How we test power consumption
- By Steve Platt
- Jun 22, 2006
WattsUp Pro power meter
When looking for a workstation PC, most agencies will be interested in getting something powerful. Whether you want to crunch some serious numbers with Excel, zip up a file to send to a co-worker or execute a large simulation, power is key. But what is the cost of that power?
We at Alterion, in consultation with NASA and GCN, have been developing a series of power performance benchmarks. These benchmarks measure the amount of power a computer consumes while executing small but significant common tasks.
The basic test bed consists of a WattsUp Pro power meter from Electronic Educational Devices (www.doubleed.com), the PC being tested and a monitoring system, which is a second computer hooked in to record results. The tested PC plugs into the WattsUp meter, which takes power measurements. A serial cable goes from the meter to the monitoring system, where the latter records power consumption at a rate of one measurement per second.
The tested PC repeatedly executes small, well-defined application-based tasks from the Alterion ALP benchmark suite, while the monitoring system records power consumption. Each test runs for a fixed period. We then take the total power consumed, divide it by the number of tasks completed, and the result is the power used (in watt-seconds) to perform a computing task.
These tests are important because they help answer a few questions. For instance, if a machine is faster but consumes more power, is it a better system? Over time, does one system cost more than another when performing office tasks? In other words, what would be more effective: using more power for shorter periods of time (i.e., a faster system) or less power for longer periods of time (i.e., a slower system)?
The current Power Measurement Test Suite consists of five tests, which measure power consumption in five situations: Running idle, Excel charting, Zip file compression, Photoshop filtering and application loading (we report three of the results in our performance chart). By most estimates, your system will be idle almost all of the time, but the remaining tests measure quick bursts of application activity. Each test uses components throughout the system, but uses them differently. For example, the Excel charting test taxes the graphics controller combined with data access and calculation processes; the compression test balances CPU and memory (to compress) with disk I/O (reading and writing). The better tuned a PC is overall, the more power-efficient it should be.
This test suite is part of the continuing Alterion ALP benchmarking program. Alterion (www.alterion.com
), based in Conshohocken, Pa., is known for its work in systems integration support and testing efforts for government agencies such as NASA and the Defense, Justice and State departments.
For more information about Alterion's Power Performance Benchmark, please e-mail Matthew Shapiro at [email protected]