Engineering overview
CMOS combines complementary n-channel and p-channel devices so ideal static logic draws little DC current except during transitions and leakage. Scaling historically improved density and speed while reducing energy per function, but modern scaling is constrained by electrostatics, leakage, interconnect delay, variability, power density and the cost of manufacturing increasingly complex structures.
Device architecture has therefore evolved from planar MOSFETs toward stronger gate control, including FinFET and gate-all-around/nanosheet approaches, while system performance increasingly depends on packaging, memory hierarchy and specialized accelerators as much as transistor shrink.
Core concepts
Engineering workflow
- Separate architectural scaling goals—density, speed, energy—from simple feature-size reduction.
- Evaluate electrostatic control, leakage and variability with device models appropriate to the process node.
- Co-optimize transistor, standard-cell, interconnect and power-delivery choices.
- Use voltage/frequency, power gating and clock gating as system-level complements to device scaling.
- Treat advanced packaging and accelerators as part of the scaling strategy when monolithic improvement is insufficient.
Tradeoffs & failure modes
- Assuming every node shrink automatically improves frequency and power.
- Ignoring interconnect and memory bottlenecks.
- Overlooking leakage and variability at low supply voltages.
- Using marketing node names as physical dimensions.