b3chain / Testing & Verification / BLAKE3 vs SHA-256 / Energy per hash

Energy per hash — SHA-256 vs BLAKE3

Joules per hash for each algorithm on different hardware classes, with the necessary caveats about why this is not the same as "PoW network energy".

Source: compare-energy.md Type: data-only

1. Two completely different numbers

Two things travel under the same name and are constantly confused:

  1. Algorithm energy — joules to compute one hash on hardware optimised for that algorithm. This is what the BLAKE3 paper measures.
  2. Network energy — total grid energy a chain spends per second on PoW. This depends almost entirely on coin price and electricity cost, not the algorithm. Bitcoin and a hypothetical identical-difficulty BLAKE3 chain would consume roughly the same network energy at the same reward level.

This page is about (1). For (2) see Cambridge Bitcoin Electricity Consumption Index. B3Chain's mainnet network energy is unknowable until launch.

2. Algorithm energy (published numbers)

AlgorithmHardwareThroughputJ/hash (est.)
SHA-256Antminer S21 XP Hyd (ASIC)~473 TH/s~6.4×10-15
SHA-256x86 single core (AVX2)~140 MH/s~5.0×10-9
BLAKE3x86 single core (AVX-512)~1.0 GH/s~6.5×10-10
BLAKE3NVIDIA RTX 4090~120 GH/s~2.7×10-12
BLAKE3 (no SIMD)x86 single core~250 MH/s~2.6×10-9

Headline: on the same general-purpose CPU, BLAKE3 is roughly an order of magnitude lower J/hash than SHA-256, primarily because of SIMD parallelism. On dedicated ASIC, the comparison flips because SHA-256 ASICs are highly tuned and BLAKE3 ASICs effectively don't exist.

3. What this means for B3Chain

  • Year one: B3Chain mined on commodity CPUs/GPUs is more energy-efficient per hash than Bitcoin mined on commodity CPUs/ GPUs. The total network energy is still set by the difficulty curve and the value of B3C, not by the algorithm.
  • Eventually: as BLAKE3 ASICs appear, this efficiency gap closes. At steady state both chains converge to roughly the same energy-per-USD spent on hashing, regardless of algorithm.

4. What we explicitly do NOT claim

  • "BLAKE3 is greener than SHA-256 PoW." False at the network level.
  • "B3Chain will use less energy than Bitcoin." Unknown; depends on B3C market value.
  • "BLAKE3 is more efficient." True only when comparing equivalent hardware tiers.

5. Methodology and caveats

  • ASIC J/hash is taken from manufacturer J/TH spec divided by 1012. Cooling overhead is not included (typically +5–15%).
  • CPU J/hash assumes the chip's TDP is fully spent on the hashing thread. Actual energy is somewhat lower because adjacent cores are idle.
  • We have not run our own physical-power measurement. The roadmap's "continuous benchmark CI" item includes wiring up an actual power meter to a benchmark machine.

6. Sources

BLAKE3 paper (O'Connor et al. 2020); OpenSSL speed sha256 benchmarks; Bitmain Antminer S21 XP Hyd spec; Bzminer release notes for RTX 4090 throughput; Cambridge CBECI for context. Full citations in compare-energy.md.