Intel aims lower
Goodbye, Xeon. Hello, lower power consumption and lower prices
The keynote at August's Intel Developer Forum could have had the ring of a concession speech. Instead, Intel President and CEO Paul Otellini rang out the Pentium 4 era with style -- considering he had to put a smiley face on the reversal of Intel's greatest strategic blunder. That was not, as AMD says, Intel's program to phase out x86-server CPUs in favor of Itanium. I know, and I believe Paul knows, too, that Intel's worst blunder was not trusting its own engineers and tech-savvy management. I saw vision in Intel's original road map that ramped down x86 to its proper position as a cheap, high-volume part for clients and bargain servers. There was no vision in Intel's blind race to obliterate AMD64 by attempting to co-opt, and thereby devalue, the unique aspects of AMD's technology.
The lesson here is "don't strategize angry." Intel thought it could knock AMD out with a combination of strong-arm marketing and just barely enough engineering to get 64-bit, dual core, fast memory, and a high-speed bus in magazines' feature tables comparing Opteron with Xeon. It went way too far, not for any unfair treatment of AMD -- that's a separate question -- but because Pentium 4 was architected from the word go to make a graceful exit and make room for Itanium.
When AMD surfaced with Opteron, Intel had three choices: stick with its Itanium road map; back up to a simpler, more extensible Pentium core -- closer to the one that Intel's shenanigans forced AMD to create -- and build it to satisfy the higher demands of the server market that AMD accurately visualized; or try to snuff AMD before the end-user value of AMD64's innovative engineering could prove itself. The first two choices were strategically sound.
Intel was wrong to shift Itanium into neutral when the market whined that it wasn't a 386. It should have put on a full-court press to bring on ISVs while making it irresistible to Microsoft. Instead, Intel was too boastful about the science of Itanium and painted it as far ahead of its time. This only heightened Itanium's risk of market rejection.
Intel also blundered in failing to spread the word internally that Pentium M, a spin-off project to create a low-power mobile x86 CPU, was the company's best work since 8086. Intel owed that engineering win to a bold effort to get the project away from Intel's stodgy home turf. Intel empowered Pentium M's engineers to keep the Intel x86 assets that made sense and toss those that didn't. The resulting chip wasn't pure magic, but the "everything's on the table" approach was, and Pentium M was pure gold in the notebook market as part of the Centrino chip set.
Intel's new emphasis is on CPU power efficiency. Intel's got a great story there in the Pentium M core, but in servers, AMD already has an overwhelming lead. AMD migrated PowerNow to Opteron -- it cuts Opteron's power and heat by as much as 75 percent during periods of reduced demand. AMD's lineup also includes Opteron HE and Opteron EE, CPUs that draw roughly one-half and one-third, respectively, of top-seated Opterons before PowerNow is applied.
The market will be happy to join Intel in its march backward toward a future where humans can actually share quarters with servers, and where unused compute cycles can't be measured in metric tons of coal.
