This early testing showed that the current Beta 4 development systems achieve only a little over 5 hours of operation from a single charge of the lithium iron phosphate battery. According to Gettys, these batteries deliver about 20 WH (watt-hours) of energy.
Gettys told me that the XO's power consumption in those tests was about 3.81 watts, and provided me with a list of pending software changes and alternative usage models that will improve this figure:
- Configuring the DCON (display controller) chip to refresh the display whenever possible, so the primary display clock source can be shut down (saving about 0.52 W)
- Turning off the backlight (saving about a watt)
- Optimizing the wireless firmware to reduce power consumption (savings unspecified)
The DCON and backlight changes alone would bring system-level power consumption down to about 2.28 watts, which would translate into a battery life of about 8 hours and 45 minutes. Saving another quarter-watt on the wireless interface, if that's possible, would bring the XO up to the 10 hours of "heavy use" predicted by OLPC's Walter Bender.
This configuration, however, is only suitable for using the XO as an electronic book reader--and outdoors, at that. That isn't what I call a "heavy use" workload, but when I said as much to Gettys, he insisted that this is what the OLPC project regards as heavy use--so by the project's own standards, it looks like the XO may be able to reach its promised goal.
But not by my standards, for whatever that's worth. I think that the usage model for a classroom environment should assume that the backlight is on and that students are typing, drawing and making their way through computer-aided learning programs. In such an environment, the figures from OLPC suggest to me that the XO will run for only 4 to 6 hours per charge.
But that's probably a good day's use for the students at the heart of OLPC's plans...good enough, I think. The OLPC project ought to stop spreading around battery-life estimates that many users won't see in real-world operation. This is one of those cases where it's better to underpromise and overdeliver.