[Hinson]

(Sorry for the length here, but I think I�m hitting on the real crux of a major concern that has also been noted by others. I wanted to spell out all my major thoughts on the subject.)

I�m looking for opinions on my opinion: Basically, I think there is one clear reason for concern when considering the Mac�s long-term price/performance outlook--that is, the lack of good ol� capitalistic competition! This, I believe, has always plagued the Mac, my long-standing platform of choice, but I think it is hard to imagine a future for the Mac that doesn�t specifically address this problem where CPU power is concerned.

What do I mean? Well, first, let�s just admit how foolish it is to argue that there�s not a very real potential for long-term price/performance deficits when comparing the probable roadmap for PowerPC improvements and those on the Wintel side of things. G4s will likely be improved, but it seems really difficult to quickly and drastically increase their clock speeds (from what we�ve seen). As for G5s, who knows when they will be coming along. On the other hand, Intel (and others) seem to continually be able to push their architecture to higher speeds, and regardless of how much GHz can be misleading, eventually we have to admit that they matter when the differences are large enough (e.g. when it gets to 4GHz Pentiums vs 1.4GHz G4s).

SO, the questions I want to address are these: What can cause this to be a continual, long-term problem and what possible way is there out of this?

I believe the problem is very obvious�on the Wintel side, you have plenty of manufacturers wanting the fastest possible chips and different chip makers competing for their attention--it�s plain and simple. There�s a strong motivation for continual improvements to chip performance. On the Mac side, we have Motorola and IBM, neither of whom makes its business from sales chips to Apple (not to mention that only one of them owns the rights to Altivec). Sells of PowerPC CPUs to the one real customer out there, Apple, do not provide a strong incentive for either of these companies (given their other major interests) to dramatically push to increase performance for those CPUs. Obviously, there�s a good chance that as time goes on, the Wintel-side of the coin will continue to see strong improvements in whatever chips are used in their platforms while PowerPC price/performance numbers will continue to lag behind (with the gap getting wider and wider).

The argument is based purely on the concept of economic motivation, but I believe its valid. The obvious question, then is as follows: How does Apple improve the relative price/performance numbers for the Mac if they have no way to cause real competition on the PowerPC side while there will continue to be great competition on the Wintel side? It�s hard for me to imagine an answer to that without eventually considering abandoning the PowerPC (which has perhaps been discussed to death, but...). Again, I argue this based solely on the economics involved, but it seems a sound argument. IF the Mac today ran on a motherboard that, though proprietary enough to keep the MacOS and Apple hardware inexorably linked, none-the-less relied on Intel�based CPUs for its raw power, then Apple would be just another competitive buyer--just another incentive to push forward the performance of those chips.

Obviously the major technological problem in such a move is that of migration. BUT the REAL PR problem is that, while it�s hard (for me, at least) to imagine another way for Apple to deal with the long-term problem (because it really is one of economics beyond Apple�s control), we haven�t heard anything from Apple on what plans it has to deal with the problem. Should Apple announce a long-term strategy to move to intel-based chips (baring some major improvements to the long-term PowerPC roadmap)? Even if such a move didn�t prompt Moto and/or IBM to push forward more with PowerPCs, perhaps it would start preparing us, the Mac faithful (not to mention programmers) for what seems to be inevitable.

Basically, one way or the other, I want Apple to acknowledge the obvious long-term problem it faces and give us a real plan for how it specifically can address it (regardless of what Moto or IBM do). That�s the type of plan that�s needed and lacking! If there�s another real option to the obvious benefits of competition in the Intel world, I want to hear about it! How long will we be stuck in a quagmire simply because, as wonderful as the PowerPC is, there is no great economics-based factors pushing it forward (especially compared to the Intel world).

Disclaimer: No, I�m not considering the horrors of switching platforms over this--I am a Mac fanatic, and though it might sound overly self-important, I�ve always believed that if I ever found reason to leave the Mac, it would have to be due to factors that truly and obviously spelled the final downfall of the platform. I just want some piece of mind that there�s a way out of the picture I�ve painted above--that Apple really does have a plan that shows its understanding of the long-term dilemma created solely by overwhelming economics-based factors.

Today it seems the strategy has become down-playing the overall price/performance issue and playing up all the nice whiz-bang apps that come with the Mac (and the high-price-or-nothing strategy of .Mac has harmed that point, IMHO). But that�s not a long-term strategy that will work. I want to know the long-term strategy(ies) Apple has in mind.

Wow, this was longer than I thought it would be. Looking forward to responses, intelligent discussions, rants, �speed doesn�t matter and never will� arguments, etc.

-Jay

[Olorin]

Apple will have some more compeditive hardware later this month. Don't worry so much Steve has it all figured out. Apple needs to leave the G4 soon but not the PowerPC. IBM has some good chips and they are working on even better ones. They would be happy to sell to apple. It will take some time but Steve is working on it. Within the next few years we will see a complete turn around ( more than we already have) in Apples hardware. Good times are ahead ( the economy is slowing things down). In time.

[mac freak]

Amen to that!

I totally agree, and the only way out that I see is, in fact, for Apple to consider using more mainstream CPU technology (as you pointed out).

However, there is one more option you haven't pointed out: Apple could buy out the entire PowerPC program, and develop new chips themselves. This has been discussed before, and IMHO would certainly be more productive than the current scheme (relying a cellular phone manufacturer). Apple would, of course, need to hire new people for the development, but they'd no longer have the no-competition issue, as Apple is always competing with the likes of Dell/Gateway/etc. to have the best-overall systems (processor speed, of course, is a factor). Motorola & IBM hardly give a sh!t about Apple (at best, a TINY one... lol).

