Home > About > Frequently Asked Questions
Lawrence Berkeley National Laboratory

Frequently Asked Questions

1. Is there commercially available high-voltage Direct-current (DC) equipment?

All of the DC distribution equipment used in the demonstration was commercially available and UL rated for 380V. DC.

2. Can you co-locate alternating-current (AC) and direct-current (DC) equipment in a data center?

Yes, with first-cost implications that must be balanced against operating-cost savings.

3. How much energy can be saved by this strategy?

Quite a bit. Our pilot project yielded 10% facility wide savings compared to the best-available AC technology, and over 20% compared to typical practices.

4. OK, but how cost-effective is it?

In fact, if these principles are incorporated in a new datacenter at the time of design we anticipate first-cost savings thanks to fewer major electronic components and HVAC downsizing. For pre-existing data centers, a small first-cost impact with large operating cost savings and lower lifecycle costs.

5. Can this DC architecture be scaled to full-scale facilities?

It can be scaled, but there would be larger scale equipment (inverters, flywheels, rectifiers, busways, etc.).

6. Why have you chosen 380 VDC?

We have chosen 380 VDC because losses are lower than 48 VDC and server power supplies today convert AC to 380 VDC. It would be a simple process to manufacture power supplies to accept 380 VDC.

7. Aren’t you worried about breaking DC current?

The DC breakers we use are UL listed, with enough capacity to extinguish the arc drawn during current interruption. We are working with industry players to develop a standardized plug connector for improved safety. See pictures [ 1 | 2 ]

8. Is anyone else doing this?

In Europe there is a production-scale DC data center operating at 350VDC, and they have agreed with our recommended 380 VDC approach. There is also a Japanese pilot project using 300 VDC, which is stepped down to to 48 VDC. Direct current infrastructure for supercomputing is an emerging area of focus. For example, IBM is distributing 350 VDC inside of their supercomputers already. Telecom systems have been using 48 VDC systems successfully and with great reliability for years. In addition, many public transport train systems use 600 VDC in the U.S.

9. How will you address the fact that many loads use induction motors?

Most of the devices we are powering here do not involve motor loads. However, there is a very large opportunity for using DC power when induction motors are controlled with adjustable speed drives. These drives inherently convert the AC to DC before conditioning the output to control frequency. Therefore, DC can easily be directly applied to the drive to power the motor.

10. Is there any chance that datacenters will use DC directly from the utility?

Many utilities have explored the possibility of supplying DC to customers. This could be a real possibility in the next 10 years.

11. Are there any standards in place?

IEEE and UL are both very well versed in the distribution of DC – telecom installations have been using 48 VDC systems successfully and with great reliability for years.

12. What is a UPS?

A UPS (uninterruptible power supply) supplies electricity without interruption to servers and other components in the event of a blackout, brownout or other utility power disturbance. Most UPS systems use massive arrays of lead-acid batteries, just like the one in your car, to store energy for just such an event. However, battery arrays require precise temperature control and ventilation to assure performance, have maintenance needs, may impose hazmat and other safety issues, and take up significant floor space.

13. What is a Flywheel-Based UPS?

A flywheel-based UPS stores energy mechanically in a quiet, spinning, composite cylinder. The Pentadyne’s VSS+DC clean energy storage flywheel used at the Sun Microsystems project requires very infrequent maintenance, is temperature-tolerant (no air conditioning needs) and can respond to hundreds of thousands of power interruptions without degradation (batteries become unreliable after a few deep discharges). Since 98% of power disturbances last less than 10 seconds, the flywheel is a perfect solution to instantaneously assume the electrical load for 10-15 seconds or more.

14. What is needed to take this to the next stage?

Voltage standardization, commercialization of server power supplies designed for DC, and development of a standard DC line cord and connector. In addition to this very modest technology development, training and education are key. The current situation is charachterized by considerable lack of awareness of the potential benefits and unwarranted fear about safety. On the benefits side, our demonstration project quantified a 10% energy savings floor compared to best available and 20% or more compared to typical practice. On the safety side, all components in the particular architecture we piloted with Sun Microsystems, Intel, and others are commercially available and UL listed, with the exception of 380VDC power supplies inside the servers. Others have proposed an architecture with 550 VDC distribution --> 48V at the rack and 48V direct configuration, for which servers are available. The straight-48 scenario requires very large wire diameters, so is probably a non starter in most cases.