EXP-072NEGATIVE
Canonical GPU engineering / SHA-256d

Fine NPT neighborhood around T512-N4

An adjacent integer NPT value 3, 5, 6 or 7 improves the confirmed NPT4 default by at least 0.2%.

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 common divisible workload and compare every neighbor against the same frozen NPT4 binary.
01 / Question & hypothesis

What was tested?

An adjacent integer NPT value 3, 5, 6 or 7 improves the confirmed NPT4 default by at least 0.2%.

02 / Scientific basis

Why the test is meaningful

Local-neighborhood testing checks whether a selected discrete geometry is a real local optimum without reopening a broad tuning search.

candidates NPT={3,5,6,7}baseline NPT=4gate R≥1.002 with robust direction
Fine NPT neighborhood around T512-N4Visual reading of the published metrics and gates for EXP-072; it summarizes the registered result, not a mining advantage.EXP-072 / FINE NPT NEIGHBORHOOD AROUND T512-N4BEST NEIGHBOR RATIO1.000374×95% INTERVAL0.999283–1.001466DATASETS PASSING0 / 4
FIGURE / RESULT READINGVisual reading of the published metrics and gates for EXP-072; it summarizes the registered result, not a mining advantage.
03 / Method

How it was tested

Use TPB512, a common exactly divisible batch of 134,184,960 hashes, four unseen headers and 40 measurements.

04 / Observed result

What happened

Best neighbor ratio1.000374×
95% interval0.999283–1.001466
Datasets passing0 / 4

NPT5 was best at 1.000374×, 95% interval 0.999283–1.001466, positive on 3/4 headers and above the gate on 0/4. All exactness and resource gates passed.

05 / Validation

Exactness and statistical controls

The confirmed NPT4 binary remained fixed; no neighbor was promoted from a sub-gate maximum.

06 / Interpretation

What the result means

T512-N4 remains the local optimum among registered integer neighbors.

Limitations

  • Only NPT 3–7 around the local point was tested.
  • A different mechanism remains possible.
  • No new energy test is warranted.
07 / Reproduction

Evidence trail

Use the common divisible workload and compare every neighbor against the same frozen NPT4 binary.

Canonical variants

CANONICAL-EXP-148PREREGISTERED-CAMPAIGNSEALED-AUDIT

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