[Benchmark] Sub-10µs RNA Ground-State Energy Minimization via Physical Silicon Ising Relaxation

Hi everyone,

We wanted to share an open empirical benchmark evaluating RNA secondary structure prediction accelerated by analog in-materio silicon charge relaxation rather than conventional dynamic programming.

Using a canonical 64-nt transfer RNA (tRNA) sequence, we mapped Turner thermodynamic base-pairing and nearest-neighbor stacking rules into an interacting Ising Hamiltonian matrix (739 coupled interaction edges). When evaluated on the room-temperature NEOMA P-CAM 003 silicon enclave:

  1. Physical Reflex Latency: 0.0061 ms (6.1 µs) hardware reflex (~2,000x acceleration vs. standard dynamic programming loops).
  2. Ground-State Free Energy: ΔG = -118.70 kcal/mol with overdamped stability (zeta = 1.0254).
  3. Attestation Proof: Immutable SEAL-PoPW cryptographic token bound directly to silicon UID.

Hardware Attestation:
SEAL-0051003C3434510C32333934-6764398500352925D88B28138BD11F6DEB36616CC7ACBCFF687EBD781D2CE123

The full reproducibility suite, open verification script, and raw hardware logs are available for peer audit:

We welcome feedback, questions, and larger blind benchmark targets from the Eterna community!