One question.
Several kinds of data.
10.8 milliseconds.
Which ships, owned through a company’s subsidiaries, had at least four consecutive hourly readings above 85°C inside a selected North Sea area during July?
Both engines returned the same ships and every qualifying timestamp. 8DB answered about 30× faster than the tuned SurrealDB path.
Linear scale from zero · same host · four CPUs per server
213 MiB vs 831 MiB at the post-query checkpoint
Synthetic, memory-resident workload. Custom Rust composition vs SurrealQL; preparation excluded.
How the query was measured
The typed time-series route was measured on an AMD Ryzen 9 9950X3D workstation under WSL Ubuntu 24.04, with the same four logical CPUs allowed for each server. Each comparison used 20 timed requests per engine. The timer includes the HTTP request and complete decoded JSON answer.
8DB used native typed operations composed in Rust, with temperature-first screening and selective location reads. SurrealDB 3.2.4 used SurrealQL and its recommended record-range layout. Neither path required authentication or durable completion. These timings do not establish a common snapshot during concurrent cross-model updates.
The memory figures come from this same comparison, not a second independent experiment. They include process runtime and retained allocations, but exclude the client and operating-system cache. They are an observed checkpoint, not a minimum device-memory requirement or an energy measurement.
Inspect tuning, correctness checks and execution choices ↗