This article describes the performance testing results of Python 3.11 compared to Python 3.10. A total of 87 various benchmark tests were conducted on computers with the AMD Ryzen 7000 series and the 13th-generation of Intel Core processors for desktops, laptops or mini PCs.
All tests were run on Windows 11 computers using the pyperformance 1.0.9 library in Python 3.10.11 and Python 3.11.6 (both 64-bit versions).
AMD Ryzen 7000 Series Desktop Processor
The first part of testing involved a desktop computer with an AMD Ryzen 9 7900 processor, RAM DDR5, and an M.2 PCIe Gen4 NVMe drive. The table below shows the results of 87 tests performed using Python 3.10 (as a reference) and Python 3.11 on this device.
| Benchmark | Python 3.10 | Python 3.11 |
|---|---|---|
| 2to3 | 216 ms | 191 ms (1.13x faster) |
| async_generators | 230 ms | 198 ms (1.16x faster) |
| async_tree_cpu_io_mixed | 685 ms | 580 ms (1.18x faster) |
| async_tree_io | 1.32 sec | 949 ms (1.39x faster) |
| async_tree_memoization | 560 ms | 420 ms (1.33x faster) |
| async_tree_none | 471 ms | 340 ms (1.39x faster) |
| asyncio_tcp | 686 ms | 708 ms (not significant) |
| asyncio_tcp_ssl | 2.04 sec | 2.02 sec (not significant) |
| bench_mp_pool | 55.4 ms | 54.9 ms (1.01x faster) |
| bench_thread_pool | 725 us | 687 us (1.05x faster) |
| chameleon | 5.89 ms | 5.23 ms (1.13x faster) |
| chaos | 57.1 ms | 45.7 ms (1.25x faster) |
| comprehensions | 14.6 us | 14.4 us (1.02x faster) |
| coroutines | 17.3 ms | 16.9 ms (1.02x faster) |
| coverage | 33.5 ms | 141 ms (4.21x slower) |
| create_gc_cycles | 637 us | 521 us (1.22x faster) |
| crypto_pyaes | 62.7 ms | 48.3 ms (1.30x faster) |
| dask | 263 ms | 235 ms (1.12x faster) |
| deepcopy | 253 us | 225 us (1.12x faster) |
| deepcopy_memo | 28.9 us | 25.1 us (1.15x faster) |
| deepcopy_reduce | 2.20 us | 2.02 us (1.09x faster) |
| deltablue | 4.27 ms | 2.63 ms (1.63x faster) |
| django_template | 28.2 ms | 23.0 ms (1.22x faster) |
| docutils | 1.63 sec | 1.39 sec (1.18x faster) |
| fannkuch | 256 ms | 240 ms (1.07x faster) |
| float | 66.7 ms | 53.5 ms (1.25x faster) |
| gc_traversal | 1.20 ms | 1.25 ms (1.04x slower) |
| generators | 41.8 ms | 41.0 ms (not significant) |
| genshi_text | 18.7 ms | 16.6 ms (1.13x faster) |
| genshi_xml | 132 ms | 95.9 ms (1.07x faster) |
| hexiom | 5.46 ms | 4.39 ms (1.25x faster) |
| html5lib | 39.3 ms | 31.2 ms (1.26x faster) |
| json_dumps | 8.09 ms | 7.72 ms (1.05x faster) |
| json_loads | 13.3 us | 12.7 us (1.05x faster) |
| logging_format | 6.74 us | 6.35 us (1.06x faster) |
| logging_silent | 102 ns | 71.6 ns (1.43x faster) |
| logging_simple | 6.36 us | 6.00 us (1.06x faster) |
| mako | 9.04 ms | 7.49 ms (1.21x faster) |
| mdp | 1.68 sec | 1.62 sec (1.04x faster) |
| meteor_contest | 65.9 ms | 65.9 ms (not significant) |
| nbody | 81.8 ms | 73.6 ms (1.11x faster) |
