An Empirical SPPR Verification of Deterministic Causal Execution Versus GPU-Based Compute
Jorge Vasconcelos · Zenodo (CERN European Organization for Nuclear Research) · 2026
We report an empirical verification of a proprietary deterministic causal execution runtime (CTI Runtime 11.9) evaluated against GPU-based CUDA execution using a latency-centric metric termed Speed Per Precise Result (SPPR). SPPR measures externally observed wall-clock time per correct causal decision under strict per-decision constraints. Unlike throughput-oriented benchmarks, this study focuses on causal, latency-dominated workloads where batching, training, and probabilistic approximation are operationally invalid. Due to security, intellectual property, and system-integrity requirements, the evaluated runtime is not publicly executable and does not accept third-party workloads. Verification is performed through deterministic behavior, fixed ground-truth evaluation, externally measured timing, and controlled audit procedures. Observed results demonstrate several orders-of-magnitude advantage in SPPR for deterministic causal execution relative to GPU-based computation under identical decision constraints. No internal mechanisms, algorithms, datasets, or executable artifacts are disclosed.