EXP-070EXPERIMENTAL
Canonical GPU engineering / SHA-256d

Initial T512-N4 energy and thermal confirmation

T512-N4 preserves its speed gain without worsening energy per hash or thermal stability.

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
ReproducibilityDo not reuse the naive aggregate; pair both order sequences within every header as in the registered successor.
01 / Question & hypothesis

What was tested?

T512-N4 preserves its speed gain without worsening energy per hash or thermal stability.

02 / Scientific basis

Why the test is meaningful

A throughput win can be operationally inferior if it consumes more energy; measurement order and heat history must be balanced.

Eratio=(J/hash)_N4/(J/hash)_N32noninferior iff header ratios≤1.002 and CI bound passesorder∈{ABBA,BAAB}
Initial T512-N4 energy and thermal confirmationVisual reading of the published metrics and gates for EXP-070; it summarizes the registered result, not a mining advantage.EXP-070 / INITIAL T512-N4 ENERGY AND THERMAL CONFIRMATIONKERNEL RATIO1.002946×ENERGY RATIO0.982385ENERGY VERDICTINCONCLUSIVE
FIGURE / RESULT READINGVisual reading of the published metrics and gates for EXP-070; it summarizes the registered result, not a mining advantage.
03 / Method

How it was tested

Measure six unseen headers, 72 processes and NVML at 100 Hz, alternating ABBA and BAAB by header.

04 / Observed result

What happened

Kernel ratio1.002946×
Energy ratio0.982385
Energy verdictINCONCLUSIVE

Speed persisted at 1.002946× kernel, but energy noninferiority did not pass: aggregate ratio 0.982385, only 3/6 headers within gate and 95% interval 0.948442–1.017544. ABBA and BAAB groups diverged strongly.

05 / Validation

Exactness and statistical controls

The order split is reported as confounding rather than averaged away. Exact B32 and thermal gates passed.

06 / Interpretation

What the result means

Speed remains credible; energy is inconclusive because position/heat dominates the small effect. A within-header paired design is required.

Limitations

  • Order is confounded with header.
  • Energy verdict is neither confirmed nor rejected.
  • NVML is device-level measurement.
07 / Reproduction

Evidence trail

Do not reuse the naive aggregate; pair both order sequences within every header as in the registered successor.

Canonical variants

CANONICAL-EXP-146PREREGISTERED-CAMPAIGNSEALED-AUDIT

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