these days! Ideally I would like everyone to boycott Intel Macs. I will be. I'm going to load up a powermac with 8GB of ram and best graphics card and whatever just as they're about to be discontinued and I shall never buy another computer, mac or not, with an x86 CPU in it. I'll last at least 5 years and I have nowhere to go for the foreseeable future so it's likely that i will run that system into the ground before having to buy something new. Hopefully by then there'll be more elegent options than x86. Few people seem to have any fighting spirit. It's sad.

In any case I'll be odering a 15" Powerbook tomorrow, and i'll be proud to be running an elegent powerpc!

Well I'm glad to hear that you still love your PPC macs! Maybe in 5 years they'll be back to using PPC or something else...

In any case do whatever's best for you man... :)

[paulypants]

Intel and AMD have covered their ugly daughters in an inch of makeup, but they're still ugly.


LOL now that's funny!

[amin]

If you developed a fondness for the PowerPC, under the assumption it was so much more powerful, then this move probably irritates you. Well, first, most of the performance gains you’ve been told to believe are actually myth and marketing hype. Oh there certainly are “whiz-bang hyper” this, and “hi-speed super” that, and “multi-ganged” stuff here, and “vector-whatnot’s” plastered all over these chips. They do serve genuine purposes, but don’t be deceived by pointless arguments of performance benefits. Opterons, Xeons and PowerPC chips are all much more powerful than what we used a few years ago, and each of them have leapfrogged the other in terms of features, performance and flexibility. If you ever thought the PowerPC chip was the heart of Apple’s superiority, or that the PowerPC was the best CPU ever built and the no other would ever catch up to it, you were mislead by marketing. It’s a “group identity” concept used to take advantage of humans’ tendency to think in terms of “us” versus “them”. Any moment along the timeline that one of these processors legitimately had the “fasted CPU” claim to fame, it was lost within a matter of weeks to another. Every favored feature you’ve known that gives PowerPC and advantage has a thousand technical “gotchas” you’ve never heard about, and so, to, it is for Hyper threading in the P4, or the “Quantispeed” of the AMD, or any other feature you can name.


Jobs is no idiot. He's not about to select a chip from TODAY'S lineup for machines to be released in 12 months or more. He's got engineering samples of chips no P4 customer will see for a year, and may just know a thing or two about what Intel performance will be compared to PowerPC's roadmap for a computer to be released in 12 months. Consider that before you spend too much energy concerned about Intel's current offerings. Have a little faith in Job's judgment here. He's had a good record about "looking ahead" for a while now. While some might think choosing Intel is the worst selection in history, don’t assume Jobs isn’t at least as aware of all those things his company printed about G3, G4 or G5 performance comparison to Intel chips. Is it that he didn’t mean it? It’s more like that exercise we had in grade school, where they made us create a fake menu, trying to make simple dishes sound wonderful. What would you expect them to have said, “PowerPC is just another wafer, it’s what I put into that counts.” – Even though that’s essentially true, it doesn’t build any kind of mental image that gives motivation to the fans. Depicting the Intel Blue dancers on fire; now that’s imagery. It wasn’t about honesty, it was about creating desire. The fact is, the PowerPC, as good as it is, is just another way to do the same thing.

I think a lot of people are sore at Apple for having "tricked" them into thinking PPCs are far superior in every way. I just want a fast thin Powerbook running OS X. No time to be mad at anyone.

[Applespider]

I think a lot of people are sore at Apple for having "tricked" them into thinking PPCs are far superior in every way. I just want a fast thin Powerbook running OS X. No time to be mad at anyone.

When the G5 was released, it was superior in many ways. Apple could brag quite happily. For certain tasks/applications, it still beats out x86 chips but for generic overall performance, it doesn't any longer. They're now saying that in 2 years time, it doesn't look like the G5 will beat anything in anything. They haven't claimed any superiority about the G4 in ages.

They haven't 'tricked' anyone - things have changed as time has passed.

[Jason Vene]

So, to you Intel is ugly, and the PowerPC is elegant.