| nqueens | 61.3 ms | 60.3 ms (not significant) |
| pathlib | 81.3 ms | 79.9 ms (1.02x faster) |
| pickle | 6.86 us | 6.66 us (1.03x faster) |
| pickle_dict | 18.5 us | 18.3 us (1.01x faster) |
| pickle_list | 2.58 us | 2.59 us (not significant) |
| pickle_pure_python | 258 us | 191 us (1.35x faster) |
| pidigits | 136 ms | 136 ms (not significant) |
| pprint_pformat | 1.17 sec | 983 ms (1.19x faster) |
| pprint_safe_repr | 568 ms | 479 ms (1.19x faster) |
| pyflate | 387 ms | 293 ms (1.32x faster) |
| python_startup | 17.4 ms | 15.2 ms (1.15x faster) |
| python_startup_no_site | 13.2 ms | 12.5 ms (1.05x faster) |
| raytrace | 275 ms | 204 ms (1.35x faster) |
| regex_compile | 89.5 ms | 77.1 ms (1.16x faster) |
| regex_dna | 111 ms | 103 ms (1.07x faster) |
| regex_effbot | 1.78 ms | 1.64 ms (1.09x faster) |
| regex_v8 | 15.0 ms | 14.7 ms (1.02x faster) |
| richards | 41.7 ms | 29.8 ms (1.40x faster) |
| richards_super | 50.9 ms | 37.0 ms (1.37x faster) |
| scimark_fft | 223 ms | 214 ms (1.04x faster) |
| scimark_lu | 94.3 ms | 68.7 ms (1.37x faster) |
| scimark_monte_carlo | 58.1 ms | 44.2 ms (1.32x faster) |
| scimark_sor | 105 ms | 80.4 ms (1.30x faster) |
| scimark_sparse_mat_mult | 3.47 ms | 3.29 ms (1.05x faster) |
| spectral_norm | 82.1 ms | 75.7 ms (1.08x faster) |
| sqlglot_normalize | 198 ms | 178 ms (1.11x faster) |
| sqlglot_optimize | 37.1 ms | 32.3 ms (1.15x faster) |
| sqlglot_parse | 1.12 ms | 891 us (1.26x faster) |
| sqlglot_transpile | 1.33 ms | 1.06 ms (1.25x faster) |
| sqlite_synth | 1.54 us | 1.41 us (1.09x faster) |
| sympy_expand | 265 ms | 248 ms (1.07x faster) |
| sympy_integrate | 12.7 ms | 11.8 ms (1.07x faster) |
| sympy_str | 162 ms | 152 ms (1.06x faster) |
| sympy_sum | 91.4 ms | 86.4 ms (1.06x faster) |
| telco | 3.80 ms | 3.91 ms (1.03x slower) |
| tomli_loads | 1.69 sec | 1.46 sec (1.15x faster) |
| tornado_http | 98.2 ms | 86.9 ms (1.13x faster) |
| typing_runtime_protocols | 314 us | 307 us (1.02x faster) |
| unpack_sequence | 39.3 ns | 35.1 ns (1.12x faster) |
| unpickle | 8.59 us | 8.13 us (1.06x faster) |
| unpickle_list | 2.87 us | 2.86 us (not significant) |
| unpickle_pure_python | 176 us | 150 us (1.18x faster) |
| xml_etree_generate | 55.2 ms | 53.6 ms (1.03x faster) |
| xml_etree_iterparse | 57.6 ms | 60.0 ms (1.04x slower) |
| xml_etree_parse | 83.4 ms | 82.3 ms (1.01x faster) |
| xml_etree_process | 44.4 ms | 38.4 ms (1.16x faster) |
| Result (geometric mean) | 1.11x faster |
The result shows that Python 3.11 has the best performance results over Python 3.10 in the following tests: deltablue (1.63x faster), logging_silent (1.43x faster), richards (1.40x faster). However, you may notice a drop in performance in some tests, especially in coverage (4.21x slower), sqlglot_parse (1.26x faster), gc_traversal (1.04x slower).
