b3chain / Testing & Verification / BLAKE3 vs SHA-256

BLAKE3 vs SHA-256 — comparative suite

Side-by-side measurements and structural analysis of the two hash functions. Empirical scripts produce numerical output; data-only documents survey the published literature. Every claim is either measured locally and JSON-logged, or sourced.

Source: contrib/testing/compare/ Output: results/*.json Latest: see individual pages

1. Empirical comparisons (run locally)

Hash throughput

SHA-256 / SHA-256d / BLAKE3 / BLAKE3d / BLAKE3-portable measured in hashes-per-second, single thread and N-thread, on the host CPU.
compare-pow-throughput.py RUN-IT

End-to-end block validation

Wall time to -reindex 2016 regtest blocks on bitcoind vs b3chaind. Most defensible single number — whole validation path, not just hashing.
compare-block-validation.py RUN-IT

Length-extension safety

Working SHA-256 length-extension forgery against a naive MAC; same attack against BLAKE3 fails by construction. Construction-level safety, not a Bitcoin attack.
compare-length-extension.py DEMO

2. Structural / data-only comparisons

ASIC landscape

Bitcoin's mature SHA-256 ASIC market vs BLAKE3's nascent one; what this means for B3Chain's launch-time decentralisation.
compare-asic-landscape.md DATA

Energy per hash

Algorithm-level energy figures from the BLAKE3 paper plus manufacturer ASIC J/TH numbers, with the necessary caveats.
compare-energy.md DATA

Attack surface & cryptanalysis

CVE history, round-reduced cryptanalysis status, structural properties (length-extension, multi-target, side channels).
compare-attack-surface.md DATA

Collision margin (MD vs Bao)

Merkle-Damgard chained compression vs BLAKE3's tree mode — what changes structurally and why it matters for protocol design.
compare-collision-margin.md DATA

3. Run all empirical comparisons

cd b3chain
bash contrib/testing/compare/run-all-compare.sh

Each script prints a markdown table on stdout and writes a JSON file under contrib/testing/compare/results/. The hub page above is regenerated from results/latest.json when the site is deployed.

4. Honesty caveats

  • Throughput numbers vary substantially by CPU, cache size, and thermal state. Each result records cpu_model and cores so you can re-interpret.
  • The BLAKE3 throughput advantage is largest on multi-thread large-input workloads. Single-thread 80-byte hashing through a Python wrapper is dominated by call overhead.
  • The length-extension demo is about construction differences. Bitcoin's SHA-256d sidesteps the LE issue intentionally; BLAKE3 sidesteps it by design.
  • SHA-256 ASICs are a mature market with a 15-year head start. BLAKE3 ASICs are nascent. This asymmetry is part of why we picked BLAKE3 for launch security; it is not a permanent property.