...and what is RISC?

It is a technological relic of the 1970's.

In the early days, say 1960 or before, it was observed that as programs became more complex, and larger, the demands on RAM were becoming too expensive. 16K of RAM was as expensive as a house. To make better use of the RAM, the CPU's language had to become more complex, so the same complexity of software could be managed with a smaller program.

Since the internal workings of the computer are, in essense, an electronic version of a mechanical process, the primatives in the CPU were capable of doing only a certain limited set of things. To add to the complexity of it's language, a common collection of assembler routines were analyzed, then grouped together into a single command - almost literally moving subroutines out of the assembler code into the CPU.

By the time of the 6502, the 6800 and the Z80 processors, the notion of using a limited set of primative circuits to do the actual work, and a list of instructions inside the CPU which could manipulate those primatives in a way that created more complex results, was typical. This is what we now call a CISC - complex instruction set computer.

In the 70's, an IBM research scientest, John Cocke, came up with the idea of returning to older days. Since RAM was no longer as precious as it once was, but speed certainly was, let the circuit complexity increase dramatically so that instead of working more primative circuits with several steps, each instruction was performed by a circuit in one step. Since there's limited real estate for transistors, this necessitated the prioritization of the assembler vocabulary, so as to keep the complexity of the result circuits accomplishable. Those instructions which couldn't fit were removed (the Reduced part of the RISC acronym). This means the programs got larger, because what once was done with one instruction (but many steps), was now done with several instructions (each taking only one step) - but there was a net gain in performance. A considerable gain at first.

Years later, IBM's work with Cyrix and Nexgen created a "RISC86" chip. The ability to cram increasing volumes of transisitors into the CPU meant that we no longer had to "reduce" the instruction vocabulary to achieve performance gains.

The point of RISC was about increasing circuit complexity, so that circuits were selected to perform more complex tasks. It wasn't about reducing the instruction set as the name implied. As the availability of transistor counts beyond the 10's of millions became available, the notion of what a RISC and CISC designs began to converge. RISC86 didn't impose the general requirement of most RISC implementations (that the instruction code always be the same length), so now a RISC implementation didn't automatically mean larger code sizes.

Today, RISC design ideas are used in nearly every modern CPU. What we have with AMD and Intel designs are hybrids, which retain benefits from both design notions.

Was the PowerPC's implementation so much better? If so, then why is that a dual G5 Apple (at 2.3 Ghz) take about the same amount of time (and sometimes more) than my dual AMD (at 2.0Ghz) to render a Maya project?

Why is the PowerPC so much better if, for example, the integer IPC for the G5 is 2 while while it is 3 or 4 for the Opteron and Xeon, or the floating point IPC for the G5 is 2, while the Opteron is 3?

Why is Altivec so much better if it doesn't deliver superior results with Maya, Photoshop or other applications?

Sure the front side bus claims are impressive, but memory bandwidth is 6.4 GBytes per second. That's the same for Opteron and Xeon, at present.


Now, perhaps it's the ugly "real" mode that gets to you. Intel and compatible chips have two modes that really do work like the ol' 8086 from 1980 - real mode itself, and it's cousin, the virtual 8086 mode. Those modes hardly see any use now.

Perhaps you're preference for "big endian" format strikes you. I can hardly see how it makes much of a difference - a bit is a bit to me.

Perhaps you find the x86 assembler more like "German" to the PowerPC's elegant "French". When was the last occasion you had to converse with the CPU in it's native tongue anyway?

For Apple's sake, I hope you do purchase a dual G5. Apple could use the sale, the machine WILL be good for several years, and it may be a collectable one day.

What I question most, though, is exactly what is really meant by "elegant" in reference to the PowerPC - and what importance does it play in the utility of a computer?

[Rocksaurus]

Great post Jason, I'm going to save that as a text file, really informative. Not to mention you just owned tilmitt.

[Tilmitt]

So, to you Intel is ugly, and the PowerPC is elegant.

...and what is RISC?

It is a technological relic of the 1970's.

