EXP-085NEGATIVE
Canonical GPU engineering / SHA-256d

P16 sync-every-two replication with complete audit

With full 2^32 exactness coverage, P16 synchronization every two launches reproduces the favorable signal of EXP-160.

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 64 audit repetitions, verify the known B32 nonce in all arms and apply all three preregistered performance gates.
01 / Question & hypothesis

What was tested?

With full 2^32 exactness coverage, P16 synchronization every two launches reproduces the favorable signal of EXP-160.

02 / Scientific basis

Why the test is meaningful

Correcting an invalid audit requires a fresh sealed replication; the earlier point estimate cannot be pooled or treated as prior evidence.

H_audit=67,108,864×64=2^32R_21=throughput(sync2)/throughput(sync1)promote iff R_21≥1.0001, CI95_low≥0.9995, retention≥0.999
P16 sync-every-two replication with complete auditVisual reading of the published metrics and gates for EXP-085; it summarizes the registered result, not a mining advantage.EXP-085 / P16 SYNC-EVERY-TWO REPLICATION WITH COMPLETE AUDITSYNC2 / SYNC10.999878×CI950.998655–1.001102NO-SYNC RETENTION0.999489
FIGURE / RESULT READINGVisual reading of the published metrics and gates for EXP-085; it summarizes the registered result, not a mining advantage.
03 / Method

How it was tested

Repeat the frozen three-arm design on four new headers, correct audit repetitions to 64 and complete 24/24 balanced benchmarks.

04 / Observed result

What happened

Sync2 / sync10.999878×
CI950.998655–1.001102
No-sync retention0.999489

All binaries found B32 nonce 3,842,969,478. Sync-two observed at 0.020290 s but reached 0.999878× versus sync-one, CI95 0.998655–1.001102, favorable on 3/4 and retained 0.999489 versus no-sync.

05 / Validation

Exactness and statistical controls

Full 2^32 coverage, zero discrepancies, exact factors, 44 registers, zero spills and 24/24 benchmarks passed before timing interpretation.

06 / Interpretation

What the result means

The invalid favorable signal did not reproduce. P16 sync-every-two is not promoted; P8-S0 remains default and P8-S1 remains an opt-in low-latency path.

Limitations

  • The result is local to registered P16 geometry.
  • It rejects promotion, not exactness.
  • Other observability primitives require new hypotheses.
07 / Reproduction

Evidence trail

Use 64 audit repetitions, verify the known B32 nonce in all arms and apply all three preregistered performance gates.

Canonical variants

CANONICAL-EXP-161PREREGISTERED-CAMPAIGNSEALED-AUDIT

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