Independent replication of T512-N4
The small T512-N4 throughput signal from EXP-144 replicates on unseen headers with frozen binaries.
Diagnostic work, exact algebra and local capabilities are not treated as end-to-end mining advantage.
What was tested?
The small T512-N4 throughput signal from EXP-144 replicates on unseen headers with frozen binaries.
Why the test is meaningful
A sub-1% effect requires independent headers, balanced order, a frozen single candidate and an interval entirely above one.
R=geomean throughput_N4/throughput_N32gate R≥1.002direction≥7/8 and CI_low>1How it was tested
Freeze N4/N32 before generating eight new headers and run 64 ABBA/BAAB measurements of 134,217,728 × 96 hashes.
What happened
N4 won 8/8 headers with individual ratios 1.002507–1.004223. Geometric kernel and wall ratio was 1.003057; 95% interval 1.002573–1.003540. Every exactness gate passed.
Exactness and statistical controls
No candidate reselection or gate adaptation occurred. Energy/thermal confirmation remained a mandatory next gate.
What the result means
The small throughput improvement is independently reproducible on this RTX 5080, but is not generalized to other hardware or probability per hash.
Limitations
- Hardware-local result.
- Energy was not measured in this campaign.
- A faster GPU kernel is not an ASIC advantage.
Evidence trail
Use the sealed binaries, eight unseen headers and identical ABBA/BAAB ordering.
Canonical variants
CANONICAL-EXP-145PREREGISTERED-CAMPAIGNSEALED-AUDITSource: internally audited canonical reports. Local filesystem structure, private headers and operational identifiers are excluded from publication.