Additionally, you can examine the performance differences between Python 3.11 and Python 3.10 based on benchmarks belonging to specific groups. The table below shows the geometric mean for benchmarks from separate groups for Python 3.11 compared to Python 3.10.
| Group of benchmarks | Python 3.11 to Python 3.10 |
|---|---|
| apps | 1.16x faster |
| asyncio | 1.32x faster |
| math | 1.12x faster |
| regex | 1.08x faster |
| serialize | 1.07x faster |
| startup | 1.10x faster |
| template | 1.16x faster |
13th Gen Intel Core Mobile Processor
The second part of testing involved a mini PC with an Intel Core i3-1315U Processor (which is also used in laptops), RAM DDR4, and an M.2 PCIe Gen4 NVMe drive. The table below shows the results of 87 tests performed using Python 3.10 (as a reference) and Python 3.11 on this device.
| Benchmark | Python 3.10 | Python 3.11 |
|---|---|---|
| 2to3 | 282 ms | 250 ms (1.13x faster) |
| async_generators | 254 ms | 208 ms (1.22x faster) |
| async_tree_cpu_io_mixed | 716 ms | 615 ms (1.16x faster) |
| async_tree_io | 1.16 sec | 864 ms (1.34x faster) |
| async_tree_memoization | 593 ms | 463 ms (1.28x faster) |
| async_tree_none | 486 ms | 373 ms (1.30x faster) |
| asyncio_tcp | 981 ms | 966 ms (not significant) |
| asyncio_tcp_ssl | 2.43 sec | 3.22 sec (1.33x slower) |
| bench_mp_pool | 89.5 ms | 87.2 ms (1.03x faster) |
| bench_thread_pool | 1.24 ms | 1.13 ms (1.09x faster) |
| chameleon | 6.47 ms | 6.01 ms (1.08x faster) |
| chaos | 68.8 ms | 55.7 ms (1.24x faster) |
| comprehensions | 18.4 us | 18.4 us (not significant) |
| coroutines | 18.0 ms | 16.8 ms (1.07x faster) |
| coverage | 33.0 ms | 163 ms (4.95x slower) |
| create_gc_cycles | 888 us | 799 us (1.11x faster) |
| crypto_pyaes | 70.3 ms | 54.9 ms (1.28x faster) |
| dask | 431 ms | 379 ms (1.14x faster) |
| deepcopy | 293 us | 282 us (1.04x faster) |
| deepcopy_memo | 32.0 us | 28.9 us (1.11x faster) |
| deepcopy_reduce | 2.47 us | 2.35 us (1.05x faster) |
| deltablue | 4.71 ms | 3.02 ms (1.56x faster) |
| django_template | 33.3 ms | 27.6 ms (1.21x faster) |
| docutils | 2.19 sec | 1.86 sec (1.18x faster) |
| fannkuch | 306 ms | 283 ms (1.08x faster) |
| float | 70.0 ms | 62.2 ms (1.13x faster) |
| gc_traversal | 1.69 ms | 1.80 ms (1.07x slower) |
| generators | 36.0 ms | 38.2 ms (1.06x slower) |
| genshi_text | 21.6 ms | 19.5 ms (1.11x faster) |
| genshi_xml | 153 ms | 118 ms (1.05x faster) |
| hexiom | 6.22 ms | 5.19 ms (1.20x faster) |
| html5lib | 56.0 ms | 45.5 ms (1.23x faster) |
| json_dumps | 9.76 ms | 8.76 ms (1.11x faster) |
| json_loads | 16.0 us | 16.3 us (1.02x slower) |
| logging_format | 8.15 us | 7.67 us (1.06x faster) |
| logging_silent | 106 ns | 81.4 ns (1.30x faster) |
| logging_simple | 7.60 us | 7.25 us (1.05x faster) |
| mako | 9.78 ms | 8.19 ms (1.19x faster) |
| mdp | 1.96 sec | 1.93 sec (1.01x faster) |
| meteor_contest | 83.7 ms | 85.9 ms (1.03x slower) |
