Moore’s Law

Moore’s Law is the semiconductor scaling frame discussed in EP270 一枚芯片的漫长征途：我们离“算力自由”还有多远？. The episode traces it to [[GordonMoore|Gordon Moore]] in the 1960s, first as an observation about component counts on integrated circuits and later as an industry rhythm often summarized as doubling roughly every 18 to 24 months.

The source treats Moore’s Law as both technology and coordination. Regular density gains improved performance and reduced cost, making chips cheap enough for ordinary devices. But as process nodes approach physical and economic limits, the industry increasingly looks to architecture, [[AdvancedPackaging|advanced packaging]], and system-level design rather than relying on simple shrinkage.

Key Claims

• Moore’s Law helped organize semiconductor investment, roadmaps, and expectations, not just describe a neutral trend.
• Smaller process nodes can improve performance and cost, but below roughly two nanometers the episode highlights leakage, quantum effects, and difficult transistor structures.
• Leading-edge fabs now require enormous capital spending, so continuing process scaling has a weaker cost-benefit profile than in earlier eras.
• Post-Moore progress can come from packaging, memory proximity, architecture, and software-hardware co-design.

Connections

• Gordon Moore and Intel — origin and company context.
• TSMC, Samsung, and SMIC — manufacturers discussed around advanced process pursuit and catch-up.
• Photolithography Bottleneck, Advanced Packaging, and Semiconductor 3D Stacking — routes and constraints around continued scaling.
• Compute Freedom / 算力自由 — user-facing cost implication of cheaper compute.