At any rate, the performance deficit today is not much worse than it ever was, and with new Power Macs a few weeks away, we can at least watch it shrink for a while!

[dilleet]

I want state of the art technology and the PowerPC consortium is not providing it and rumours, roadmaps & Mac dreams will never allow Moto & IBM to catch up with AMD & Intel. I have no problem using Intel chips if they are running a Mac OS.
Whoever suggested Apple take over chip production hasn't a clue what a foundry costs, or even sub contracting the production out. Its the OS and the Mac community I love, all of the other components are standardized now, ie: HDs, Graphic cards, memory & peripherals. Just give me a Mac with an AMD or Intel Mobo and I will be beaming and Apple can stop pleading with IBM & Motorola to waste their time and money on their non-profitable CPU development for our platform

[Ken_F2]

On the other hand, Intel (and others) seem to continually be able to push their architecture to higher speeds, and regardless of how much GHz can be misleading, eventually we have to admit that they matter when the differences are large enough (e.g. when it gets to 4GHz Pentiums vs 1.4GHz G4s).

Before I address the issue, I want to point out that the situation is not quite as bad as you make it. We'll not be a situation of a 4GHz P4 vs 1.4GHz G4. Motorola is still manufacturing silicon on the old .18um process that Intel and AMD have abandoned, which is why they are lagging so bad right now. As soon as Motorola moves to the smaller .13um process (and they're in the process of doing so, but it takes time), we'll see a boost to 1.4GHz, with scaling to at least 1.6GHz, and possibly 1.8GHz or 2.0GHz. I expect we'll be at 1.5GHz G4 on this process by Macworld in January. So we'll be looking at dual 1.5GHz G4 systems vs. P4 3.0GHz (or in worst case, 3.2GHz) systems in late January.

The P4 isn't expected to hit 4.0GHz until the end of next year, at least 14-15 months away. Motorola will still be on the .13um process at that time (while Intel is on the newer 0.09um), but they should be producing G4 processors at >1.6GHz, and more likely a G5 at 1.6GHz to 2.0GHz. Apple will no doubt continue it's dual configurations. At that point, we'll probably find ourselves in another MHz rut while we wait...and wait..and wait...for Motorola to move to the .09um process. Worst case scenario is a 2.0GHz G5 vs. an improved 6.0GHz P4, but that's a long time off (late 2005 - early 2006).

I believe the problem is very obvious�on the Wintel side, you have plenty of manufacturers wanting the fastest possible chips and different chip makers competing for their attention--it�s plain and simple.

That's certainly true. PC users have the Intel vs. AMD war to thank for their fast and cheap systems. Were it not for AMD, Intel would probably be pushing 2.0GHz now (if even that) rather than 3.0GHz in October. And AMD has announced a roadmap with a 4000+ "Hammer" Athlon slated for the second half of next year, so Intel will surely do everything it can to match/exceed that.

Obviously the major technological problem in such a move is that of migration. BUT the REAL PR problem is that, while it�s hard (for me, at least) to imagine another way for Apple to deal with the long-term problem (because it really is one of economics beyond Apple�s control), we haven�t heard anything from Apple on what plans it has to deal with the problem. Should Apple announce a long-term strategy to move to intel-based chips (baring some major improvements to the long-term PowerPC roadmap)?

The best thing for Apple to do is exactly what it's doing now. Encourage developers to abandon legacy OS9 code and move to 100% native Cocoa for OSX. At the same time, continue to internally develop OSX for x86 and Cocoa development/porting tools for x86. There's no doubt Apple is doing this right now. Here's how I see it playing out:

In the short term, Apple will continue with the PowerPC. Motorola will soon be moving to .13um, which will provide a much needed boost for the PowerPC. For 2003, the gap (in % terms) should not get any worse; the % gap should remain the same or slightly improve (decrease) while Motorola is able to churn out faster processors on the newer .13um process. Through much of 2004, Apple should be just fine with the PowerPC.

Also important it that Microsoft does not have anything major planned in 2003 or 2004, meaning there will not be some new "killer app" functionality that will be possible on faster PCs, and not possible on Macs (other than a few games). That's the real issue here. So long as Apple can do the same basic things as Windows PCs with the processors they have, then there is really no significant pressure to abandon the PowerPC.

Things will change in 2005. First, Intel and AMD will move to the a .05 or 0.06 process. As soon as they do that, that performace gap will widen, and widen FAST. Motorola will have just moved to 0.09 (or be in the process of moving from 0.13 to 0.09). For every 100MHz increase on the PowerPC, AMD/Intel will move 400MHz.

Moreover, Microsoft will unveil its next-generation "Longhorn" Windows (recently delayed from early 2004 to early 2005). Microsoft claims that this version of Windows will be a far bigger step than Windows95/98 -> Windows XP. Microsoft has said it will bear little resemblence to today's Windows. Everything will be redone. The minimum hardware required could include a 3.0GHz P4 and 8X AGP 3D card (at the time, Intel and AMD will be pushing close to 6.0GHz, and will begin moving toward 8+GHz fast with their new 0.05m process). Microsoft is working on this OS right now with >2000 developers (and likely 5000+ on project next year), and it is still 2.5 years away. [In the interim, MS has two new versions of XP planned, a la 98->ME type releases for XP which most people could care less about.]

