If no G5...then what?
Ok, this is an exercise in futility. I want to know what some of you AI gurus can brainstorm in the following scenario. Please, I don't want this to turn into another G5 thread. Instead, I want to explore Apple's future without the G5. For the sake of argument, let's assume there is no G5. Let's consider the following scenario:
There is no G5 PPC. Motorola will continue with G4 development in the short term, but has no commitments to desktop PPC for the long term. Furthermore, IBM does not have any PPC chips in development that use Altivec, and they have no plans to implement Altivec or anything compatible with Altivec. Of course IBM will continue "G3" development, and even now the G3 has scaled to an impressive 1 GHz! The mere thought of such power is enough to blow your hair back!
Given this future, what will Apple do?
They can continue to use the G4, but even the rosiest estimates have the G4 scaling to 1.5 GHz or so by MWSF at the very earliest. By Jan 2003, the Pentium 4 will be at 3.0-3.5 GHz.
I don't have an answer, I'm looking to those "in the know" (or just plain knowledgeable) for some morsels to chew on.
[ 04-03-2002: Message edited by: Junkyard Dawg ]</p>
There is no G5 PPC. Motorola will continue with G4 development in the short term, but has no commitments to desktop PPC for the long term. Furthermore, IBM does not have any PPC chips in development that use Altivec, and they have no plans to implement Altivec or anything compatible with Altivec. Of course IBM will continue "G3" development, and even now the G3 has scaled to an impressive 1 GHz! The mere thought of such power is enough to blow your hair back!
Given this future, what will Apple do?
They can continue to use the G4, but even the rosiest estimates have the G4 scaling to 1.5 GHz or so by MWSF at the very earliest. By Jan 2003, the Pentium 4 will be at 3.0-3.5 GHz.
I don't have an answer, I'm looking to those "in the know" (or just plain knowledgeable) for some morsels to chew on.
[ 04-03-2002: Message edited by: Junkyard Dawg ]</p>
Comments
Interestingly, the chip with the SIMD on that road map has, AFAIK, never been introduced by IBM, but the <a href="http://www-3.ibm.com/chips/products/powerpc/newsletter/oct2001/new-prod1.html"; target="_blank">Sahara of course has</a>. And Sahara includes the other features associated with that new SIMD/RapidIO chip, i.e., .13µ fab and low-k dielectric. So why have they talked about the Sahara, but not this new chip?
I actually think the G5 is closer than we think. I don't know if we'll first see it in the form of a G4 with new features or an actual G5. And remember that G5 doesn't necessarily mean the same as when they announced it after the G4 release, it simply means 5th Generation PPC. It's not a moniker attached to an abosulte design.
And if we go by your simple minded thinking of only MHz than Intel might as well give up on their Itanium since it's only going to be 800 or 900 MHz and the 64 bit G5 will be 1.5GHz+
Maybe it's time you learn to grade processor performance on more than just MHz, that's absolutely foolish. You don't judge a car engine on horsepower alone, do you? Do you take into account other things such as number of cylinders, head size, torque, efficiency, etc... CPU's are the same, MHz is becoming one of the least important measures in performance. I guess that's why AMD can kick Intel's butt when they are 20% slower in clock speed.
*If* Motorola has to sell the PPC assets in the summer it seems likely the new IBM/Sony/Toshiba alliance would be interested, especially for PS3 stuff.
This might be good or bad for Apple. Jobs needs to get Sony on-side which might be tricky as they both sell on style. Sony could feasably block IBM from selling to Apple in extreme cases.
Apple needs to move away from AltiVec but keep the speed the provide. IBMs SIMD engine works on all code not just the optimised bits so that might be the answer (there are few mentions of G4 at WWDC this year).
So for marketing reasons I think we'll see something called a G5 this year, but it will not be the mythical 8500. I think it will be either the IBM 1Ghz+ chip, the PPC7500 or an Apple chip designed 100% in-house.
Jobs said by the end of 2002 there would be no speed gap (can't find the damn quote) so I reckon they've got things planned we know nothing about.
One of the risks they're taking, to cut costs, is reducing the rate at which air is replaced within the fabs. This does indeed cut costs up front, but as Eskimo has said, it means their fabs are dirty. So they're essentially hoping that they'll be lucky enough that the dirt getting into their chips and equipment won't cost them more than they're saving. The technical term for this sort of tactic, I believe, is "Hail Mary."
