Research
Measurements and published findings.
Research records document what FARPY tested, what the evidence showed, and where conclusions stop.
Studies
- Octane Pixel Determinism Versus Runtime InstabilityView study
Decoded pixels matched across runs; runtime variance failed the acceptance gate, so no performance result was published.
- Blender GPU Runtime Reproducibility Under a Frozen Workload ContractView study
The frozen Blender contract passed artifact and scoring checks. V5 has not been physically executed.
- Reproducibility of a Seeded Bell-State Measurement SimulatorView study
A seeded Bell-state simulator reproduced byte-for-byte. No physical QPU was used.
Research Notes
- Why Identical Pixels Can Produce Different PNG File HashesView study
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.
- Why Runtime Variance Can Invalidate an Otherwise Correct BenchmarkView study
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.
Compatibility / Technical material
- Blender 4.1.1 compatibility testing on FARPY RTX 3090 workersView study
4 of 4 registered RTX 3090 GPUs qualified for isolated Blender 4.1.1 research. Production eligibility NOT_GRANTED.
- FARPY Public Technical BriefView study
Historical review-candidate public-model summary. Not an FRS result.