Chip Speed Vs. Power Demand

Shipments of dual-core x86 processors are expected to outstrip single-core versions this year as IT departments embrace them as the surest way to increase data center performance while containing energy

February 27, 2006

1 Min Read
Network Computing logo

Shipments of dual-core x86 processors are expected to outstrip single-core versions this year as IT departments embrace them as the surest way to increase data center performance while containing energy costs.

For 38 years, Moore's law was demonstrated by a near doubling of processor clock-cycle performance every two years. In the age of multi-gigahertz performance, however, while the number of transistors continues to grow, speed increases are being constrained by concerns over associated increases in a chip's power dissipation.

The newest processors operate at slightly reduced clock frequencies and generate less heat. They provide increased performance by placing two processing engines within the same silicon area. Increasingly, the key metric for processors is performance per watt.

AMD offers processors with three levels of power dissipation, which is also a measure of heat loss: 95, 68, and 55 watts. A general rule is that 55-watt processors, while cooler, are two "speed bumps" reduced from an equivalent 95-watt device.

The next step is quad-core chips. Intel has said it will ship a quad-core Xeon early next year, and AMD has indicated its first quad-core Opteron will come late next year.Multicore processors already are a reality outside the x86 world. Sun Microsystems' UltraSparc T1 has eight processing cores running at 1.2 GHz each and a power dissipation of 72 watts. The chip's power dissipation could be reduced further by turning off two or more of the cores and running in quad- or dual-core mode.

Continue to the sidebar:
Uncool Data Centers Need To Chill Out
Return to the story:
Power Surge

SUBSCRIBE TO OUR NEWSLETTER
Stay informed! Sign up to get expert advice and insight delivered direct to your inbox
More Insights