RESEARCH
Published findings from verified experiments.
Research records document what FARPY tested, what the evidence showed, and where conclusions stop.
Studies
Study · FRS-2026-001
Octane Pixel Determinism Versus Runtime Instability
- Type
- Study
- ID
- FRS-2026-001
- Date
- 2026-07-20
- Status
- REJECTED BENCHMARK RESULT
- Conclusion
- REJECTED
- Evidence count
- 2
Study · FRS-2026-002
Blender GPU Runtime Reproducibility Under a Frozen Workload Contract
- Type
- Study
- ID
- FRS-2026-002
- Date
- 2026-07-20
- Status
- VERIFIED RESEARCH FINDING
- Conclusion
- VERIFIED RESEARCH FINDING
- Evidence count
- 3
Study · FQR-2026-001
Reproducibility of a Seeded Bell-State Measurement Simulator
- Type
- Study
- ID
- FQR-2026-001
- Date
- 2026-07-29
- Status
- PUBLISHED_LOCAL_SIMULATOR_RESEARCH
- Conclusion
- LOCAL_SIMULATOR_RESEARCH
- Evidence count
- 5
Research Notes
Research Note · FRN-2026-001
Why Identical Pixels Can Produce Different PNG File Hashes
- Type
- Research Note
- ID
- FRN-2026-001
- Date
- 2026-07-21
- Status
- PUBLISHED
- Conclusion
- Yes, in the documented Octane experiment. The decoded pixel payloads matched across the measured outputs, while the complete PNG file hashes differed. The canonical evidence attributed the container-level byte differences to changing PNG metadata.
- Evidence count
- 1
Research Note · FRN-2026-002
Why Runtime Variance Can Invalidate an Otherwise Correct Benchmark
- Type
- Research Note
- ID
- FRN-2026-002
- Date
- 2026-07-21
- Status
- PUBLISHED
- Conclusion
- Yes, under the tested Octane contract and configuration. The decoded pixel payloads matched, but timing did not satisfy the frozen acceptance rule. The runtime coefficient-of-variation gate was at most 3%. The best consecutive three-pass CV was 15.579055%, and additional warmups did not stabilize the measurements. FARPY therefore rejected performance publication and published no verified Octane benchmark result.
- Evidence count
- 1
Compatibility / Technical material
Compatibility study
Blender 4.1.1 compatibility testing on FARPY RTX 3090 workers
- Type
- Compatibility study
- Date
- 2026-07-29
- Status
- SCOPED COMPATIBILITY EVIDENCE
- Conclusion
- 4 of 4 registered RTX 3090 GPUs qualified for isolated Blender 4.1.1 research. Production eligibility NOT_GRANTED.
- Evidence count
- 14
Technical Brief
FARPY Public Technical Brief
- Type
- Technical Brief
- Date
- 2026-07-22
- Status
- REVIEW CANDIDATE · VERSION 0.4
- Conclusion
- Historical review-candidate public-model summary. Not an FRS result.
- Evidence count
- 1