In the early days, say 1960 or before, it was observed that as programs became more complex, and larger, the demands on RAM were becoming too expensive. 16K of RAM was as expensive as a house. To make better use of the RAM, the CPU's language had to become more complex, so the same complexity of software could be managed with a smaller program.

Since the internal workings of the computer are, in essense, an electronic version of a mechanical process, the primatives in the CPU were capable of doing only a certain limited set of things. To add to the complexity of it's language, a common collection of assembler routines were analyzed, then grouped together into a single command - almost literally moving subroutines out of the assembler code into the CPU.

By the time of the 6502, the 6800 and the Z80 processors, the notion of using a limited set of primative circuits to do the actual work, and a list of instructions inside the CPU which could manipulate those primatives in a way that created more complex results, was typical. This is what we now call a CISC - complex instruction set computer.

In the 70's, an IBM research scientest, John Cocke, came up with the idea of returning to older days. Since RAM was no longer as precious as it once was, but speed certainly was, let the circuit complexity increase dramatically so that instead of working more primative circuits with several steps, each instruction was performed by a circuit in one step. Since there's limited real estate for transistors, this necessitated the prioritization of the assembler vocabulary, so as to keep the complexity of the result circuits accomplishable. Those instructions which couldn't fit were removed (the Reduced part of the RISC acronym). This means the programs got larger, because what once was done with one instruction (but many steps), was now done with several instructions (each taking only one step) - but there was a net gain in performance. A considerable gain at first.

Years later, IBM's work with Cyrix and Nexgen created a "RISC86" chip. The ability to cram increasing volumes of transisitors into the CPU meant that we no longer had to "reduce" the instruction vocabulary to achieve performance gains.

The point of RISC was about increasing circuit complexity, so that circuits were selected to perform more complex tasks. It wasn't about reducing the instruction set as the name implied. As the availability of transistor counts beyond the 10's of millions became available, the notion of what a RISC and CISC designs began to converge. RISC86 didn't impose the general requirement of most RISC implementations (that the instruction code always be the same length), so now a RISC implementation didn't automatically mean larger code sizes.

Today, RISC design ideas are used in nearly every modern CPU. What we have with AMD and Intel designs are hybrids, which retain benefits from both design notions.

Was the PowerPC's implementation so much better? If so, then why is that a dual G5 Apple (at 2.3 Ghz) take about the same amount of time (and sometimes more) than my dual AMD (at 2.0Ghz) to render a Maya project?

Why is the PowerPC so much better if, for example, the integer IPC for the G5 is 2 while while it is 3 or 4 for the Opteron and Xeon, or the floating point IPC for the G5 is 2, while the Opteron is 3?

Why is Altivec so much better if it doesn't deliver superior results with Maya, Photoshop or other applications?

Sure the front side bus claims are impressive, but memory bandwidth is 6.4 GBytes per second. That's the same for Opteron and Xeon, at present.


Now, perhaps it's the ugly "real" mode that gets to you. Intel and compatible chips have two modes that really do work like the ol' 8086 from 1980 - real mode itself, and it's cousin, the virtual 8086 mode. Those modes hardly see any use now.

Perhaps you're preference for "big endian" format strikes you. I can hardly see how it makes much of a difference - a bit is a bit to me.

Perhaps you find the x86 assembler more like "German" to the PowerPC's elegant "French". When was the last occasion you had to converse with the CPU in it's native tongue anyway?

For Apple's sake, I hope you do purchase a dual G5. Apple could use the sale, the machine WILL be good for several years, and it may be a collectable one day.

What I question most, though, is exactly what is really meant by "elegant" in reference to the PowerPC - and what importance does it play in the utility of a computer?

You are certainly right that many and if not all of the innovations maybe by people designing RISC CPU's have be reincorporated into x86. Apparently much of modern x86 do things like a new RISC CPU would do in it's higher level parts and only in the very low level does it convert it's instructions to standard x86. But I resent that such a situation exists. Even Intel does, it's Itanium was an attempt to move on to a more elegent architecture. Alas the market proved too much for even them and Itanium has been a pretty big failure.

