Results show how one solution's query (and data load) performance scales with the number of threads it is given. Instead of comparing solutions at a fixed thread count, every thread count of the selected solution is put side by side on one machine. Data resides in memory, so scaling here reflects how well the engine parallelises over CPU cores and memory bandwidth. Filters update the summary charts and detailed comparison below.
Benchmark filters
| Threads: | All |
|---|---|
| Baseline: | |
| Solution: | |
| Machine: | |
| NUMA node: | |
| Metric: | Cold Run Hot Run Second Hot Run |
| Data: | Size: Data date: |
| Threads (engine): | |
| Query Filters | |
| Instrument filter: | |
| Complexity filter: | |
| Include tags: | |
| Exclude tags: | |
Benchmark summary
| Threads |
Geometric mean of per-query time ratios relative to (lower is better). |
|
|---|---|---|
Wins per thread count
on how many of the compared queries that thread count posts the fastest time
on how many of the compared queries that thread count posts the fastest time
No results match the selected filters. Select a thread count above or use Reset filters.
Detailed Comparison
Load Times[1]
[1] The "load a partition into memory:" phase duration depends primarily on disk speed, not CPU/memory performance. It can be ignored when comparing thread counts in this in-memory benchmark.