Intel sets cruise control on Pentium 4

Intel Corp. President and Chief Operating Officer Paul Otellini served notice at the company's recent developer conference that the megahertz era was coming to a close, and the decision Thursday to remove the 4GHz Pentium 4 processor from its road map is clear evidence that Intel has once and for all kicked its speed habit, according to industry analysts.

The remaining single-core chips in Intel's desktop road map will top out at 3.8GHz, falling short of the 4GHz target Otellini promised at the company's financial analyst meeting in November of 2003. The Santa Clara, California, company now plans to increase the performance of its single-core chips by adding cache memory and to enhance the user experience with a number of new features, Intel representatives said. This will have the added benefit of reducing the impact of a short-term chip capacity glut, analysts said.

For years, Intel and rival Advanced Micro Devices Inc. (AMD) focused on the clock speed of a processor as the most important indicator of system performance, believing that consumers would have a difficult time trying to understand the other aspects of processor performance. AMD started to emphasize the amount of work its chips do per clock cycle, rather than the actual speed of the chip, as early as 2001. Intel has only recently changed its marketing strategy to accommodate a processor numbering system and increased consumer education about overall performance metrics.

Concerns about the power dissipation of chips at 3GHz and beyond forced Intel away from pushing ever-higher clock speeds. Intel's guidelines for the current Prescott Pentium 4 design specified that the faster chips in the Prescott lineup could consume as much as 115 watts of power under maximum operating conditions. There is a direct relationship between a processor's clock speed and the amount of power it consumes, and a similar relationship between power consumption and the amount of heat given off by a PC.

When faced with speed constraints, adding cache memory is often the easiest way to improve performance, said Dean McCarron, principal analyst with Mercury Research Inc. in Cave Creek, Arizona. Cache memory stores frequently accessed data close to the processor, where it can be retrieved more quickly than data stored in the main memory.

The process of increasing the clock speed of chips with a mature architecture is a painstaking exercise of finding bottlenecks in the chip's pathways and making subtle changes to enable the higher frequency, McCarron said. This is not especially challenging, but takes up time that could be spent working on more fruitful designs, he said.

A 3.8GHz Pentium 4 processor with 2M bytes of cache will probably outperform a 4GHz processor with 1M-byte of cache, and it will be significantly easier to produce in large volumes, McCarron said.

Sign up for ITworld's Daily newsletter
Follow ITworld on Twitter @IT_world

• Email
• Print
• Share
  • Slashdot
  • Digg
  • Stumble
  • Reddit
  • Newsvine