Home
About
Apple
Career
Experiences
General
Graphics
Hardware
History
Humor
Interface
Networking
OS
Opinion
Politics
Programming
Quotes
Reviews
Security
Software
Sound
Thought
Web
Cheap International
Airfare Online
Wachovia online banking
Get Free Coupons Online
Finding the perfect
discount hot tub
Payday Loans
Stock Trading Online
Stuffed Animals
Smart Investing Online
|
|
What happened to the G5?
Or what's going to happen?
By: David K. Every
|
Kind:
Created:
Size: |
Article
2002-11-05 06:34:25
7 KB |
|
|
 |
 ou'd have to ask Motorola or Apple where the G5 is; I'm not a rumor site. I assume that there will be something in the not too distant future; and that those parts may help Apple to have something until the 970 takes over the top end, and even after that, will probably make good chips later for the low end. Either that, or Apple is so fed up with Motorola that they are doing a hail-mary and going for IBM all the way. I just don't know how bad things are in Motorola. They were bad a few years ago, and heading worse. I haven't seen anything that makes me believe things there are radically different.
The problem is that if the G5 had been delivered when it was initially targeted, I think it would have had a much bigger impact. The more that time goes by, the less it matters. Now, it would have to be pretty spectacular, and probably more aggressive than what it sounded like they were working on, to be much more than middle of the pack chip. And with Motorola missing every delivery and promise, my confidence in them is evaporated.
Frankly, Motorola has done a lot to hurt itself and Apple. And Apple hasn't helped. The G4 was a nice chip, most of the problems weren't in the design, but the implementation; getting the kinks worked out, and manufacturing the thing to be fast enough. Frankly, Motorola didn't feel there was enough market in that to make the effort (or make enough effort to be successful), and their management drove many of the best people out the door and to competitors. Then they couldn't compete as well, and wouldn't put the money in to be able to compete in the future. Without that investment, they're doomed (long term).
The good news is that when you kill R&D, run lean, and stop innovating, you can be very profitable on your descent and often survive for years or decades, because you are just milking what you developed years or decades before. But long term, you're just waiting for the end. Up or out, innovate or become irrelevant. Motorola seems to be opting for the latter: and Apple facilitated them too long.
Most of that investment doesn't seem too hard. Heck even multi-cored G4's, which isn't exactly rocket science, would have made them look more innovative. But Apple and Motorola both point fingers. In the end, I believe Apple, and those who left Motorola, more than I do those left in Motorola who appear to be making excuses. In the end, there was a ton of wasted opportunities that were never realized; and I grieve that loss.
            
IBM has been the pioneer of the industry; look at the Power3 and Power4, and you saw all sorts of pioneering concepts. IBM was the first to break the 1 Ghz barrier, was the first to offer dual-headed variants of their chips (what Intel calls hyper-threading), and IBM learned its limitations and abandoned that design approach already too. IBM pioneers SOI and Copper interconnects. IBM took branch prediction to the next level, long ago. I love IBM's R&D, and if they can ever bring it down to a consumer level, the industry is going to win big. Hopefully the 970 is the start of that trend.
IBM also has a couple "G5's" - meaning 5th generation PowerPC's. Certainly the 970 could be considered that; but IBM considers it more the consumer version of their Power4 (with a few tweaks). IBM has a mainframe product line that they called the G5 because it was more than their Power4; but it is really a different product line altogether. IBM is also working on their Power5; and assumably, there will be a single-chip consumer version if their move to the 970 is at all successful. But most of us would think of that as a G6.
So the IBM products are not the G5 that we had heard of. In theory, IBM could manufacture that chip for Apple. But in practice, I suspect it was co-designed with Motorola, and used their processes and tools, and has legal and political co-design issues that would make that completely implausible.
        
I think it matters. It is just how much. If they came out with it soon, and it was great, it could matter a lot. Heck, there is some headroom in that theoretical design, and it could be better than the 970. The 970 is a lot more complex and powerful chip; but there are many ways to speed. And remember, in chips it is all where you spend your transistors.
The Power4 philosophy was to make the most pioneering processor on many levels, and think big. They didn't have to worry about gates (transistors) as much; this was going to be big and expensive; so they were making the best processor core they could, for single stream execution, and would support many processors, and so on. This is to processors what the Ford Expedition is to minivans (see big). Moore's law, has just made it viable to move it down to the consumer level.
The G5 was more limited in some ways. It was how to redesign a much smaller core, to make it faster and better. It was to lengthen the pipes, and support 64 bits, and so on; so there are similarities. But different companies, design philosophies and times. IBM could afford to throw 50% more transistors at a problem area to get 20% more performance, Motorola (and thus Apple) were thinking far more conservative and were far more concerned with gate count and power/heat and so on. The results are that the G5 could easily end up being 80% the chip that the 970 is, for half the transistors and a much smaller power budget. (Conceptually, I know nothing in specific). And there are certainly places where both processors make sense.
Personally, there are cases where gate count really matters, and the G5 could have been a far better chip than the 970. Let's say that it is 75% the chip performance for 50% the transistors; that means you can do dual cores for the same cost, and in many ways have far better performance overall.
So in the best case world, both are likely to happen and next year or two will be great. Moto will come out with single core G5's and it will be a big step forward, and a little later they will go multi-core for the same transistor count (cost). That would be good for both Motorola and Apple. Around the same time as Moto would be thinking multi-core, IBM could be releasing the 970, and Apple has a kick-ass desktop and server system, and better single thread execution than they can get out of a G5. And they have many design decisions and choices that they can make in creating good solutions. But I've lost so much confidence in Motorola in ever hitting a schedule or following through, that I'm not holding my breath.
I'm quickly descending past explaining technical concepts and tradeoffs, and getting into rumors and speculation, which isn't my goal; so it's time to come back to the real world and get back to work. My point is just that one design doesn't obviate the other; and two radically different approaches can be successful, even at the same time. And there are different needs even in the same market; I do think that the potential is there for both a 970 type processor and a lighter G5 type processor. But we'll just have to wait and see what happens. 
 Format for Printing  Mail
|
|
|
|
