At this year’s Portable Power Conference in San Francisco, speakers examined some of the options for maintaining or extending portable device runtimes in the face of growing demands for power. As several presenters observed, the demand for power in more functionally rich portable products is growing much faster than the incremental gains in energy density being achieved by Li-ion batteries. However, new power management techniques applied to microprocessors and displays may alleviate the looming energy shortage in portable equipment.
In his talk, “Empowering Portability,” Art Swift of Transmeta Corp. provided some examples of how the CPU could be fine tuned to achieve longer battery runtimes.
In assessing the power problem, Swift noted that CPU and graphics power consumption are growing with the transition to finer process geometries and smaller transistors. According to Swift, the CPU peak power consumed in the leading mobile PC will rise to 45 W next year. Compounding the problem is the predicted skyrocketing of leakage currents with the migration to finer semiconductor process geometries.
Swift discussed two power management options for stemming the rise in power consumption. One is the use of code morphing software, which profiles and translates X86 instructions into optimized native VLIW (very long instruction words). This technique effectively trades power-consuming transistors for software. Swift also pointed to his company’s LongRun power management technique, which changes the CPU’s voltage and frequency hundreds of times per second to match computing requirements.
In what sounds like a novel twist on power management, Partha Ranganathan of HP described a concept similar to tuning the CPU, but applied to displays. As Ranganathan explained in his talk “Low Power PDA Displays,” the displays found in mobile systems consume significant amounts of the system power—50% in laptops and 61% in handhelds. Previous approaches to reducing the display power have involved either turning off the display or using a smaller display.
Ranganathan argues that a better approach would be to match the display power consumption to the actual screen usage and user requirements. Producing such an energy-adaptive display is challenging because it demands an understanding of screen usage combined with the appropriate display technology and software design. In the studies he conducted, Ranganathan found that on average only 60% of the screen is being used, referring to the segment or window within the screen that the user is actually focusing on. But despite the partial usage of the display, it still consumes full power.
In his talk, Ranganathan presented an energy-adaptive hardware design based on OLEDs. The use of OLEDs allowed pixel power to be varied by varying the brightness and color of the pixels. This capability allowed Ranganathan to dim the portions of display that were less important to the viewer. Ranganathan found that energy usage could be reduced by 22% to 88% without intruding on the user’s experience. Results, in terms of energy savings and viewer acceptance, varied according to the extent to which the display was being dimmed. Surprisingly, some viewers preferred having portions of the screen dimmed because it allowed them to focus on text more easily.
For more information on either presentation, see www.portablepowerconference.com.