I see Apple formerly announcing the switch in January 2005, with OSX II. By that time, all popular Mac applications will be written in Cocoa. Apple will have perfected x86 versions of its development and porting software. Key developers will have had these porting tools (under NDA) for some time, and versions of all major applications will be announced for the new platform. Apple will announce/demonstrate a significantly enhanced OS that will make use of the new faster ~6.0GHz hardware (with single and dual configs available). Apple will show off new killer apps that require this sort of horsepower. The OS and apps will look like such a substantial upgrade that everyone will want to upgrade.

The new OSXII will do everything (and more) as Microsoft's next-generation Windows, but will be released several months beforehand.

[rampant]

I think you're optomistic given history.

[BZ]

Let me first go on record by saying I know very little (mostly read from these and other forums). What I have put together is that the G4 still has life in it if it is given the right RAM/Bus speed to talk on.

For example: Take the Xserve. In all the tests I have seen a 2x1Ghz Xserve just destroys a 2x1GHz PM. The only thing different is the Northbridge bus, DDR Ram (266Mhz) and the drives.

<img src="http://www.xinet.com/images/benchmark.graphs/2002/xserve.reads.gif" alt=" - " />
<img src="http://www.xinet.com/images/benchmark.graphs/2002/macosx.reads.gif" alt=" - " />

<a href="http://www.xinet.com/benchmarks/benchmarks.2002/" target="_blank">http://www.xinet.com/benchmarks/benchmarks.2002/</a>

Now take a look at the current BareFeats benchmarks that show a Dual 1Ghz PM keeping up with a Athlon MP.
<a href="http://www.barefeats.com/pentium4.html" target="_blank">http://www.barefeats.com/pentium4.html</a>

Now, if this is on 6 month old hardware, where we know Apple has a G4 bump, DDR ram and a new mobo/bus I think the next PM release will do very well. Combine 1.4Ghz G4s, 266 DDR Ram, 1 Ghz Northbridge bus, 10.2 all of which I think will get us about a 100% increase in speed (measurable).

Not bad for six months. Now lets see what they can pull out of IBM for next year?

BZ

[OverclockedHomoSapien]

The key is that Apple is ditching Moto for their high end chips. IBM is going to be Apple's new supplier, and unlike Moto, IBM does NOT have difficulties in fabbing and scaling their chps. They are foundry gods in every sense of the word.

Two possibilities for the future:

1. IBM produces a scaled-down version of their Power 5 chip which is being designed for workstations.

2. Apple designs the "G5" in house and has IBM fab it.

Either way, Apple comes out with chainsaws roaring an axes swooshing. The Power 4, at only 1 GHz, is CONSIDERABLY faster than any Pentium available, and IBM is only going to continue to scale it.

[CodeWarrior]

</font><blockquote><font size="1" face="Geneva, Verdana, Arial, sans-serif">quote:</font><hr /><font size="1" face="Geneva, Verdana, Arial, sans-serif">Originally posted by OverclockedHomoSapien:
<strong>Two possibilities for the future:

1. IBM produces a scaled-down version of their Power 5 chip which is being designed for workstations.

2. Apple designs the "G5" in house and has IBM fab it.

Either way, Apple comes out with chainsaws roaring an axes swooshing. The Power 4, at only 1 GHz, is CONSIDERABLY faster than any Pentium available, and IBM is only going to continue to scale it.</strong></font><hr /></blockquote><font size="1" face="Geneva, Verdana, Arial, sans-serif">From what I've read, the POWER5 is aimed at the 1U servers AND workstation markets so I don't think they will have to scale it down to fit in an Apple PowerMac case especially if it is the new rumored one with lots of cooling. Either way, I want Apple to use a chip that is designed for workstation markets and not embedded controller markets. The only question remaining is what to do about SMID - Velocity Engine (aka Altivec). It would need to work with legacy stuff so companies wouldn't have to recode everything.

[MindFad]

BZ's got it right. The G4 is not a slow chip, and has plenty of life in it. Damn Apple for not adopting DDR SDRAM and faster buses very long ago. A dual 1-GHz G4 performing a few *seconds less* in Photoshop (at a few tests) than a 2.4-GHz P4 doesn't scare me -- you know things will improve greatly when Apple decides not to castrate the hardware with the current bus and SDR SDRAM. The performance is what I care about, and though it isn't great, I don't think it's terrible, even if I do pay a thousand or more extra for the Mac. Of course, this is a business, and we will probably still buy our Macs anyway, you know. And Apple's just fine with that. Apple's just hitting another rough time right now -- the hardware will get back up there. It better, damn it.

[Ken_F2]

Either way, Apple comes out with chainsaws roaring an axes swooshing. The Power 4, at only 1 GHz, is CONSIDERABLY faster than any Pentium available, and IBM is only going to continue to scale it.

Ehh hehe not quite...



Even with 1.4Mb of L2 cache and 128Mb of L3 cache, the 1.3GHz POWER4 still manages to be outperformed by the latest P4 in integer ops (which constitutes 90% of the work people do with Macs and PCs). The 1.0GHz POWER4 with 128Mb L3 cache offers considerably less performance, comparable to the P3 1.4GHz (results for POWER4 1GHz just released so yet reflected in above graph).

Nodoubt, POWER4 and Itanium2 do excel in floating point, but most consumers aren't simulating nuclear explosions, weather patterns, and the like. Practical application here is 3D rendering and games.

From what I've read, the POWER5 is aimed at the 1U servers AND workstation markets so I don't think they will have to scale it down to fit in an Apple PowerMac case especially if it is the new rumored one with lots of cooling.

I haven't read that. If you ever come across a link/url, please post.