Here i'll quote an article (http://arstechnica.com/columns/mac/mac-20050607.ars) that helps explain my point of view.

"Sure, x86 has the market-share and usually the speed, but is it elegant? Does it turn the CPU geek knobs all the way to 11? Is it sexy? No, not really. In fact, it's pretty darned ugly.

Yes, this all sounds silly, but it's a real phenomenon. Right or wrong, sensible or not, this is how a lot of people feel about PowerPC vs. x86 (or 68K vs. x86, for that matter). I'm one of the biggest x86 haters. I've often argued that the collective human effort spent making fast implementations of the bass-ackwards x86 ISA would be much better spent elsewhere. Oh, I fully realize the market realities that conspire to make all of this x86 effort worthwhile, but this is about emotion, not reason. And if I didn't give significant weight to my feelings when it comes to my platform choice, would I really have been a Mac user for the past 21 years?

So yeah, I'm sad that the PowerPC is leaving the Mac platform, even more so because of the really interesting things going on with that ISA in the upcoming crop of game console CPUs. It will pain me to know the contortions that instructions are going through in an x86 CPU inside a Mac. I will miss the idea of AltiVec, and the promise of powerful new CPUs arriving "out of nowhere" to power new Macs, a la the G5. I'll miss the interesting things that "only Macs can do" thanks to clever CPU features like AltiVec, or even just a particularly fast barrel shifter. I'll miss the dream, even if it was always destined to be just that."

I'm pleasantly surprised the way you guys haven't flamed me or anything, ye just argued back with a clear head. I certainly will still be trumpeting my new macs abilities to everyone i meet and perhaps someday I'll even be enthusiastically rumouring about the latest RISC cpu's for Apple, even if they come from Intel.

[mukansa monkey]

Some comments and a question.
After going through the phases of denial and anger, I have managed acceptance. From a software point of view, this switch should be a good thing. A handful of smaller software vendors will doubtlesly be lost. However, all reports are saying that rewriting for the new chip will be relatively painless. Supposedly Apple's software is already ready to go on Intel chips... and certainly all the apps should be by the time the new hardware goes on sale.
There will probably be a generation of major third-party software that will be installable on both hardware types. Then the version after that will require Intel chips. Which is really no different from most other two-generation intervals out there... some software that was brand-new when the 266Mhz G3's were first released doesn't run in Classic on a G5 running Tiger. Another half-dozen years from now, supporting PPC chips will be irrelevant.
Touching on what PaulyPants and Mr. Vene said, this move should encourage vendors to port more software to the Mac, because it will probably be a lot easier with Intel chips underneath. Results will vary, of course, depending on the software. But game makers in particular, who I'm told rely heavily on low-level interaction with the chips, will hopefully find it much easier to make Mac versions with Intel chips underneath. I don't think we'll see OSX go the way of OS/2, for two reasons. One is that Apple makes hardware. You buy an Apple machine that comes with OSX, and no Windows. You'd then have to go out of your way to put Windows on it. The other reason is that OS/2 wasn't superior to Windows the way OSX is. I seriously doubt many folks who buy a Mac are going to say, "Man, this sucks. I'm just gonna use Windows for everything I can."

This assumes that OSX won't run on non-MacIntel machines, which people seem to be generally taking for granted. My guess is that as part of this deal Apple is giving Intel some chip tech involving hperthreading, Altivec, or some such. Then the MacIntel chips will have features that standard Wintel chips do not. Thus OSX won't run on Wintel chips, while Windows will run on MacIntel, only without being able to fully utilize the chip.

