Reference setup · in-house test
Analytics performance based on the TPC-H benchmark: PostgreSQL vs GritDB
On 22 analytics queries derived from TPC-H, GritDB was up to 43.5 times faster than PostgreSQL, and 9.5 times faster in total (SF20). GritDB also finished the 2 queries that PostgreSQL did not finish, each in just over 1 second. Both databases ran on the same server specification.
- Query with the largest speedup · Q18 · SF20 (PostgreSQL 69.6 s → GritDB 1.6 s)
- Up to 43.5 times
- SF20 · total of 20 queries (PostgreSQL 282.2 s → GritDB 29.8 s)
- 9.5 times in total
- Queries PostgreSQL did not finish (Q17 · Q20) · GritDB 1.0 s to 1.3 s
- 2 queries
In-house test (2026-09) · Amazon EC2 r8i.16xlarge (64 vCPU, 512 GB). PostgreSQL ran with the default parallel and work_mem settings (2 parallel workers per query, work_mem 4MB). Speedups are computed from times before rounding. Totals and speedups count the 20 queries that both databases finished. Speedup per query: 1.8 to 21.3 times at SF10 and 3.4 to 43.5 times at SF20. These results are derived from TPC-H. They are not official TPC-H results and cannot be compared with published TPC-H results.
The gap grows with the data
For the total of 20 queries, GritDB was 5.4 times faster at SF10 and 9.5 times faster at SF20. When the data doubled from SF10 to SF20, the total time grew 2.1 times for PostgreSQL and 1.2 times for GritDB. At SF20, GritDB still finished every one of the 22 queries within 2.6 s.
Time per query
Show time per query as a table (SF10 · SF20)Hide time per query as a table (SF10 · SF20)
| SF10 | SF20 | |||||
|---|---|---|---|---|---|---|
| Query | PostgreSQL | GritDB | Speedup | PostgreSQL | GritDB | Speedup |
| Q1 | 18.7 s | 1.4 s | 13.0 times | 37.4 s | 1.6 s | 24.0 times |
| Q2 | 2.1 s | 0.8 s | 2.5 times | 4.6 s | 1.1 s | 4.2 times |
| Q3 | 6.5 s | 1.2 s | 5.4 times | 13.8 s | 1.5 s | 9.3 times |
| Q4 | 2.2 s | 1.2 s | 1.8 times | 4.3 s | 1.3 s | 3.4 times |
| Q5 | 4.7 s | 1.2 s | 3.9 times | 6.1 s | 1.4 s | 4.3 times |
| Q6 | 2.9 s | 0.5 s | 5.5 times | 5.7 s | 0.6 s | 9.8 times |
| Q7 | 4.7 s | 1.2 s | 3.8 times | 9.4 s | 1.3 s | 7.1 times |
| Q8 | 3.9 s | 1.4 s | 2.7 times | 9.1 s | 1.6 s | 5.7 times |
| Q9 | 15.1 s | 1.7 s | 9.1 times | 34.3 s | 2.3 s | 14.6 times |
| Q10 | 5.8 s | 1.5 s | 3.8 times | 11.7 s | 2.4 s | 4.9 times |
| Q11 | 1.4 s | 0.4 s | 3.4 times | 2.8 s | 0.7 s | 3.8 times |
| Q12 | 5.4 s | 1.6 s | 3.4 times | 10.7 s | 1.7 s | 6.4 times |
| Q13 | 5.4 s | 2.3 s | 2.4 times | 11.0 s | 2.6 s | 4.3 times |
| Q14 | 2.8 s | 0.8 s | 3.6 times | 5.6 s | 1.0 s | 5.8 times |
| Q15 | 3.3 s | 0.6 s | 5.5 times | 6.6 s | 0.6 s | 10.4 times |
| Q16 | 2.0 s | 0.6 s | 3.5 times | 4.2 s | 0.8 s | 5.0 times |
| Q17 | Did not finish | 1.0 s | — | Did not finish | 1.3 s | — |
| Q18 | 32.6 s | 1.5 s | 21.3 times | 69.6 s | 1.6 s | 43.5 times |
| Q19 | 3.7 s | 1.8 s | 2.1 times | 7.3 s | 2.1 s | 3.4 times |
| Q20 | Did not finish | 1.0 s | — | Did not finish | 1.3 s | — |
| Q21 | 7.5 s | 2.4 s | 3.1 times | 19.3 s | 2.5 s | 7.6 times |
| Q22 | 3.1 s | 0.6 s | 5.2 times | 8.8 s | 1.0 s | 8.6 times |
| Total (20 queries) | 134.0 s | 24.8 s | 5.4 times | 282.2 s | 29.8 s | 9.5 times |
How we compared
PostgreSQLComparison
GritDBAnalytics sessionShow comparison conditionsHide comparison conditions
- Queries: 22 queries derived from TPC-H. Each time is the time to run one query, and totals count only the queries that both databases finished.
- GritDB ran the queries in an analytics session. GritDB's transaction engine (PostgreSQL) is the same version as the PostgreSQL it was compared with.
- Schema: CHAR columns were changed to VARCHAR. This differs from the original TPC-H schema.
- Data load time is not included in this comparison.
Show the test environmentHide the test environment
| Role | Setup | Server | CPU | Memory |
|---|---|---|---|---|
| Comparison | PostgreSQL 17.4 | Amazon EC2 r8i.16xlarge | 64 vCPU | 512 GB |
| Target | GritDB (transaction engine: PostgreSQL 17.4) | Amazon EC2 r8i.16xlarge | 64 vCPU | 512 GB |
Data: TPC-H data at SF10 (86.58 million rows) and SF20 (twice the size of SF10). Measured in 2026-09.
The CDC sync from the same test is in Adding an analytics system to an Oracle® operational database with CDC, and the setup of the two engines is in Architecture.
TPC-H is a trademark of the Transaction Processing Performance Council. Amazon EC2 is a trademark of Amazon.com, Inc. or its affiliates. Oracle is a registered trademark of Oracle and/or its affiliates. Postgres, PostgreSQL and the Slonik Logo are trademarks or registered trademarks of the PostgreSQL Community Association of Canada, and used with their permission.
See GritDB for yourself
Ask for a discovery meeting or a PoC, and we will get in touch.