energy-delay product
**Energy-Delay Product (EDP)** is a **composite metric that quantifies the energy efficiency of a computation by multiplying the energy consumed per operation by the time taken to complete it** — penalizing both energy-wasteful designs (high energy) and slow designs (high delay) equally, providing a single figure of merit that captures the fundamental tradeoff between power consumption and performance in digital circuit and processor design.
**What Is Energy-Delay Product?**
- **Definition**: EDP = Energy × Delay = (Power × Time) × Time = Power × Time², measured in joule-seconds (J·s) or picojoule-nanoseconds (pJ·ns) — lower EDP indicates a more efficient design that achieves a better balance between energy consumption and computation speed.
- **Why Multiply**: Simply minimizing energy is trivial (run at the lowest possible voltage and frequency), and simply minimizing delay is trivial (run at maximum voltage regardless of power) — EDP captures the insight that a good design must be both fast AND efficient.
- **Voltage Scaling**: EDP has a minimum at an optimal supply voltage — below this voltage, the delay increase outweighs the energy savings; above it, the energy increase outweighs the speed improvement. This optimal point is typically 0.4-0.6V for modern CMOS.
- **Technology Comparison**: EDP enables fair comparison between different technology nodes, architectures, and circuit styles by normalizing for both speed and energy — a design with 2× lower EDP is fundamentally more efficient regardless of whether it achieved this through speed or energy improvement.
**Why EDP Matters**
- **Optimal Voltage Finding**: EDP analysis reveals the supply voltage that provides the best energy-performance tradeoff — critical for battery-powered devices where both battery life (energy) and responsiveness (delay) matter.
- **Architecture Evaluation**: Comparing EDP across different processor architectures (in-order vs. out-of-order, RISC vs. CISC) reveals which architecture is fundamentally more efficient for a given workload.
- **Technology Node Assessment**: EDP improvement per technology node generation quantifies the true efficiency gain — a node that improves speed by 20% but increases energy by 10% has a net EDP improvement of only 12%.
- **Circuit Design**: At the circuit level, EDP guides the choice between static CMOS, dynamic logic, pass-transistor logic, and other circuit families for each function.
**EDP Analysis**
- **EDP vs. Voltage**: For CMOS circuits, EDP = C_L × V_dd² × t_delay, where delay ∝ V_dd/(V_dd - V_th)^α — the EDP curve has a clear minimum at the optimal operating voltage.
- **EDP² (Energy-Delay² Product)**: A variant that weights delay more heavily — EDP² = Energy × Delay² — used when performance is more important than energy, shifting the optimal voltage higher.
- **EDAP (Energy-Delay-Area Product)**: Extends EDP to include silicon area cost — EDP × Area — used when die cost is a significant factor (mobile SoCs, IoT).
- **Workload Dependence**: EDP varies with workload — compute-intensive tasks have different optimal operating points than memory-intensive tasks, motivating dynamic voltage and frequency scaling (DVFS).
| Metric | Formula | Optimizes For | Optimal Vdd | Best For |
|--------|---------|-------------|------------|---------|
| Energy | C·V² | Minimum energy | V_th (near threshold) | Ultra-low power |
| EDP | Energy × Delay | Energy-speed balance | ~0.4-0.6V | Battery devices |
| EDP² | Energy × Delay² | Performance-weighted | ~0.6-0.8V | Performance + efficiency |
| Delay | t_pd | Minimum delay | V_dd,max | Maximum performance |
**Energy-Delay Product is the fundamental efficiency metric for digital computation** — capturing the essential tradeoff between energy consumption and speed in a single number that enables fair comparison across technologies, architectures, and operating conditions, guiding the voltage scaling and design decisions that optimize semiconductor products for their target applications.