With the simple benchmark script on GitHub, below is the performance scores on Windows

Access mdb database

The seconds used to operate 20,000 records on my Lenovo Dual-Core Celeron 2.1G , 3G RAM laptop :

pypyodbc on CPython 2.7.6

Running with pypyodbc 1.3.0

Write time: 9.97799992561 R & W time: 11.7869999409 Read time: 0.910000085831

pypyodbc on CPython 2.7.6

Running with pypyodbc 1.2.1

Write time: 9.7380001545 R & W time: 12.2130000591 Read time: 1.60899996758

pyodbc on CPython 2.7.6

Running with pyodbc 3.0.6

Write time: 12.9730000496 R & W time: 12.5540001392 Read time: 0.18799996376

pypyodbc on PyPy 2.3 (after JIT started)

Running with pypyodbc 1.3.2 odbcjt32.dll

Write time: 7.72200012207 R & W time: 9.39100003242 Read time: 0.483999967575

pypyodbc on PyPy 2.2 (after JIT started)

Running with pypyodbc 1.2.1

Write time: 9.57299995422 R & W time: 10.9049999714 Read time: 0.707999944687

pypyodbc on IronPython 2.7.4

Running with pypyodbc 1.2.1

Write time: 10.2185821533 R & W time: 15.8019104004 Read time: 2.62014770508

SQLServer

The seconds used to operate 20,000 records:

pypyodbc on CPython 2.7.6

Running with pypyodbc 1.3.0 sqlncli10.dll

Write time: 13.2389998436 R & W time: 15.9130001068 Read time: 0.920000076294

pypyodbc on CPython 2.7.6

Running with pypyodbc 1.2.1 sqlncli10.dll

Write time: 12.4639999866 R & W time: 15.6489999294 Read time: 1.54399991035

pyodbc on CPython 2.7.6

Running with pypyodbc 3.0.6 sqlncli10.dll

Write time: 10.4210000038 R & W time: 10.6079998016 Read time: 0.141000032425

pypyodbc on PyPy 2.2 (after JIT started)

Running with pypyodbc 1.3.0 sqlncli10.dll

Write time: 12.7769999504 R & W time: 15.1629998684 Read time: 0.470999956131

pypyodbc on PyPy 2.2 (after JIT started)

Running with pypyodbc 1.2.1 sqlncli10.dll

Write time: 12.1679999828 R & W time: 14.9449999332 Read time: 0.763999938965