As of a month or two ago, the POWER5 was to debut in the second half of 2004 in a new high-end box called "Armada," which is to replace the current "Regatta" series boxes (also known as p690, which starts at about $500,000). The POWER5 is supposed to feature both SMT (symmetrical multithreading, like the P4) and on-chip enhancements for TCPIP and possibly encryption.

And of course, this still doesn't change the fact that POWER4 and POWER5 are not PowerPC processors, and will not run OSX, OS9, or any current Mac applications. Once again, IBM moved away from PowerPC architecture with the introduction of the POWER3 some years ago.

[Hinson]

Nice to see some optimism and various points being made. There seems to be some agreement out there that a switch to a CPU used by the vast majority of the PC industry may be an obviously needed long-term goal. For those that disagree and argue otherwise, let me note a couple of points:

Concerning IBM stepping up and improving the PowerPC--I still don't see how the incentive for them to push forward continually with PowerPC speeds matches that on the Intel side in the long term. I get that they are probably better for the long term than Moto (because of the focuses of each) but the real power of major competition still just isn't there given the how either company views its sell of PowerPC chips to Apple in the large scheme of things. Am I so wrong that the obvious incentive-advantage on the Intel side (to push forward their PC CPUs) really doesn't overpower the long term incentives for IBM to push the PowerPC forward?

As for Apple taking over the chip-making or designing, I don't quite see it. I am NO expert here, but it seems to me that a major chip maker has to reap a lot of benefits to make up the cost of designing, manufacturing, and testing a chip--more so than might be made back by the single chip (which, I'm just guessing here, might just be one of a variety of chips that they make that benefit from the whole design and manufacture process), Again, I could be wrong, but I just imagine that its not really cost effective for Apple.

Finally, I know that though there is an obvious gap that has developed in price/performance, some "coming soon" Macs will help lower the gap, and that future bumps may help, in the short term, keep the gap from growing at the rate I might have suggested, but I still think that in the long term (in computer time), Apple HAS to go with industry standard chips (to benefit from the competition, etc.--just like it's benefited from other standards it's embraced). Again, I'm not making an argument about the technology, just from the standpoint of capitalistic economic incentives.

If there are other thoughts, Id the replies.


-Jason

[suhail]

Just say NO to Intel!!
You think OS-X is going to work properly if Apple brings it out for Intel. We will have a TWO year lag behind Windows until Apple squashes the major Bugs. Even if Apple has a running version of OS-X, at this minute in their labs, and their tests show minor problems, once you put it in the publics' hands� behold, major problems.

Anyway Intel and AMD have many, MANY, transistors that are useless to Mac OS-X and 9. Why do you think these processors are so huge, and consume so much power, while the G4 is so much smaller and cooler? Is it because Motorola has smarter people working for them? No. Its because the processor was not made for Windoofus. Try putting the desktop vesion of AMD or Intel in a notebook, the battery will run-out while the computer is still booting in DOS!

Apple should have customized versions of AMD with Altivec chips for OS-X. And I'll bet, that the Mac chips will finish production first.

If AMD believes that Apple is going to dramatically increase it's market share, like I do, they should jump in NOW!

[TNproud2b]

How many Apple users even know anything about the architecture of their CPU?

I'll say .025% of them do.

What if ALL current G4s were actually Pentium3s ??

Who would know unless they looked?

Apple isn't concerned AT ALL about what users will think if they switch to x86. Ain't nobody gonna notice except the .025%.

[driven]

Originally posted by MindFad
BZ's got it right. The G4 is not a slow chip, and has plenty of life in it. Damn Apple for not adopting DDR SDRAM and faster buses very long ago. A dual 1-GHz G4 performing a few *seconds less* in Photoshop (at a few tests) than a 2.4-GHz P4 doesn't scare me -- you know things will improve greatly when Apple decides not to castrate the hardware with the current bus and SDR SDRAM. The performance is what I care about, and though it isn't great, I don't think it's terrible, even if I do pay a thousand or more extra for the Mac. Of course, this is a business, and we will probably still buy our Macs anyway, you know. And Apple's just fine with that. Apple's just hitting another rough time right now -- the hardware will get back up there. It better, damn it.

Not everyone has adopted DDR on the PC side either.

I have a 2.2 Ghz P4 on my desk at my office and it runs SDRAM, NOT DDR. Go figure.

[OreoCookie]

Although I can see Intel cranking up the GHz, they have also pointed out what their competitors have been trying to all the time. GHz is not everything.

Oh no, I am not one of those Mac fans. Not at all.

The trouble is the heat dissipation and especially the heat dissipation per square centimeter. It continues to rise and rise so that it will be physically impossible to cool a CPU with reasonable technology (forget about the -40 �C �freezers' here).

The Itanic and Itanic 2 are clocked around 800 MHz and 1 GHz, respectively.
These clock speeds will unlikely rise as fast as the P4s, because these chips were designed with another goal in mind.

IMHO, Intel will have a little bit of a problem when introducing IA-64 for mainstream computing (if it ever will).

My point is that future chips will slow down when it comes to the clock speed, but feature different improvements that still speed up the system considerably.
E. g. the Hammer is believed to be introduced at (real) 1,8 GHz. Yet the performance is increased considerably compared to any PC CPU.

Bringing this longer post to the point: No matter what CPU Apple will use in future system, the development of CPUs will certainly go in a different direction. NEC is producing vector chips at ridiculous 125 MHz that consume about 1 W of power. Their top speed is 1 GFlops. Many of them will be the base for their future vector supercomputers.

Another interesting thing is that whenever Apple will make the shift to a 64-bit CPU, then there is no established 64-bit technology on the market right now.

