A computer or processor in standby mode does not perform any
computations but is ready to resume normal operation, usually
within a specified minimum amount of time.
The percentage of time at which the computer is in standby mode
is the standby ratio
.
The goal of optimum power efficiency at a given performance remains
the same, but the term ``power'' now refers to the total average power
consumption
which is the weighted sum of active-mode power consumption and standby-mode power consumption:
The simplest form of standby mode is to operate the processor
in the normal active state but with a lower or zero clock frequency
in an idle mode.
In this case the switching energy can be expressed as
The essential problem associated with this simple standby operation is that the required low for a high standby ratio, which is quite common in portable applications such as notebooks or palmtops, leads to a tradeoff between active and standby mode, which may effectively prevent an Ultra-Low-Power technology. Potential ways to tackle tackle this problem are:
This approach may be too expensive because of the chip area consumed by the power switches. However, when the switches are already present as part of a power-supply circuit the only additional circuitry needed for the power-down mode is a suitable control of the power supply (cf. Section 2.8).
Furthermore, certain logic styles provide an implicit infrastructure for power-down modes. One example is pseudo-NMOS logic (see Section A.2.3.3) which allows to turn off individual logic blocks with the gate of the PMOS pull-up transistor
Internal states of the high-speed core of a chip are transferred to a memory which uses high- transistors before the core is turned off.
In standby mode is decreased to a minimum of to reduce the standby power by a factor of . As the for normal operation is larger there is still a tradeoff between active and standby power.
A reverse bias on the wells is used to increase the the threshold voltage so as to turn the devices off (cf. Section 2.7.2). This approach, however, has some disadvantages, especially, it cannot be implemented effectively for fast deep-sub-micron devices.