So: Apple ditches a chip maker that's increasingly unable to provide the production and improvements they need to remain competitive. They gain software titles due to low-level porting being easier. In order to create a bigger product base for the online media-purchasing systems they already dominate, they pick up Intel's digital rights management capabilities. Imagine a day when you no longer need cable TV, because you can watch what you want cheaper through Apple's iMedia store.
And what about Microsoft, who you'd think would be unhappy about sharing a chipmaker with their OS enemy? The thing is, Microsoft isn't a hardware maker, but Apple is. How much does Microsoft really care who sold the hardware, as long as the customer is still buying MS Office? This move will probably cut into sales of Windows, as people realize how much better OS X is. However, Windows is already hugely pirated. Office is doubtlessly more profitable. The way Longhorn is (or isn't) coming, Microsoft may have decided that they don't have much hope of coming out with an OS that's superior to OSX. Combine all that with the fact that Windows is losing more market share to Linux than they are to Mac, and I can see M$ deciding that it's better to accept OSX gaining market share in order to reduce the ranks of people running Linux and buying nothing from M$ at all. As Jobs pointed out, Office is the Microsoft product nobody really wants to mess with.

My question is: what is it about Intel chips that's so bad compared to AMD? Unlike Tilmett I'm not particularly convinced of the G5's superiority. I've read that the G5 was a bit of a kludge, with stuff like the Altivec engine just tacked on instead of properly incorporated. It certainly hasn't been improved upon significantly in the last two years, and from what Apple's saying it's not going to be in the future. So, lack of elegance notwithstanding, Intel and AMD are going to make chips that perform better than anything PPC has to offer. To me that is all that matters. Apart from Intel's having the DRM business and better expected production ability, any thoughts on why Intel instead of AMD? I don't know all that much about the hardware differences. I've never had to care before. :|

PS: Having just read the last couple posts, I'd like to point out that many of us haven't been Mac owners for years due to primarily emotional reasons. As far back as 1988 I compared Apple software to M$ software and found that what was written for Apple was simply better. Years later I was agog to see that Win95 had a ugly, blocky 16-color interface when Mac had been 256 for years already. For me this has always been about superior functionality, which is mostly software. And who's to say that MacIntel chips won't have some sexy additions like keeping Altivec? I really doubt Macs will have generic Pentiums inside. I expect there will always be things that "only Macs can do". And finally we'll be able to put to rest some of the arguments over benchmarking, with OSX and Windows on the same machine the differences wil become clearer.

[wrldwzrd89]

The big thing against Intel right now is their power-hungry "Netburst" architecture (which is the philosophy behind the Pentium 4). Fortunately, Intel realized that Netburst is a dead-end before it was too late, and have started taking steps to purge it from their lineup. Once this happens, Intel and AMD will be on more equal terms.

[psycho bob]

I think my biggest fear wasn't the fact that PPC CPU's were going to be dropped but the tie up solely with Intel. I'm all for OS X on x86 if the speed gains are going to be there but Intel do not make the best chips for desktops and haven't for years. I'm not a real techie but I know enough. None of my PC using friends, both those who use them for work in the programming and aerospace fields as well as those who just use them as general purpose machines will choose Intel over AMD. Intel are currently the company to go to for mobile technology I don't think anyone would debate this but in the high end workstation and desktop arena AMD rule the roost. The Opteron and new Athlon X2 are world beating chips. Now obviously if supply is going to be an issue then fair enough but there have been no real complaints from with in the resale industry about not being able to get AMD offerings.

I think Apple could have put a lot more people at ease by just announcing a move to x86. Unless Intel will be making customised chips for Apple why do they have to be tied in with a specific company? AMD's Dresden FAB 36 is supposed to enter into full volume production by 2006 this should solve any of the supposed supply issues.

When the G5 moved to 90nm a lot of people said that it was IBM's inability to master that new process. I feel that statement needs to be revised, it must surely be IBM's inability to produce the G5 under the new 90nm process. While AMD's new FAB is under production it is my understading that IBM took up some of their production at the Fishkill plant. AMD suffered no major problems with their transition. AMD are also on course to introduce 65nm next year. It has always intrigued me how Steve Jobs and Apple never really made any mention of AMD. Steve stood on stage and always compared Apple/IBM chips to Intel. Most PC techies laughed at this knowing full well the performance kings were neither of those parties.