Intel Itanic or successor: Pros: Fast and new CPU. Radically new design with no compatibility compromises. Cons: Very expensive. Large executables. People not used to Itanic code.

AMD pushes the Hammers that are IA-32 CPUs with a 64-bit enhancement. Pros: compatible to all existing code with no performance penalty. Cons: Not established and still features the many drawbacks of x86 CPUs (very few registers, etc.).

Sun's SPARCs: Pro: relatively old and mature CPUs; Sun has the second highest R&D budget of any CPU manufacturer! Cons: SPARCs have never been the performance leaders (at least in terms of SPEC marks). Image of slower (relative to competitors) but extremely reliable machines.

IBM: Pros: Power4 is the fastest CPU on the market (Itanium 2 is supposed to be faster, but not yet available). Advances CPU design. Still a PPC derivative, i. e. probably the easiest port for OS X's upper layers due to accustomed code. Cons: Very expensive in its current incarnation. Availability?

If I were Apple I wouldn't be too sure what to do.

[Ken_F2]

Anyway Intel and AMD have many, MANY, transistors that are useless to Mac OS-X and 9. Why do you think these processors are so huge, and consume so much power, while the G4 is so much smaller and cooler?

You have no clue what you are talking about.

The current Motorola G4 processors (74xx) are 106mm^2 large and use 33 million transistors. The AMD Athlon is 84mm^2 in size with 37 million transistors. The P4 2.0GHz is 42 million transistors; the P4 die is currently 131mm^2 (and that includes twice as much L2 cache as the G4). The P4 die is larger than you would expect, given its transistors, because of the design necessary in order to get such high frequencies across the chip. The G4 is manufactured on the older .18um process.

In summary, the Motorola G4 is 26% larger than the AMD Athlon and 23% smaller than the latest Intel P4. The Motorola G4 has 8% less transistors than the AMD Athlon, and 27% fewer transistors than the Intel P4 (which has twice the cache of the G4). The Athlon and P4 use a newer manufacturing process, so they can fit more transistors in a smaller die area.

You want to see a big processor? Take a look at the IBM POWER4. The POWER4 is a whopping 170 million transistors, or more than five times as many transistors as the Athlon, and four times as many as the P4. Thought the P4 was hot? At 1.2GHz, the IBM POWER4 dissipates 125W of heat at 1.5V, which is more than twice as much heat as a P4 at 2.80GHz, and four times as much heat as the Motorola G4 1GHz. In terms of size, the smallest IBM POWER4 processor is many times the size of a P4 2.8GHz.

Try putting the desktop vesion of AMD or Intel in a notebook, the battery will run-out while the computer is still booting in DOS!

The latest PC notebooks get 2.5hrs on a battery, or ~5.0 hours with two batteries. The notebook screen consumes far more power than the processor.

If AMD believes that Apple is going to dramatically increase it's market share, like I do, they should jump in NOW!

Remember, Motorola and Apple have far greater resources (with exception to engineering talent) than AMD. AMD is currently using all the resources it has to compete with Intel. Let's not forget, AMD is currently engaged in a chip performance and price war with Intel. All their resources right now are dedicated to getting the next-generation "Hammer" Athlon out on time. In case you haven't been paying attention, AMD is burning through serious cash (losing money) right now.

[suhail]

You said exactly what I was saying, The P4 and AMD have way more transistors than the G4. But they are useless to OS-X and 9.

Also the Intel processor inside a PC notebook is called a Pentium Modility, it is slower than its desktop counterpart. Moreover, PC notebooks are having a hard time keeping-up with the PB-G4 battery life.

Oh... and the G4 has a bigger screen

By the way, the G4 notebook is faster than most Pentium notebooks. Hence, slower Mobility Pentium.

The G4 is soooo much better than a Mobility Pentium in notebooks. The main reason is� see my first point.

[Ken_F2]

suhail,

Your first point doesn't really apply if there is a version of OSX for x86. And I wouldn't call 8% more transistors "many more;" the next PowerPC from Motorola will probably overtake the Athlon in terms of transistor count.

Also the Intel processor inside a PC notebook is called a Pentium Modility, it is slower than its desktop counterpart. Moreover, PC notebooks are having a hard time keeping-up with the PB-G4 battery life.

No doubt, the current mobile chips from Intel and AMD are just hacks. They are desktop processors with a few power management features tacked on. Meanwhile, the slower G4 processors were designed for embedded applications, so they tend to consume less power. However, the top end PowerPC processors used in Macs are desktop processors; they are no more efficient in terms of power and heat, MHz-for-MHz, than the P4 and Athlon. Consider that the G4 1.0GHz is spec'd to dissipate a maximum of 30 watts of heat, while the P4 at 2.5GHz (more than twice the frequency) dissipates around 50 watts of heat (less than twice as much).

Intel's first designed-from-the-ground-up notebook processor, codenamed "Banias," is coming out next January. We shall see how much performance Intel gives up for a truly mobile processor. As far as screens, Dell offers >=15" screens on many of their notebooks, and several PC vendors have announced notebooks based on new 16" screens.

[driven]

Originally posted by Ken_F2:
suhail,

Your first point doesn't really apply if there is a version of OSX for x86. And I wouldn't call 8% more transistors "many more;" the next PowerPC from Motorola will probably overtake the Athlon in terms of transistor count.
No doubt, the current mobile chips from Intel and AMD are just hacks. They are desktop processors with a few power management features tacked on. Meanwhile, the slower G4 processors were designed for embedded applications, so they tend to consume less power. However, the top end PowerPC processors used in Macs are desktop processors; they are no more efficient in terms of power and heat, MHz-for-MHz, than the P4 and Athlon. Consider that the G4 1.0GHz is spec'd to dissipate a maximum of 30 watts of heat, while the P4 at 2.5GHz (more than twice the frequency) dissipates around 50 watts of heat (less than twice as much).

