SiCortex Updates Green Computing Performance Index (GCPI) to Include First Available Intel Xeon 5500 “Nehalem EP” Benchmarks; Metrics Confirm Processor Falls Short on Energy Efficiency
MAYNARD, Mass.--Out-of-control energy consumption has become data centers managers’ biggest headache. Until now, there have been no tools available to objectively measure energy efficiency at the computer level – only marketing hype from vendors clamoring to be labeled as “green.” To help, computer manufacturer SiCortex recently proposed the Green Computing Performance Index (GCPI), a set of metrics for comparing energy efficiency in the High-Productivity Computing (HPC) segment, defined as systems that perform complex scientific and industrial calculations. The GCPI measures overall system performance-per-watt, based on the industry standard HPCC benchmark suite.
In response to HPC community feedback, SiCortex has updated the GCPI to include only the most relevant metrics from the HPCC benchmark suite. Benchmarks that measure latency alone have been excluded in favor of focusing the Index on actual work performed by computing systems. The GCPI remains the first and only tool to measure and rank computers on a broad range of performance metrics relative to energy consumed, providing a highly valuable business tool to help data center managers make more informed decisions.
Intel’s new Xeon 5500 (Nehalem EP) processor was among the first systems to be evaluated using the new benchmarks. Testing on the Endeavor Xeon 5560 revealed a GCPI score of 9.85, second behind the SiCortex SC5832, which scores 12.32, backing up SiCortex’s claims that the chip has failed to meet the market’s increased energy efficiency demands.
“With a crowded marketplace full of companies touting energy efficiency, it’s crucial that businesses have a tool to help cut through the marketing noise and get to the hard facts,” said Chris Stone, president and CEO of SiCortex. “We designed the GCPI in partnership with industry experts to meet exactly this need, and we welcome ongoing HPC community input to ensure its continued evolution as a strategic — and accurate — business tool.”
The initial response to the GCPI from members of the HPC user community revealed the need for a tool more closely aligned to specific application requirements, in addition to the broad metrics provided by the HPCC benchmark suite. As such, SiCortex has developed an online tool that enables HPC purchasers to customize and create their own index. Available online, the GCPI Calculator allows users to change the weighting on specific HPCC benchmarks and obtain a ranking of relative energy efficiency for the leading HPC system vendors. In seconds, the tool provides an important data point to support an informed purchase decision.
“Energy efficiency is an important criterion when selecting HPC systems,” said Phil Dickens, assistant professor at the University of Maine and SiCortex customer. “We recognize the need to go far beyond changing light bulbs to reduce our power consumption, and choosing the most energy-efficient computers is an important consideration. The GCPI is a useful tool to guide our investment in computing systems – to cut our consumption and to control our costs.”
Beyond the cost considerations, Stone notes that data centers’ growing electricity demands are fast becoming a legislative issue. The U.S. government has taken notice of data centers’ impact on the greenhouse gas problem, as evidenced by President Obama’s pending cap-and-trade energy plan.
About the Green Computing Performance Index (GCPI):
Developed by SiCortex engineers in cooperation with industry experts, the GCPI analyzes computing performance-per-watt across a spectrum of industry-standard benchmarks, providing organizations with much-needed guidance in the era of out-of-control data center energy consumption. It is derived from a straightforward calculation that currently examines a representative sample of established high-productivity computers, and catalogues each according to HPCC industry benchmarks and energy consumption data. The index will expand over time as more data becomes publicly available. The HPCC benchmark suite, developed by a team led by Dr. Jack Dongarra at the University of Tennessee, is a well-respected set of benchmarks that measure the full range of computer system performance.