I hope as Jason has said on this thread that Apple have left themselves open to the possibility of using AMD further down the line.

[mukansa monkey]

I'm pretty much assuming that Apple is going with Intel in part because they're going to make customizations (or already have some, like DRM) that Apple wants. Having more power-efficient chips is a big deal for Apple due to their desire, as a hardware manufacturer, to make the best laptops they can. The whole Yonah thing is doubtlessly a significant part of Apple's future. I'm just curious what the essence is of AMD's performance superiority from a technical standpoint.
IMHO, absolute power matters less and less as fewer and fewer people need their whole processor to perform their usual tasks. I think Apple's decision was based mostly on factors other than maximum top end power.

[llamatronique]

Why a partnership with intel can yield good results (http://www.sgi.com/products/visualization/prism/).

[Jason Vene]

Mukansa monkey:

In line with your comments;

I’m not yet convinced that Intel will be motivated to alter their line solely for Mac or Apple. My reasoning is simply based on the notion that Intel has it’s own interests in mind, and has never demonstrated this trait in the past. To Intel, while the Mac news is a PR landslide, it isn’t a major business boost compared to the rest of their business. I think, instead, Mac represents one of a range of high end customers to which they intend to cater, and for which they likely have technological advancements in plan. Whether we’ll see those plans bear fruit is another matter, I must agree.

As to game development – I think the trend will move toward “physic engines” in the future. This is in the form of a drop in card which performs simulation calculation similar to the graphics engines we see now, but in the domain of physical simulation which is not in the graphic’s chip at present, and which even dual or quad CPU systems could use a boost. Though it might be best if physics simulation features WERE built into the main CPU, I don’t yet get the sense that this is the direction of either Intel or AMD. If the physics engine does take the market, then Apple has, by virtue of the platform, brought the physics engine card developers a little closer to the Apple platform, without making them develop dual product pipelines (though by the nature of a drop in card, it might not have been much extra work anyway).


As to your question, “What makes Intel chips so bad?” Well, I too use AMD chips on my systems, and the comparisons I’ve made with Intel based machines makes it clear to me that Intel isn’t the speed king at present. Intel and Amd have leapfrogged each other over the years, but AMD has made rapid comebacks.

Intel based their main thrust on raw clock speed, and sacrificed circuit complexity to do it. Generally, the more complex the circuit, the tougher it is to ramp it’s raw speed upward. AMD realized that gains in IPC, even at the expense of clock speed, would provide more dramatic gains when clock speeds can be raised, and they knew other “physics” were as much at the heart of that issue than simply compromising early rather than later. Intel was short sighted in this manner.

Hyper threading hasn’t been much of a gain either, though in theory it seemed a good idea, real dual core is by far more productive. AMD recognized this as well.

PowerPC, by nature of it’s RISC roots, pushed IPC over clock speed too. There are compromises in all designs, despite the advertising to the contrary, because physics, electronics, logic and the difficulties of mass production impose them. Intel simply didn’t make great choices on those compromises. Still, I wouldn’t agree to the extent that they’re chips are garbage, or seriously bad.

I also don’t quite agree that there’s any significant mangling of the stream of code and logic moving through the processor. It certainly isn’t as convoluted as the Windows operating system, which certainly is a mangled mess to a great degree. Windows 95 and it’s cousins were the worst at this, and while NT is vastly superior to Window 95/98/ME, it still bears the marks of legacy which give it warts everywhere. Still, there’s a decent potential in there if developers set aside most of the bad, and it’s no excuse for developing bad applications for it. Photoshop on the PC is one example – it works very well on my system, as do 3DS Max, Steinberg’s Wavelab, and a range of other applications. Adobe’s Premiere, however, can’t match Final Cut – but that’s not the chip or the OS at fault (well, Adobe might have had some complications from the OS, but it still isn’t a good excuse).

What is quite true, though, is that Intel’s offerings aren’t the fastest, and that does bring to puzzle just why Jobs went Intel only. I’d like to learn if his contract with them has the “gotcha’s” that Dell has (which is apparently why they’re all Intel to date).