Intel's first designed-from-the-ground-up notebook processor, codenamed "Banias," is coming out next January. We shall see how much performance Intel gives up for a truly mobile processor. As far as screens, Dell offers >=15" screens on many of their notebooks, and several PC vendors have announced notebooks based on new 16" screens.

In fact Sony actually has a notebook with a 16" screen running 1600x1200.

I doubt I'd have room to open it in the steerage seats on US Airliners but it would be nice to have on my desktop.
(Now if I could just get it to run OS X!)
<GRIN>

[suhail]

Ken,
42 Million for a P4, and 33 Million for a G4, this info was from you own post.

Well� here in Washington, that info comes to 21.4%
9 Million transistors more!!

Pentium and AMD still have win32 transistors which are useless to OSX and 9.

Maybe Apple could have another chip manufacturer continue the G series development.

[istallion]

"The trouble is the heat dissipation and especially the heat dissipation per square centimeter. It continues to rise and rise so that it will be physically impossible to cool a CPU with reasonable technology (forget about the -40 �C �freezers' here)."

Eventually this problem will create new challenges for cooling, but there are already methods to help overcome it. Shuttle's small form factor cases use only a mini fan in the power supply and 1 fan on the heat pipe radiator, and it can run a 2.5Ghz p4 almost silently. Friday, Isonics announced that a wafer foundry had been certified to produce its isotopically purified silicon to AMD. This silicon has thermal transfer capabilities 60% better than natural silicon. Supposedly prototype Athlons @ 1.5Ghz that had this silicon ran with passive cooling.

[TNproud2b]

Sushail, based on that logic, the G4 needs more transistors. It's getting its ass handed to it by every x86 CPU made in the last 3 years.

[cowerd]

Pentium and AMD still have win32 transistors which are useless to OSX and 9

What's a win32 transistor? Only one logic state, i.e. always off?

[suhail]

OK, Let's say Apple brings out a P4 PowerMac...

Ofcourse, this would mean that Windows software could run on the new Mac without much pre-processing, just like Classic on OSX.

And, Connectix would, most probably, make a version of their VPC that would run just like classic runs on OSX; where the MS-Windows Apps would show in the dock without having Windows in a seperate window.

So in this case, WHY would application developers bother to dish-out big sacks of cash to develop an OSX version of their application, when their Windows version already runs on the new Mac?

[OreoCookie]

Originally posted by istallion:
"The trouble is the heat dissipation and especially the heat dissipation per square centimeter. It continues to rise and rise so that it will be physically impossible to cool a CPU with reasonable technology (forget about the -40 �C �freezers' here)."

Eventually this problem will create new challenges for cooling, but there are already methods to help overcome it. Shuttle's small form factor cases use only a mini fan in the power supply and 1 fan on the heat pipe radiator, and it can run a 2.5Ghz p4 almost silently. Friday, Isonics announced that a wafer foundry had been certified to produce its isotopically purified silicon to AMD. This silicon has thermal transfer capabilities 60% better than natural silicon. Supposedly prototype Athlons @ 1.5Ghz that had this silicon ran with passive cooling.

Yes, but these methods of cooling are rather expensive.
There are physical limitations as well -- it isn't so much the amount of heat anymore, but the �heat density' (heat per area) that is critical. AMD *requires* copper base plates in all new coolers for their next-gen CPUs. Until now they have been used by overclockers.

That's what I'm talking about.

Sooner or later this will be a determining factor of CPU development. 100 W is still feasible, 125 W of heat dissipation is a major problem (you need a high quality case and a good ventilation, good bye all ya home-brewed PC builders).

[Superchicken]

Cause apple will cut off conecke something like that!

As far as chips, I'm running a 333Mhz chip. I'm not really all that concerned with speed, this computer's gona do me for another year too!
after that I'll buy a new iMac and iBook (which hopefully will have a G4 in it) and then I'll wait for that to get outdated in comparison.
My current iMac isn't to bad either, it's a nice ocmputer that runs great!
I have no idea why so many people go psyco about this junk.
Yes AMD's releaseing the hammer who cares?
IT doesn't run Mac OS does it?
Well that's what I like to use.

And as far as most Mac users not knowing their chips. I htink most know their Mhz, and most know somethin about their comps, Mac users are more tech savvy than PC users in general.

But from what I understand all the developers would have to code their new apps to run on x86 proccessors.
Which would be really bad for anyone who bought a G4 or something recently, and new iMac owners would be furriated!

[OreoCookie]

No matter what happens, if there is a CPU architecture transition such as one to x86 aka IA-32 or whatever (I have listed all candidates), then it will be no earlier than two years from now.

Why?

1. Apple has to complete the OS X transition.
2. Apple has to develop its own hardware around that (motherboard, etc.).

In two, three years a lot can happen. Apple buys Moto's CPU business, maybe a together with IBM, they could go to IBM and ask them to adjust their Power4/Power5 to their needs, they could use IA-32 -- we'll see what happens.

