2.5. Step 3: Sub-100nm Scaling – Leakage Takes Over

With the advent of 90nm and below: ● Static power due to subthreshold leakage, gate oxide tunneling, and junction leakage became substantial.

As semiconductor technology progressed to the 90nm process node and smaller, new challenges arose in managing power consumption. Specifically, the term 'leakage' refers to unwanted current that flows through a transistor even when it is turned off. Subthreshold leakage occurs in transistors that are near the threshold of being turned on. Gate oxide tunneling involves leakage through the thin insulating layer around the gate of a transistor, and junction leakage happens at the regions where different semiconductor materials meet. These kinds of leakage currents contribute significantly to the overall static power consumption in chips, especially as transistors are made smaller.

Imagine a leaky faucet that drips water when it’s turned off. Even though the faucet is supposed to be closed (like the transistor being off), water (or current, in this case) still manages to escape, wasting resources just like leakage currents waste power in electronic devices.

● Supply voltage scaling hit limits due to noise margins and variability.

As efforts were made to reduce the supply voltage (Vdd) to lessen power consumption, engineers encountered significant obstacles. Lowering the voltage improves power efficiency but can also decrease the noise margins of a circuit, making it more susceptible to errors. Additionally, manufacturing variations can cause some transistors to respond differently to voltage adjustments, which leads to inconsistencies in performance across chips. Hence, simply reducing Vdd was no longer a viable solution for high-performance designs.

Think of a tightrope walker trying to balance on a dish that becomes increasingly smaller. The lower the dish gets (analogous to reducing Vdd), the harder it becomes to maintain balance without tipping over (representing increased susceptibility to errors).

● High-performance designs could no longer rely on just reducing Vdd.

As power consumption from leakage became a more pressing issue, engineers had to innovate beyond just lowering the voltage of their devices. The complexity of modern chips required more sophisticated strategies to manage both performance and power consumption. Instead of relying solely on voltage scaling for power efficiency, designers began to explore a variety of other solutions to address the growing leakage issue while still achieving desired performance levels.

Imagine a car that gets poorer mileage as the speed increases; simply driving slower isn't enough if you want to keep speed and still save fuel. In this case, you have to focus on improving other aspects of the car, like aerodynamics or engine efficiency, paralleling how engineers had to adapt their designs beyond voltage reduction.

Solutions included: ● High-Vt transistors for standby. ● Body biasing for leakage control. ● Power gating and sleep transistors to fully disconnect power in idle modes.

To combat leakage issues, engineers developed several strategies. High-Vt (high threshold voltage) transistors are employed to reduce leakage when the circuit is inactive; these transistors allow for lower leakage currents in standby states. Body biasing is another technique used to control the threshold voltage of a transistor dynamically, effectively managing leakage during different operational modes. Power gating uses sleep transistors that completely cut off power to sections of the chip that are not in use, ensuring that no energy is wasted when those parts are inactive.

Think of these solutions like a smart home: when you're not in a room, you can turn off the lights (power gating) and adjust the temperature on your thermostat (body biasing), ensuring energy isn't wasted when it's not needed, similar to how these techniques minimize leakage power in circuitry.

Insight: In 65nm CMOS, leakage power could account for >30% of total power in idle state.

As chips evolved to the 65nm node, the significance of leakage power became increasingly apparent. Data indicated that more than 30% of the total power consumed by the chip while in a non-active (idle) state came from leakage alone. This highlighted the urgency for efficient designs and strategies to minimize leakage, as the cumulative effect of idle power leakage could critically impact overall device performance and energy consumption.

Just like a smartphone that continues to drain battery life even when not in use due to background processes, leakage power in chips represents wasted energy that can lead to significant efficiency losses, demanding advanced management to keep power usage low.

--