A lot of people here have attributed behavior and fault of unreliable operation to Intel. This simply isn’t a factual representation. That’s strictly in the realm of motherboard/chipset/driver and occasionally at fault when integrators (like Dell) toss in marginal power supplies, drives or other peripherals. Of all the P4 1.4 through 2.5Ghz Dell’s I’ve known (dozens in various offices, for example), they all failed miserably because of bad disk drives (IBM Deskstars), bad motherboards, or something similar.

Now, to the contrary – I know a manufacturing plant that had, until about a year ago, an IBM PC XT in use since 1981. Yep, darn thing ran every day, 12 hours a day, for 23+ years! They also have an AT (original 6Mhz build) which is still running, from 1986 (or is that 87?). My own, old (old) dual P2-333Mhz box is still running in it’s present home, which I original bought (as parts) in 1998. The AMD based 233, 300 and 450 (K6 variety) I bought in 1996 are still working. One served my son (when he was 3, now he’s 4 so I moved him up to a P2-450 from 1998). In other words, Apple isn’t the only company that makes a PC capable of running every day, all day (many of these machines are 24/7) for years.

My own Win2k development machine was last booted on May 9th, when I last updated it with Windows service updates. It’s not been shut down since then. I typically leave it running for weeks at a time, often with debugging sessions that are continued for days at a time.

I would expect OS X to be easier to keep reliable than Windows of any flavor, for years ongoing. My point is that nothing about Intel makes the machine inherently unreliable or flawed, nor is AMD any less capable in this regard.

As to the attitude about preference for the PowerPC with any of the adjectives like elegant, superior, more advanced…well – here’s my thought. I’ve heard that cussing in Italian is better than cussing in English. A character in the Matrix films preferring cussing in French for reasons that made me grin, but I speak neither language well. My Colombian wife manages a searing retort in Spanish now and then, which I thankfully don’t understand, but when I feel the need to berate my fellow man, I haven’t found a lack from the English language, though I admit I must not really know what I’m missing. A few words my wife taught me from Spanish colloquial insults sound impressive to me because they’re new. Once she mentioned, at a business party, how someone was “kissing arse” to gain favors, I looked at an acquaintance who’s ears burned a bit and said, “Well, it sounds prettier in Spanish, I’m sure.”

The fact is, CPU’s are all about turning a language of bits and bytes into function. They all use some kind of ALU, bit shifters, comparators and a host of other electronic gizmos which come to us from decades past (and a few from centuries past). Does it matter, really, if a cylinder head has a hemispherical top, or a pyramidal one? There may be minor performance differences in one technological approach or another, but what really makes a difference is when performance jumps by orders of magnitude. A mere 20 or 30% difference is hardly noticeable for most situations; double and triple (or that major 10K+ jump we’ve made since the old 6502) does make major, meaningful differences.

PowerPC is still used in some of IBM’s own machines, but they also have the RISC6000 design to consider. Since they’re competing for customers in an arena where Intel and now even AMD compete, we see things happening to IBM that have happened to HP, Sun and SGI. Sun’s own Sparc, an interesting concept in it’s day, is giving way to Opterons. HP has several Opteron servers, and IBM can’t find a way to compete in the price/performance space all that well.

I have no doubt that innovation in the CPU will move forward by bounds in the year to come. I really thought Transmeta had some great ideas, but fizzled out with licenses to the big guys, and got out entirely. Development in this industry is about genius and cash. John Cocke is gone, now. I don’t know where his counterpart is employed, but someone at AMD is no slouch, and I think there may be at least one still at Intel, just moved out of the Itanium group. The rest is up to the physics guys, and some of them are elsewhere (unknown to me).

HP has a bright idea that might make big news in a few years (a new idea to replace the transistor, shrink size to 1/10th of today’s best silicon, and run faster than anything we know at present). So there’s at least a few ideas in physics we aren’t yet even able to taste in product. Guess who’s got the most cash to license that gem?