The secret's out.
Intel is re-interpreting Moore's Law. Not repealing it. Not rejecting it.
They're reinterpreting it. That's the significance of the change incoming CEO Paul Otellini (right) is making.
Before Moore's Law was like Samuel Gomper's famous quote about what labor wanted. More. More circuits, more speed, more cycles, more bits. More.
This led to mistakes like the Itanium, which Intel is still living down.
As of today Intel's new direction is better. Better doesn't always mean more. In the case of microprocessors it can mean putting more computers on each chip (multi-core) or running with lower power. In terms of communications it can mean a host of attributes, from security to coverage to throughput.
Much is being made of the fact that incoming Paul Otellini was not originally an engineer. Much should be. Because better is a subjective ideal, whereas more is objective. Better can be understood by the businessmen and the market. More can be understood by engineers.
This doesn't mean, and can't mean, that Intel is no longer run by engineers. Engineers must still do all the heavy lifting, and the heavy thinking. But it does mean that there are more people at Intel interested in defining, and explaining, better -- to OEMs, to customers, to consumers.
It's a balancing act. But business always is. And Intel remains a very profitable business.
But it also means that things from here on out will be a lot more complicated. That's cool. Complexity is good.
Think of it as Moore's Law in action.
Dana Blankenhorn has been a business journalist for over 25 years and has covered the online world professionally since 1985. He founded the "Interactive Age Daily" for CMP Media, and has written for the Chicago Tribune, Advertising Age, and dozens of other publications over the years.
1. Athlon64 on May 5, 2005 02:40 PM writes...
Moore's Law is alive and well. Each shrink gives double the number of transistors. Unfortunately the intel transistors leak more than 100 Watts without doing anything. Unfortunately intel has no useful design for twice as many transistors.
AMD long ago planned to use the extra transistors for 64 bits, and dual core. AMD long ago planned how to prevent transistor leakage current and the result is a 25Watt Turion64 notebook processor that runs better than the top of the line intel 150 Watt Xeon server processor.
Permalink to Comment2. two_center on May 5, 2005 03:07 PM writes...
What the author says is that Intel has turned to subject measures, like "better", to sell its processosrs, as symbolized by Paul "marketing guy" Otellini's ascension to the top post at Intel.
Subjective measures are what you use when you start losing on all the objective measures. Even Intel-slanted benchmarks are unable to provide objective evidence in Intel's favor. The trouble for Intel is that convincing customers that Intel processors/platforms are better is more expensive than letting objective numbers speak for themselves, most notably clock speed.
Intel's actions are not entirely AMD driven. The Pentium-M (aka P3) is a "better" processor than the P4 for notebooks. The realization a couple of years ago that the Pentium-M was going to play a large role in Intel's future provided the catalyst for Intel to turn to subjective measures when defining "better". The primary characteristic of Intel's numbering system is that it forces a subjective evaluation rather than an objective measure of performance, like AMD's PR numbers.
The final aspect of Intel's push toward subjectivity is platformaization. Intel doesn't really want "better" to be determined though a comparison of processors alone (AMD vs. Intel). Intel wants the comparison to be on the basis of the platform, the package that supports and works with the CPU. Intel is telling customers that they need to consider more than just the processor. This allows them to dilute the case for AMD processors.
It's a good plan by Intel given their processor design constraints in the middle of this decade. It's up to AMD to make it's price-performance numbers trump the Intel schmooze about "better".
Permalink to Comment3. Jim on May 12, 2005 12:13 PM writes...
What we have is more of a raw physics problem. Over time we've gone from 5um gate lengths, to 50nm, a 100x change. A more narrow gate means a faster transistor, higher clock frequency. So could we get to 0.5nm lengths, another 100x improvement? My answer - no. The atomic diameter of silicon is 0.3nm, so 5um = 16,700 atoms, 50nm = 167 atoms, 0.5nm
From this fact, the problems Intel solved over the past 20+ years are necessarily different than the ones they will solve over the next 20.
Otellini has some very big shoes to fill. Noyce, Moore, Grove, and Barrett are all engineering titans, beyond even the inventor of "Just Do It". He also is in the quandary of not knowing exactly what problems Intel should solve.
Permalink to Comment