Performance

Six of seven suites. Faster than Trino.

Differentially validated against Trino, with DuckDB's official dsdgen as an independent data oracle.

6/7suites won vs Trino 465
222/222queries pass, differentially validated
18.5×top suite speedup (TPC-E)
Where we stand

Decisive up to SF1, closing fast at SF10+

Decisive wins up to SF1 — across six of seven industry suites.

SF10+ (distributed) is closing fast: on a forced-distribution rig, TPC-DS SF1 went from 4.4x slower than Trino to 1.7x faster after the dynamic-filter pushdown fix. SSB is next — shipping build-side key sets (bloom filters) to workers.

Scale factor ≈ dataset size: SF0.1 ≈ 100 MB, SF1 ≈ 1 GB, SF10 ≈ 10 GB. The cross-scale charts plot SF0.1 / SF1 / SF10 on one time axis (log Y).

TPC-H run duration across SF0.1 / SF1 / SF10 over time SQE distributed execution: coordinator and workers
Receipts

The seven suites — what each measures, and how we do

Suite & purposeSQETrino 465SpeedupPass
TPC-E (11)Financial-trading workload: date-range filters, point lookups against trade-history, aggregations across many small dimensions. 9.3s 172.0s 18.5x 11/11
TPC-BB (10)BigBench mixed workload: structured analytics plus JSON and text operations over a retail warehouse schema. 28.0s 255.7s 9.1x 10/10
TPC-C (8 read)The read SELECTs from the TPC-C OLTP workload — point lookups against small Iceberg tables (SQE runs the read subset, not transactions). 0.41s 2.65s 6.5x 8/8
TPC-DS (99)99-query decision-support benchmark — larger and harder than TPC-H; pushes joins, window functions, and grouping sets. 13.4s 45.6s 3.4x 93/99
ClickBench (43)43 queries over one wide table — log analytics: heavy string filters, point lookups, and counts. SQE's caching shines here. 1.3s 4.46s 3.4x 43/43
TPC-H (22)22-query decision-support benchmark — the most widely benchmarked SQL workload and SQE's primary comparison target. 16.8s 26.7s 1.6x 22/22
SSB (13)Star Schema Benchmark — 13 queries over one fact joined to four dimensions; the shape data warehouses are optimised for. 8.3s 5.8s 0.70x · we trail 13/13

SSB is the one suite we still trail — lineorder's uniform foreign-key distribution defeats row-group pruning, so the runtime filter only helps at row level. We show it honestly. Full per-query method + the cross-scale charts: docs, or run it yourself via the benchmark quickstart.

The leaps

How it got fast — and still is

  1. 12 Jun 2026

    The benchmark that lied

    Differential validation with DuckDB as an independent oracle: caught 16 vacuous TPC-DS queries and a Trino rounding bug (q75).

    Read the story →
  2. 16 Apr 2026

    Our nemesis: TPC-DS q72

    Runtime-filter pushdown drops q72 from 10.7s to 0.77s.

    Read the story →
  3. 14 Apr 2026

    DataFusion 53 + vendored Iceberg fork

    ~40x faster query planning.

    Read the story →
  4. 12 Apr 2026

    Five layers of caching

    8.8x faster than Trino across the suite.

    Read the story →

Current as of 4 July 2026 · synced from the SQE source