<strong>
Maybe it's time you learn to grade processor performance on more than just MHz, that's absolutely foolish. You don't judge a car engine on horsepower alone, do you? Do you take into account other things such as number of cylinders, head size, torque, efficiency, etc... CPU's are the same, MHz is becoming one of the least important measures in performance. I guess that's why AMD can kick Intel's butt when they are 20% slower in clock speed.</strong><hr></blockquote>
I agree. I can put a 426 ci Hemi in a Hugo but I wouldn't want to drive it every day and going around corners and general manuevering would be tough.
As someone have said here, G5 refer only to a new generation of chip, i think your question is what happens if Motorola stop the developpement of new PPC chips for Apple.
<strong>Maybe it's time you learn to grade processor performance on more than just MHz, that's absolutely foolish. You don't judge a car engine on horsepower alone, do you? Do you take into account other things such as number of cylinders, head size, torque, efficiency, etc... CPU's are the same, MHz is becoming one of the least important measures in performance. I guess that's why AMD can kick Intel's butt when they are 20% slower in clock speed.</strong><hr></blockquote>
Nothing is more annoying than someone bringing up the Megahertz Myth myth on here.
Yes, AMD can kick Intel's butt at a 20% difference in clock speed. The question is what Motorola can do with a 50% difference in clock speed.
Yes, yes, workflow, total cost of ownership, efficiency of software, AltiVec, etc. etc. The point is, in general, Intel or AMD chips are now dramatically and demonstrably faster than the fastest which Motorola can offer. Arguing whether or not megahertz is a good way of expressing that difference completely ignores the issue.
To answer JD's question - I think you'd see an announcement from Apple that they were switching to AMD chips. The obvious problem is Carbon. I think we might see some kind of weird solution like hybrid PowerMacs with one AMD processor and one G4 (at least until nobody used Carbon apps anymore), or a really complicated on-the-fly emulator.
<strong>
Nothing is more annoying than someone bringing up the Megahertz Myth myth on here.
Yes, AMD can kick Intel's butt at a 20% difference in clock speed. The question is what Motorola can do with a 50% difference in clock speed.
Yes, yes, workflow, total cost of ownership, efficiency of software, AltiVec, etc. etc. The point is, in general, Intel or AMD chips are now dramatically and demonstrably faster than the fastest which Motorola can offer. Arguing whether or not megahertz is a good way of expressing that difference completely ignores the issue.
To answer JD's question - I think you'd see an announcement from Apple that they were switching to AMD chips. The obvious problem is Carbon. I think we might see some kind of weird solution like hybrid PowerMacs with one AMD processor and one G4 (at least until nobody used Carbon apps anymore), or a really complicated on-the-fly emulator.</strong><hr></blockquote>
if Apple move to the X86 chips or a different chip from the PPC chips they are dead, a new emulator would be a disaster. If it arrive i would leave the mac, dispite my great affection for this mervellous computer.
<strong>
if Apple move to the X86 chips or a different chip from the PPC chips they are dead, a new emulator would be a disaster. If it arrive i would leave the mac, dispite my great affection for this mervellous computer.</strong><hr></blockquote>
You'll note I didn't say it would be a good idea, merely that that's what I'm pretty sure they'd do. I think I remember reading that the Wallstreet was designed to accommodate a low-power Pentium if needed.
Thats what every loyal mac owner dreams of (including me) but is it possible, just say that the next generation of mac processor is called the G5, but from apple in house, with partnership with IBM, and apple have ditched motorola when the contract runs out, (at about July?).
Ok, so the G5 is said to be at speeds of 800Mhz to 2Ghz +, however we know now that it would be suicidal for apple to release the G5 at a clock frequency lower than any G4. So it would probably begin at around 1.4Ghz to 1.8Ghz.
Now there are two scenarios, the first one is that apple doesn't want to shoot itself in the foot by releasing super fast clock speed chips, say around 2.5Ghz, because it would hurt the G4 sales in iMacs and their reputation.
On the other hand, apple could release the G5 at clock speeds much as high as this, if they get suffecient yields, just to prove that macs are not slow, but then they would have to have 333Mhz or 400Mhz DDR for memory and an 800Mhz effective FSB if they want to be in league with the AMD hammer series that will be released Q1 2003.