[bigv]

It has taken almost to years aready to make the transition from the classic OS to OSX. Even if Apple had a version of OSX for x86 right this minute ready to go the transition for adoption of the new platform software developers would still have to recompile their apps, hardware manufacturers completely rewrite their device drivers, and EVERY mac owner that wants to continue with the Mac will have to replace ALL of their software in addition to buying a new MAC. All of this just a couple of years after having to do replace their apps for OSX. In actual practice the transition would take much longer that just two years. What do you people think? Apple says '
now we have a x86 Mac with OSX available today' and everything just falls into place immediately and flawlessly? Look at how bumpy a ride it has been just migrating to OSX from OS9. A two to four year time frame I think is very optimistic for such a transition, not to mention I don't think current Mac users would put up with two major changes in as many years.

In other words I find it very highly unlikely that moving to x86 will happen at all. Much more likely is Apple finding another manufacturer to produce the PPC chips.

[suhail]

Right-on Bigv
I could not have said it better

Mac users went through so many unstabilities and transitions with the Mac-OS (Not to mention embarrasments like IE). Starting with Mr. unstable 8.0 all the way to Mr. no applications and very SLOW OS-X 10.0

Now that we are almost at 10.2 and have a reasonable amount of apps, we are discussing another transition!!

I'm not willing to go through another transition, and I've been a Mac user since 84, and an Apple II user since 1981.

I think Pentium is a stinky chip, and Apple should take the G series to another shop.

[nana4]

Originally posted by suhail:

Also the Intel processor inside a PC notebook is called a Pentium Modility, it is slower than its desktop counterpart. Moreover, PC notebooks are having a hard time keeping-up with the PB-G4 battery life.

Oh... and the G4 has a bigger screen

By the way, the G4 notebook is faster than most Pentium notebooks. Hence, slower Mobility Pentium.

The G4 is soooo much better than a Mobility Pentium in notebooks. The main reason is� see my first point.

The G4 does not have a bigger screen than most PC laptops. For example, a Dell Inspiron 8200 (or any other PC laptop) with 15" UXGA (1600*1200) screen has a total of 1.92 million pixels, and a resolution of 133 DPI.

The DVI G4 PowerBook has a 1280*854 pixel screen. Giving ~1.09 million pixels. I'm sure someone can tell me what the resolution is. Not even close. Yes it is 0.2" wider physically, but it is not as "tall". Just like a 50" 4:3 TV has far more viewable area than a 55" 16:9 TV.

P.S. Most laptops are now using the Pentium 4-M CPU, which is available at speeds of upto 2GHz. There is also the AthlonXP available at upto 1.53GHz. Both using DDR SDRAM, which provides for lower power consumption than your beloved PC133.

The Pentium 3 Mobility which I assume you are talking about, is not slower than it's corresponding desktop counterpart. It merely has added functionality in SpeedStep (allowing it to slow down when not required to save battery life). In fact the P3-M was the first to have 512KB L2 cache, the desktop versions came later. Of course the overall system performance is lower than a desktop, due to such things as slothy laptop hard drives.

The G4 does have low power consumption, but that is the side affect of low performance. I'm sure with screen brightness at full the G4 powerbook doesn't last 3.5 hours when playing a DVD. Perhaps 2 hours?

[RGray02]

This seems to have been well debated, but my two cents:

moving to x86 is ludicrous.

As it is, the x86 architecture wastes at least two pipeline cycles translating to micro-ops, which means that in a 9 stage pipeline of the athlon, 2 cycles are lost, hence 2/9ths of the entire execution cycle of a single x86 instruction is wasted!

That is actually the least of the concerns, when you consider that the original architecture specification has only 8(!) floating point registers, without register renaming theyd be toast! In addition, the pathetic FP performance is a result of the x87 addon in 1981, which was a hack to begin with.

Which is precisely why the G4, its measly 7 stage pipeline (which used to be 4!), castrated by a poor memory subsystem, and impaled on a tiny cache is still competitive with systems running on bleeding edge subsystems.

Seriously, going to x86 would be a step back. As far as I am concerned the people are attacking the wrong problem, its not the PPC platform, but rather the lack of deep pockets by anyone involved in the AIM consortium, and the stupid legal issues involved with IP. Face it, this is more of a court battle.

I blame apple directly for their hardware deficiencies. Moto doesnt design subsystem controllers, apple does. I think that resources have been pulled from everywhere to complete jaguar, once it ships, i expect hardware to jump back dramatically.

Calling for a platform switch would be ludicrous now, especially after Apple broke itself trying to optimize OSX for one architecture.

[Hinson]

Just a few points...

First, for those arguing (as I knew some would) that speed doesn't matter, note that in the LONG TERM, it very much does. If Mac speed gains lag behind PC speed gains for long enough, the Mac will be obsolete compared to the PC for running the latest whiz-bang programs. A computer platform will loose its market share over time if its speed doesn't keep up with others in the market. Like it or not, the "speed doesn't matter" argument just doesn't work in the long term. I have an older Mac (about 4 years), and its lack of speed is very important (because I know what's out there that can outrun it quite well). Further, because Mac speeds have been stalled for a while, I haven't been willing to upgrade, costing Apple my money. You just have to get past the "speed doesn't matter" mindset, because for the PLATFORM, in the long run, it very much does!

Second, for those arguing over why the architecture of the PowerPC is better than x86, you're missing my whole point. My point is that it's all about capitalism and economics! Unless the PowerPC becomes so much better than the Pentium that it ALONE draws people towards a platform built on the PowerPC, then the market will always favor faster improvements on the Intel/AMD side than on the PowerPC side. There's more market share to be had, more competition, more momentum, etc. Those forces ALONE have already overcome architecture arguments made in the early 90s (remember, the PowerPC was going to blow past the x86 simply based on its superior RISC technology--the CISC chip would soon fall way behind and never catch up???). That argument ignored the economics. The fact is that the forces behind pushing forward the Pentium side have been much stronger than those pushing forward the PowerPC side. It's hard to overcome that capitalistic strength, and so far the PowerPC has not shown signs that it will ever be able to do so.

