Skip to content

Hardware Fundamentals

  • Moore's law and the end of frequency scaling: why parallelism matters
  • CPU architecture recap: superscalar execution, out-of-order execution, branch prediction, speculative execution
  • SIMD concept: single instruction, multiple data, data-level parallelism
  • Vector registers and vector width: 128-bit, 256-bit, 512-bit
  • Memory bandwidth vs compute: roofline model, arithmetic intensity
  • Latency vs throughput: pipelining, instruction-level parallelism
  • Chip families overview: x86 (Intel, AMD), ARM, RISC-V, Apple Silicon
  • Thermal and power constraints: TDP, power efficiency, dark silicon