Also, there is a prospect of co-processors that has crept (sp?) up recently, and I think this is a brilliant idea, as it would be innovative, and there would be another factor to speeding up mac as a back up if the processors aren't up to scratch. My only question is, would this just be like a second processor, and be a dual, or would it be specific like speeding up quartz, renderings without loss of memory bandwidth etc? Does anyone know how this would work?
If all these ideas were incoperated into a next generation mac, then wouldn't it be enough for the 3D market and broadcasing market, which apple has leaned its focus on?
There are parts inside a CPU that are more important than the MHZ, but the MHZ is an abstraction of the truth inside. I think Apple's software solutions will have to continue to provide the same layer of abstraction to keep the underlying MHZ gap out of the picture.
There are three possible sources for new chips that Apple will use: Motorola, IBM, and internally developed at Apple. Nobody else is in the PowerPC business, or even seems remotely interested in it. Any of those three could do the design and use an external fab -- Moto and Apple are likely to use this approach, while IBM seems to prefer their own excellent foundries. My guess is that Apple is pushing both Motorola and IBM to provide alternatives since the whole point of AIM originally was to provided multiple sources of processors for Apple machines. I'm a little doubtful we'll see an Apple designed PowerPC, but they likely have people consulting with both Moto & IBM.
One possibility which is likely (and mentioned above) is that future Apple processors will be specialized for Apple. An industry-wide trend in processor design is to have a modular scheme where application specific modules can be rapidly added to the core processor. Apple could be designing the add-on modules while Moto and/or IBM are designing the core, integrating the modules, and building the things. This may relate to Apple's membership in HyperTransport.org, their purchase of Raycer, etc. Their eventual goal will no doubt be some kind of "system on a chip", giving them a high degree of integration and specialized capabilities... which is a perfect match for Apple's traditional strength of delivering a tightly integrated hardware/software package. It could allow them to get much better product differentiation from the PC market where these things are slower in coming. The PC world has a harder time doing this sort of thing because it requires very close cooperation between the individual PC makers, Intel/AMD, and Microsoft.
Apple will not move away from AltiVec, and there is at least one WWDC session on the subject this year. Future PowerPC chips might have alternative implementations of the AltiVec instruction set, but it is critical that it be compatible. Apple had quite a bit to do with the original AltiVec instruction set design so I'm sure that they either have at least partial ownership, or they can twist Motorola's arm to license it. I'm sure that IBM's SIMD will be AltiVec compatible -- they were rumoured to be quite impressed when the G4 first shipped. Considering AltiVec is what currently keeps them in the performance game at all, I expect it'll be around for quite a while.
On the subject of co-processors... aside from the graphics co-processor I don't think we'll see anything here unless it is integrated directly into the CPU. The G4 is already memory bound so adding a co-processor will likely just make the situation worse. Better to improve the CPU in some general way so that everything improves automatically, rather than putting specialized hardware in to do something specific. Graphics are the exception because the machine is always doing graphics for every user.
[ 04-03-2002: Message edited by: Programmer ]</p>
Final Cut Pro is great, but the only reason it doesn't run on Windows is it's owned by Apple. Nice as Altivec is, FCP runs just fine on a G3, and would run just fine on a P4 or Athlon. Application-specific accelerators cancel out the Altivec advantage for less than the cost difference of buying a Mac over a PC.
Apple has to figure out a way to market the Mac's TRUE advantage -- ease of use, low cost of support, low overall cost of ownership, flexibility, etc. -- and eliminate the Mac's TRUE disadvantage -- poor performance.
<strong>Actually hardware-based DVD encoders for Windows are faster and (according to some) better than the software-based options available on the Mac.
Final Cut Pro is great, but the only reason it doesn't run on Windows is it's owned by Apple. Nice as Altivec is, FCP runs just fine on a G3, and would run just fine on a P4 or Athlon. Application-specific accelerators cancel out the Altivec advantage for less than the cost difference of buying a Mac over a PC.
Apple has to figure out a way to market the Mac's TRUE advantage -- ease of use, low cost of support, low overall cost of ownership, flexibility, etc. -- and eliminate the Mac's TRUE disadvantage -- poor performance.</strong><hr></blockquote>
I just can't understand all the people that say poor performance on the Mac. Most of the Pro Apps that people use Macs for are faster on the Mac. Photoshop being the most used example, but yes DVD encoding on a mac blows away any Wintel machine. FCP does run fine on a G3 but not if you want the realtime effects.