Finally, I understand how much of a headache it will be to switch to x86 architecture (or whatever the rest of the world is running on a few years from now). But the alternative is to ignore the powers that REALLY push forward speed advances in the PC world and, in all likelihood, continue to loose ground in the price/performance arena. I once thought that by going with "better" technology, Apple could eventually be running on far superior machines, and that that could make the price/performance ratio far superior for Macs. However, I now believe that because the Mac market share is small and (obviously) Apple-only, there are no great economic incentives that will push forward whatever technology Apple decides to go with. The only way out is to go with what DOES have the economic forces behind it, and right now that's Intel and their competition.

It will be a long, hard road (assuming Apple takes it), but there are strong reasons (based on the reality of economics and market forces) for eventually making the switch. The real question is then how to do it in a way that keeps you in control of your hardware, keeps current users from having to spend large amounts of money in a short time to switch, and draws more people to your platform. I KNOW that will be very hard, but I believe the platform might very well become obsolete in the long run if it doesn�t move towards the architecture that has the capitalistic advantage of being "the standard" in the industry.

-Jay

[OreoCookie]

Told ya the little Power4 would be comin' ... for the good of IBM and Apple

[TNproud2b]

AMD Opteron will be here first, however.

And I've got money riding on Apple's adoption of x86.

[Ken_F2]

which means that in a 9 stage pipeline of the athlon, 2 cycles are lost, hence 2/9ths of the entire execution cycle of a single x86 instruction is wasted!

It sounds like you have no idea what you are talking about...

But in any event, pipelines will only grow with time, as such is necessary to reach higher clocks. The POWER4, upon which a new IBM desktop processor (possibly for Mac) will be based, has 14 pipelines. IBM's documents show 16-18 pipelines for future versions of the POWER architecture. These processors maintain IPC by being wide issue.

The next-generation "Hammer" Athlon improves performance-per-mhz (IPC) about 25%, despite the fact that it adds two more pipelines. The extra performance penalty of those two extra pipelines for branch misprediction is vastly overcome by a) significantly improved branch prediction, which was already at 95+%, and b) the reduction is memory latencies through an on-die controller. In the current Athlon, the processor is starved and operating with much of its logic idle most of the time.

That is actually the least of the concerns, when you consider that the original architecture specification has only 8(!) floating point registers, without register renaming theyd be toast! In addition, the pathetic FP performance is a result of the x87 addon in 1981, which was a hack to begin with.

Pathetic fp performance? What planet are you living on? You need a reality check.

Moreover, x87 isn't the future on the PC, SSE2 and other SIMD implementations are. All new apps use these SIMD units. The only reason anyone would ever need to use x87 is for 80-bit arithmetic, and even that may be addressed by future SIMD implementations.

Vendors like SUN are also moving this way with their future RISC processors, and so is IBM, judging from the news this past week.

Which is precisely why the G4, its measly 7 stage pipeline (which used to be 4!), castrated by a poor memory subsystem, and impaled on a tiny cache is still competitive with systems running on bleeding edge subsystems.

I would hardly call the G4 competitive. It offers approximately 2/5 to 1/3 the performance of the fastest processors from AMD and Intel under UNIX. When the G4 is competitive, it's only thanks to its SIMD unit (Altivec).

And as you noted, modern x86 processors have access to an abundance of registers through renaming. This still leaves something to be desired, but as you know doubt know, the next-generation AMD "Hammer" Athlon, coming in December, gets access to a whole slew of new registers thanks to its 64-bit extensions.

Seriously, going to x86 would be a step back. As far as I am concerned the people are attacking the wrong problem, its not the PPC platform, but rather the lack of deep pockets by anyone involved in the AIM consortium, and the stupid legal issues involved with IP. Face it, this is more of a court battle.

While Intel and AMD are used in 97% of computers and servers, the x86 and future PC platforms will always be at an advantage with R&D. What IP issues are you talking about?

Moto doesnt design subsystem controllers, apple does. I think that resources have been pulled from everywhere to complete jaguar, once it ships, i expect hardware to jump back dramatically.

The PowerPC G4 itself only supports truly pathetic 133MHz FSB. This is a processor limitation. Apple doesn't have any control over this. No matter how fast a memory controller Apple implements with its chipsets, no matter how fast the memory you stick in a G4 system, the processor will never see more than 1.064 Gb/s thanks to this design limitation. In contrast, the current Athlon has a system bus of 2.1Gb/s, the 333FSB Athlon coming in October will have a 3.2Gb/s bus, the current P4 has a 4.2GB/s system bus, and the next-generation P4 coming in the first half of next year will have a 5.3Gb/s bus.

Even when the processor can't accept any more information thanks to an obsolete system bus, faster memory can reduce latency and the like, but the benefits here are minimal; certainly, the XServe benchmarks show that this is very little improvement until you hit 20-50 simultaneous users. When people quote the XServe benchmarks, sometimes I wonder if they actually bother to read the graphs; how many people will be manipulating 20 to 50 different images in Photoshop at once?

[suhail]

BOTTOM LINE!

If Apple adopts Intel, software developers will abandon the Mac OS. WHY?? Because VPC will run Windows apps at close to 100% speed, just like Classic runs on OSX.

If you want Intel, go buy one, they are not hard to find.

[cube-dude]

Topic: general
Forum: Lounge

Thread moved