AC vs. DC in a battle for data center efficiency

A new study by Schneider Electric finds AC is as efficient as DC in today’s data centers.

For a few years now, there’s been a debate over which power distribution architectures – DC or AC – are the most efficient for data centers. I know I don’t have to tell you, but for the record DC is direct current, which is a type of electrical current that travels through a circuit in only one direction. AC is alternating current (the most commonly used), which is an electrical current that frequently reverses direction electricity is measured according to its cycles, with one complete cycle being counted each time a given current travels in one direction and then doubles back on itself.

I recently talked with Kevin Brown, VP of Schneider Electric’s IT Business, about DC vs. AC. The company conducted an internal study through its Data Center Science Center, which conducts various research. All of the studies findings are presented in a paper, called “A Quantitative Comparison of High Efficiency AC vs. DC Power Distribution for Data Centers.” You can see the paper here.

Brown recounted the findings of a study published in 2008 that reported a 28% efficiency improvement of 380 volt DC versus AC distribution. The trouble is, Brown says, that paper compared an outdated AC architecture against a prototype DC architecture. “The reality is, this and a lot of other studies don’t do a true comparison,” Brown says, adding that the studies – while widely picked up by the media – generally provided too simplistic of a view.

The latest study by Schneider Electric compares the most efficient DC and AC power distribution methods, and in the end, find that both have practically the same efficiency.

Brown says that he and his colleagues at Schneider Electric are not aware of any data centers moving off of their established, traditional power distribution to DC. But he says customers continue to raise the question. The study is meant to help them determine the answers that best suit their needs.

The study acknowledges that there are data centers using AC power distribution that are also operating at lower efficiencies, but points to poor designs and aging AC technologies. According to the study, the observed inefficiencies are primarily due to: inefficient IT device power supplies; inefficient transformer-based power distribution units (PDUs; inefficient UPS systems; and operation at loads well below the design rating of the system, which amplifies all of the above losses.

There have been quite a few improvements in the last three years in device power supplies and UPS systems, the study points out, that make installations installed now much more efficient. Also, modular scalable UPS systems make it easier for organizations to install the right UPS for their load, preventing electrical inefficiency due to gross underutilization frequently seen in the past.

The study’s bottom line, by the way: the data demonstrates that the best AC power distribution systems today already achieve essentially the same efficiency as hypothetical future DC systems. Brown says he’s been having a lot of discussions with data center managers about improving efficiencies, and much of the focus I on cooling and capacity planning that keeps over-building in check.

One Schneider Electric customer, Bryant University near Providence, R.I., recently modified its data center strategy, consolidating operations and deploying Schneider Electric’s InfraStruxure solution. The university’s goal was to reduce energy consumption and operational expenses by 20 percent. The implementation included a new integrated modular IT room architecture, the InfraStruxure for high-density applications with UPS, cooling and IT racks, Schneider Electric’s data center monitoring software platform, new overhead cabling and piping to eliminate the need for a raised floor, and a new chiller and diesel generator, according to Schneider Electric. Work continues, but already the university has achieved its target of a 20 percent decrease in operational expenses.

What’s wrong? The new clean desk test
Join the discussion
Be the first to comment on this article. Our Commenting Policies