Skip to main content

Pick a database on evidence

We want a database choice anyone can check: the same open tests for every engine, every result public.

Find your databaseMethodology

Benchmarks

Explore your database’s speed

TPC-C on one 32-core server. Every row opens the full run report.

Open benchmarks →
LeaderboardTPC-C 200 WH · 256 VU · 30 min · 32 vCPU · 128 GB
  1. PostgreSQL 18single node5,655 tps
  2. MariaDB 11.8single node5,443 tps
  3. MySQL 8.4single node4,935 tps
  4. 4OrioleDB beta19-pg17single node4,785 tps
Latest run Nov 3, 2026All ratings →

Compatibility

“PostgreSQL-compatible” rarely is

We run the same checks on PostgreSQL and on every engine that claims its protocol.

16 of 20score under 80 of 100

Open the conformity map →
  1. alloydb-omni98.9
  2. cloudberry98.7
  3. yugabytedb97.7
  4. pgbouncer91.3
  5. cockroachdb68.2
  6. doltgres64.4
  7. serenedb62.7
  8. cedardb46.6
  9. duckgres33.8
  10. materialize31.4
  11. ydb21.5
  12. cratedb20.1
  13. risingwave18.1
  14. arcadedb15.2
  15. questdb13.7
  16. greptimedb13.3
  17. spanner-pg10.5
  18. picodata7
  19. clickhouse4.3
  20. immudb3.3

All open

Check any number yourself

The tools behind these results are open source.

StroppyThe open-source load generator: one binary, TPC workloadsStroppy ConformityDoes a database really speak the protocol it claimsGraphene CIThe control plane that brings up the stand and runs the runStroppy CloudPick a database and a workload, get a report

Don’t want to set up the machines? Try Stroppy Cloud