Wow has this thread drifted over the map or what?
At the risk of interrupting the argument between asarsun, JCheng and others, let me pick up on the theme Ravenflight has introduced recently.
But first, a point to:
Mac-Xpert:
they (Intel) will never be ahead of the competition anymore.
We dont know this with any degree of certainty, any more than we know IBM WILL be ahead of the competion once again. What we do know is that there is room for improvement, and considering some of the capabilities of the Itanium2, they do have a few tricks in their bag as yet undispensed.
Now, as to Ravenflights theme:
So where do we stand? We stand at the end of an era. The end of the great race. There will probably be no more transitions to new architectures like the Cell or something better. R&D into new PC architectures will dry up, because there will be no major OS player to lead the way.
As a matter of history, processor innovation has NOT been entirely related to the competition between rivals like Intel and Apple/PowerPC. Let me trace that a little.
In the 60s, John Cocke worked on projects like Stretch, which wandered and were eventually cancelled by IBM. Later, he moved on to the 801, which as I understand it led to the patents IBM once owned on RISC. His work moved on to make the RT, a full RISC design in a workstation class product (in its day) that flopped for IBM. The technology, though, was not a flop just the product. This work was further advanced into the RISC/6000, which was, alas, a reasonably successful product line for IBM at the time. In the RISC/6000, John Cocke implemented the full range of his ideas (if in early form), developed over a decade or more. His genius considered the puzzle of performance from the compiler through the logic circuits. Along the way he created a logic simulator, which was used to test design ideas before investing in production, and its the mainstay to this day.
John Cockes contributions were a matter of research, and the result of inspiration and energies of a mind exploding with new ideas. They were not fashioned on the notion of product per se, they were much closer to pure research. IBM paid his salary because they recognized these ideas would become product, but Cockes motives were intellectual and emotional outlets; self actualization, the very heart of modern psychology. He wasnt in it for the money so much he was in it to produce the fruit of inspiration to be himself.
His influence is akin to Einstein, on a smaller scale I suppose, and limited to the domain of CPU design, but similar in ways to Einstein. Einsteins influence on physics was explosive a single mind, and a new set of ideas transformed physics. Since Einstein, progress has continued, but not from a single individual with such potency. Progress has been incremental, many times in tributaries that lead nowhere. Thousands of minds expanding the sphere of knowledge, and taking that knowledge into practice, have advanced science by leaps, but not by such a leap as Einstein himself produced. The post-Einstein advancements have been incremental by comparison, and are due to the brute force of so many followers working from Einsteins ideas (sometimes even ones Einstein discarded).
It is so with Cockes inventions. Within IBM, and now throughout the industry, Cockes ideas are at the heart of most innovations that deliver increased power. The accompanying advancements in production (shrinking circuit sizes, increasing transistor counts) are NOT the innovative advancements of the kind Cocke introduced. They are incremental, tributary, and though important (without doubt), they are not the explosive product of a genius in research.
The PowerPC is the result of IBMs interest to repackage the otherwise cumbersome RISC/6000 processor into a more convenient product, fitting better with trends and conventions of the mass produced computer products. It worked well, and it benefits greatly from the more incremental advancements that have come since the 92 retirement of John Cocke, but nothing has been quite as explosive with potential since Cocke.
Until another John Cocke is found, and just where that person will have employment is unpredictable, we are left with the more brute force, slower moving advancement made by the thousands of followers which take his ideas and move them forward, sometimes with tributaries that lead nowhere, and sometimes with genuine bursts of performance gains, but none that match the monumental leaps reaped from the full deployment of Cockes ideas.
I submit, as a result of that review, that the end of the era began with Cockes retirement from IBM, and the completion of the RS/6000. We have been coasting forward from that momentum, which has lasted many years.
The forward force of development progress, therefore, is at a pace set by market forces. Money, brute force trial and error are the driving forces at work, until another bright genius takes the stage again. Where money drives the opportunistic development pace of the brute force approach, IBM is not nearly as well positioned as Intel. In other words, IBM isnt as focused on directions that deliver what we want from the next PowerPC. We have been receiving latent benefits from better implementation of Cockes original proposals, mixed with more incremental benefits from related technologies. Cockes ideas are now found in all CPU development, from all companies. Since the level of performance from each offering is similar (none are double or triple the others), we are left at the mercy of money pushing the progress.
I do agree that without competitive forces from multiple sources, innovation might stall. If Intel were able to sit back, drivel out similar processors from one year to the next without concern over competitive offerings, then aside from another genius bursting with new ideas, only one other force would motivate change: software.
One thing that happens over time is that our expectation of the machine increases. Software does the actual work in meeting that expectation, and when something comes along that requires more real horsepower, the only answer is increased hardware performance. At some point a stagnant Intel, without competition, would have to survive in a market saturated with chips at the then current levels of performance. Who among us would buy new machines if they offered little more than our current ones? To spur new sales, new levels of performance, required by increasing expectation of software capability, would drive the development wave, motivating all of us to upgrade again.
Then, theres another John Cocke due any year now. When Cocke worked on the RISC concepts, IBM was all but the only player in the game at that level. While competition drives the brute force, incremental advancements, it is the energetic bursts from unique minds that provide the next truly great advancements. After all, Einstein wasnt in it for the money, either.