Processor design is the big difference here. Most people here claim to understand it, but they really don't. We can argue about SIMD units and parrellism, # of units, etc. Every part of chip design has a trade off involved somewhere.
I bet I could get a 3GHz G4 chip working off a fab within 6 months. I would increase the pipeline depth to 20 stages like a P4, sacrafice some parrellelism in the design to add more branch prediction units. Greatly increase the die size, power consumption, heat output, and just turn up the clock speed.
You forget that one of the goals of PPC was to follow a different path of design. Maybe you should go read up on the guidelines that were orignally set up by the AIM alliance and see what they have accomplished vs what they said they would do.
Here's the thought experiment:
Resolved, Moto has completely dropped the ball, there is no G5. Development is waay off track. What to do?
Well, if I'm Steve, I pack my toys and go play somewhere else.
If I remember correctly, Apple is co-developing chips with Moto. I'd take my designers and the work so far, and go shopping for a fab. Buying would be to expensive, but out-sourcing the development (actually, partnering on it) and later fab could be done outside of Apple.
As has been suggested before, look to AMD, maybe. What about TI? IBM?
Product wise, I'd start hammering all the enhancements I could into the existing product line. DDR, RapidIO, etc. Someone suggested (in another JYD thread) using the old Amiga model of specialized chipsets for video,etc processing. Not just a romping nVidia chip, but for rendering and other floating point functions. Kind of like adding dedicated processors for various functions.
I'd begin looking really hard at what the most needed software is for the platform. What can really provide a boost? Then, I'd use some of that 4 billion sitting in the bank to support development. I'd work on development of tools to allow greater support options for IT departments. SNMP would be great. Increase integration with WinTel networks, ie network drives, printing, file transfer.
Spend $ on consumer education. I know Apple has a pretty good ad campaign right now. Double it. Triple it. Spread the word far and wide.
Finally, using the current CPU's roll out new product segments. MP servers. Rackmount servers. Headless Imacs (rebirth of the cube?) ProSumer machines. Not to clutter the product matrix necessarily, just add a couple of squares. 8)
Just my two cents.
Flame on!
And if we go by your simple minded thinking of only MHz than Intel might as well give up on their Itanium since it's only going to be 800 or 900 MHz and the 64 bit G5 will be 1.5GHz+
<hr></blockquote>
I don't need to be lectured on the relationship between MHz and performance. You're thinking in simple minded terms if you actually think that a G4 is faster than a Pentium 4 irrespective of MHz. Speed scales with MHz for any given architecture, and what it comes down to is that Intel has scaled the x86 architecture better than AIM has done with the PPC. Yes the PPC is a more efficient CPU and is capable of completing more instructions per clock cycle (ICC) under most circumstances, but the ICC is not a boundless number. ICC is constant for the PPC and even falls as Moto adds more stages to the pipeline.
Think of it this way: CPU performance can be expressed as instructions per clock cycle (ICC) times cycles per second (GHz). This is drastically oversimplified but it serves for explaining the idea to your feeble mind. Now, suppose for the sake of argument that the PPC can do 40% more instructions per clock cycle than the x86. But at the same time, the PPC runs at 1.2 GHz, and the x86 is at 3.0 GHz. So we have something like this:
x=instructions per clock cycle.
PPC: (1.40x)(1.2 GHz) = 1.68 x
x86: (1.00x)(3.0 GHz) = 3.00 x
So even though the PPC is 40% more efficient, the x86 completes ( 3.00/1.68 = 1.79) 79% more instructions per second by the use of brute GHz.
Does this help you understand the looming performance gap between Macs and Wintels? I hope so.
Anyways there's some really interesting replies to this thread, and I'm going back to reading them.
It's nice to see some optimists here who support their optimism with facts. Being a pessimist by nature, I need that sort of perspective to keep me going sometimes.
[ 04-03-2002: Message edited by: Junkyard Dawg ]</p>
Ahh, I dunno. <img src="graemlins/hmmm.gif" border="0" alt="[Hmmm]" />
<strong>
.</strong><hr></blockquote>
Note taken