Reproduce “45-year CPU evolution: one law and two equations” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Daniel Etiemble
Daniel Etiemble's paper, from LRI-CNRS, University Paris Sud, provides a unifying framework to explain the major shifts in microprocessor design since the 1970s. It demonstrates how Moore's Law, the CPU Performance Equation, and the CMOS Power Dissipation Equation collectively account for the transition from frequency scaling to multicore architectures driven by power constraints.
Key insights
- The "heat wall," resulting from escalating power density, forced the industry's pivot from increasing clock frequency to pursuing multicore designs.
- Instruction-level parallelism (ILP) and memory latency (the "memory wall") reached diminishing returns, necessitating complex cache hierarchies and eventually driving the move to explicit parallelism.
- CPU evolution can be fundamentally understood as a continuous effort to optimize execution time while respecting power constraints, all enabled by Moore's Law.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.