EXP-076EXPERIMENTAL
Canonical GPU engineering / SHA-256d

Externally balanced P8 energy measurement

P8 preserves its confirmed throughput gain while remaining energy-noninferior when every header receives perfectly balanced external run order.

Claim statusNO END-TO-END MINING ADVANTAGE DEMONSTRATED

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

Date2026-08-16
HardwareNVIDIA GeForce RTX 5080 · exact CUDA SHA-256d verifier
Run scopePreregistered paired campaign with sealed validation and audit
ReproducibilityUse the sealed P1/P8 binaries, four balanced cycles per header, identical work and independent B32 verification; compare with direct cumulative-energy measurement.
01 / Question & hypothesis

What was tested?

P8 preserves its confirmed throughput gain while remaining energy-noninferior when every header receives perfectly balanced external run order.

02 / Scientific basis

Why the test is meaningful

Power integration over short processes is noisy; pairing and equal ABBA→BAAB / BAAB→ABBA exposure remove positional confounding but cannot remove telemetry resolution error.

R_E=(J/hash)_P8/(J/hash)_P1noninferiority: CI95_upper(R_E)≤1.003external order balance=2 ABBA→BAAB+2 BAAB→ABBA per header
Externally balanced P8 energy measurementVisual reading of the published metrics and gates for EXP-076; it summarizes the registered result, not a mining advantage.EXP-076 / EXTERNALLY BALANCED P8 ENERGY MEASUREMENTKERNEL SPEEDUP1.003033×ENERGY RATIO0.999959ENERGY CI950.989204–1.010830
FIGURE / RESULT READINGVisual reading of the published metrics and gates for EXP-076; it summarizes the registered result, not a mining advantage.
03 / Method

How it was tested

Run four cycles per header, 128 total processes and 100 Hz NVML sampling with frozen exact P1/P8 binaries; audit B32, sample count and temperature.

04 / Observed result

What happened

Kernel speedup1.003033×
Energy ratio0.999959
Energy CI950.989204–1.010830

P8 retained 1.003033× kernel and 1.002749× process speed. Energy/hash was 0.999959, but only 2/4 headers passed and CI95 was 0.989204–1.010830. Runs had 137–141 samples and reached at most 66°C.

05 / Validation

Exactness and statistical controls

External position was exactly balanced and SHA-256d/B32 checks passed. The wide interval is therefore retained as measurement uncertainty, not averaged into a favorable claim.

06 / Interpretation

What the result means

Throughput, exactness and thermal safety remain confirmed; energy noninferiority is unresolved. The next test must use the GPU cumulative energy counter.

Limitations

  • 100 Hz integration is low-resolution for ~2.2 s processes.
  • Energy passed on only two of four headers.
  • The result is local to one RTX 5080 and frozen binaries.
07 / Reproduction

Evidence trail

Use the sealed P1/P8 binaries, four balanced cycles per header, identical work and independent B32 verification; compare with direct cumulative-energy measurement.

Canonical variants

CANONICAL-EXP-152PREREGISTERED-CAMPAIGNSEALED-AUDIT

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