Legacy fixture · not a performance claim

Consumer Hardware Benchmark Audit

The earlier page presented a prescribed 15-femtosecond timestep as a discovered NNC speedup. The source script chooses that timestep in advance, so the resulting step-count reduction is true by construction and does not establish equal accuracy, stability, or HPC equivalence.

Retracted claims

  • Universal optimal k for any physics problem
  • HPC-class capability from the timing fixture
  • Same-accuracy speedup from a larger imposed timestep

What the demo can show

  • A browser-friendly soft-core molecular illustration
  • Session-local stability and force metrics
  • Why prescribed timestep ratios require a controlled convergence study

Required next benchmark

Compare both solvers over the same physical duration, sweep timestep independently, report energy drift and trajectory error against a converged reference, include warm-up, repeat timings, and publish the raw artifact. Until then, the old timing arrays remain an unversioned legacy fixture.