Donate to Science & Enterprise

S&E on Mastodon

S&E on LinkedIn

S&E on Flipboard

Please share Science & Enterprise

New Data Encoding Method Cuts Energy for Memory Cards

Azalia Mirhoseini (Rice University)

Azalia Mirhoseini (Rice University)

Engineers and computer scientists at Rice University in Houston and University of California in Los Angeles have discovered a way to write data on computer memory cards that cuts the energy needed for the task by 30 percent. The team from Rice’s Adaptive Computing and Embedded Systems Laboratory, led by Ph.D. student Azalia Mirhoseini (pictured right), presented their findings earlier this week at the IEEE/ACM Design Automation Conference in San Francisco.

The researchers based their technique on a memory process called phase change memory (PCM). This technology offers a type of non-volatile memory — it keeps the data stored even when power is turned off — considered a successor to the familiar flash memory drives, but faster, cheaper and more energy-efficient. Phase change memory works because of its asymmetric physical properties that make it easy to measure or distinguish one electronic state from another, in this case from low to high resistance, a requirement for coding 1 and 0 in computer memory.

The Rice/UCLA process takes advantage of these asymmetric properties. In this approach, data are read before new data are written. This step makes it possible to scan the configured data and overwrite only the parts of the structure that need changing. “We developed an optimization framework,” says Mirhoseini, “that exploits asymmetries in PCM read/write to minimize the number of bit transitions, which in turns yields energy and endurance efficiency.”

The second part of the new method makes use of integer-linear programming, a mathematical and analytical technique that aims to find optimal solutions. The more complex the solution, the longer it takes for integer-linear programming to find the optimal solution, so the team found a shortcut by using a temporary store of small codes that could be quickly combined for more complex solutions.

UCLA computer science professor Miodrag Potkonjak, a collaborator on the project, says this is a practical use of integer-linear programming since the there is only one search process for the small codes, during the design phase. “The codes are stored for later use during [the] PCM operation,” Potkonjak adds.

In addition to the reduction in energy required, the researchers also found the new encoding scheme extended the life of the memory cells that can handle a finite number of rewrite cycles before they become unusable. The team reports that the new encoding method cut more than 40 percent off the memory-wear rate of the cells that results in exhaustion of memory due to rewrites.

The researchers say the feasibility, low overhead, and efficiency of the proposed optimization methods were demonstrated with a series of evaluations on benchmark data sets.

Read more:

*     *     *

Comments are closed.