External ccminer baseline against M2

The local M2 kernel should outperform a fixed, externally maintained CUDA SHA-256d baseline under equivalent work.

Claim statusNO END-TO-END MINING ADVANTAGE DEMONSTRATED

Diagnostic work, exact algebra and local capabilities are not treated as end-to-end mining advantage.

Date2026-08-14

HardwareNVIDIA GeForce RTX 5080 · exact SHA-256d verifier

Run scopePreregistered campaign with sealed result and audit

ReproducibilityUse the frozen upstream commit, equal work, eight unseen headers, balanced run order and independent hashlib verification.

01 / Question & hypothesis

What was tested?

The local M2 kernel should outperform a fixed, externally maintained CUDA SHA-256d baseline under equivalent work.

02 / Scientific basis

Why the test is meaningful

A credible speed claim requires comparison with a strong external implementation, identical hash counts, independent headers and exact nonce verification.

Rsteady = throughput(M2) / throughput(ccminer)Renergy = joules/hash(M2) / joules/hash(ccminer)promotion requires exact hashes and robust Rsteady > 1

FIGURE / RESULT READINGVisual reading of the published metrics and gates for EXP-056; it summarizes the registered result, not a mining advantage.

03 / Method

How it was tested

Freeze KlausT/ccminer commit 5bd8273, preserve its arithmetic byte-for-byte, run eight fresh headers in 96 ABBA cycles, hash 134,217,728 × 96 candidates per run and sample NVML at 100 Hz.

04 / Observed result

What happened

Steady-state M2 / external0.914953

External speed lead9.30%

Hash discrepancies0

ccminer was 9.30% faster in steady state and won 8/8 datasets. M2/ccminer steady-state ratio was 0.914953. Process-inclusive timing and energy were inconclusive because initialization/JIT outliers widened their intervals. Five B32 nonces were independently verified with zero discrepancies.

05 / Validation

Exactness and statistical controls

The conclusion was manually narrowed from broad dominance to steady-state dominance without changing measurements. Process-inclusive and energy endpoints remain explicitly inconclusive.

06 / Interpretation

What the result means

M2 does not beat the frozen external baseline. ccminer becomes the operational reference; no mining-advantage claim is enabled.

Limitations

Energy and process-inclusive endpoints were confounded by startup outliers.
The result is specific to the RTX 5080 and frozen binaries.
It does not compare ASIC hardware.

07 / Reproduction

Evidence trail

Use the frozen upstream commit, equal work, eight unseen headers, balanced run order and independent hashlib verification.

Canonical variants

CANONICAL-EXP-132PREREGISTERED-CAMPAIGNSEALED-AUDIT

Source: internally audited canonical reports. Local filesystem structure, private headers and operational identifiers are excluded from publication.