Long direct-energy decision for P8 polling

Long direct-counter runs confirm that P8 improves throughput without worsening joules per hash beyond the preregistered margin.

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

ReproducibilityRebuild from sealed inputs and repeat four new headers, equal long work, direct joule counting, paired ordering and exact B32 audit.

01 / Question & hypothesis

What was tested?

Long direct-counter runs confirm that P8 improves throughput without worsening joules per hash beyond the preregistered margin.

02 / Scientific basis

Why the test is meaningful

Increasing work per process improves energy-counter resolution; a paired noninferiority decision requires header consistency and a corrected confidence bound below the fixed margin.

H_run=134,217,728×384R_E=(J/hash)_P8/(J/hash)_P1confirm iff R_E favorable 4/4 and CI95_upper<1.003

FIGURE / RESULT READINGVisual reading of the published metrics and gates for EXP-078; it summarizes the registered result, not a mining advantage.

03 / Method

How it was tested

Use four new headers, two exactly balanced cycles each and 64 frozen-binary runs; collect 507–517 thermal samples per process and use cumulative NVML energy as primary endpoint.

04 / Observed result

What happened

Process speedup1.003495×

Energy ratio0.997595

Corrected CI950.994435–1.000764

P8 won 4/4 headers: kernel 1.003333× and process 1.003495×. Energy/hash was 0.997595 with corrected paired CI95 0.994435–1.000764. Mean power rose 238.47→239.01 W, but shorter duration reduced energy/work; maximum was 66°C.

05 / Validation

Exactness and statistical controls

Exact SHA-256d/B32, paired order, thermal coverage and the preregistered 1.003 upper-bound gate all passed.

06 / Interpretation

What the result means

T512-N4-P8 is the confirmed local engineering default for this RTX 5080. It changes orchestration, not success probability per hash; jobs invalidating in under ~25 s need a latency policy or P1.

Limitations

Maximum host polling interval is 0.162203 s.
The finding is hardware- and software-specific.
It is not evidence of an ASIC or probabilistic mining advantage.

07 / Reproduction

Evidence trail

Rebuild from sealed inputs and repeat four new headers, equal long work, direct joule counting, paired ordering and exact B32 audit.

Canonical variants

CANONICAL-EXP-154PREREGISTERED-CAMPAIGNSEALED-AUDIT

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