| nbody | 80.5 ms | 82.0 ms (not significant) |
| nqueens | 75.9 ms | 73.8 ms (1.03x faster) |
| pathlib | 77.5 ms | 73.4 ms (1.06x faster) |
| pickle | 7.85 us | 7.71 us (1.02x faster) |
| pickle_dict | 20.9 us | 22.0 us (1.05x slower) |
| pickle_list | 3.11 us | 3.23 us (1.04x slower) |
| pickle_pure_python | 302 us | 229 us (1.32x faster) |
| pidigits | 167 ms | 169 ms (1.01x slower) |
| pprint_pformat | 1.39 sec | 1.19 sec (1.17x faster) |
| pprint_safe_repr | 676 ms | 581 ms (1.16x faster) |
| pyflate | 461 ms | 359 ms (1.28x faster) |
| python_startup | 25.0 ms | 23.1 ms (1.08x faster) |
| python_startup_no_site | 19.4 ms | 19.8 ms (1.02x slower) |
| raytrace | 320 ms | 233 ms (1.38x faster) |
| regex_compile | 118 ms | 102 ms (1.16x faster) |
| regex_dna | 148 ms | 136 ms (1.09x faster) |
| regex_effbot | 1.88 ms | 1.72 ms (1.10x faster) |
| regex_v8 | 17.0 ms | 15.5 ms (1.10x faster) |
| richards | 46.6 ms | 35.0 ms (1.33x faster) |
| richards_super | 58.5 ms | 43.1 ms (1.36x faster) |
| scimark_fft | 214 ms | 212 ms (1.01x faster) |
| scimark_lu | 93.7 ms | 71.5 ms (1.31x faster) |
| scimark_monte_carlo | 64.1 ms | 53.3 ms (1.20x faster) |
| scimark_sor | 114 ms | 87.5 ms (1.30x faster) |
| scimark_sparse_mat_mult | 2.97 ms | 2.97 ms (not significant) |
| spectral_norm | 84.8 ms | 77.2 ms (1.10x faster) |
| sqlglot_normalize | 234 ms | 219 ms (1.07x faster) |
| sqlglot_optimize | 44.8 ms | 40.6 ms (1.10x faster) |
| sqlglot_parse | 1.40 ms | 1.07 ms (1.30x faster) |
| sqlglot_transpile | 1.67 ms | 1.31 ms (1.28x faster) |
| sqlite_synth | 2.16 us | 1.97 us (1.10x faster) |
| sympy_expand | 367 ms | 350 ms (1.05x faster) |
| sympy_integrate | 17.4 ms | 15.9 ms (1.09x faster) |
| sympy_str | 223 ms | 215 ms (1.04x faster) |
| sympy_sum | 124 ms | 118 ms (1.05x faster) |
| telco | 4.42 ms | 4.60 ms (1.04x slower) |
| tomli_loads | 1.89 sec | 1.60 sec (1.18x faster) |
| tornado_http | 140 ms | 116 ms (1.20x faster) |
| typing_runtime_protocols | 378 us | 377 us (not significant) |
| unpack_sequence | 44.9 ns | 52.6 ns (1.17x slower) |
| unpickle | 9.21 us | 9.04 us (1.02x faster) |
| unpickle_list | 3.16 us | 2.99 us (1.06x faster) |
| unpickle_pure_python | 209 us | 173 us (1.21x faster) |
| xml_etree_generate | 61.8 ms | 62.2 ms (not significant) |
| xml_etree_iterparse | 74.4 ms | 75.5 ms (1.01x slower) |
| xml_etree_parse | 114 ms | 116 ms (1.02x slower) |
| xml_etree_process | 49.5 ms | 42.7 ms (1.16x faster) |
| Result (geometric mean) | 1.09x faster |
The result shows that Python 3.11 has the best performance results over Python 3.10 in the following tests: deltablue (1.56x faster), raytrace (1.38x faster), richards_super (1.36x faster). However, you may notice a drop in performance in some tests, especially in coverage (4.95x slower), asyncio_tcp_ssl (1.33x slower), unpack_sequence (1.17x slower).
Additionally, you can examine the performance differences between Python 3.11 and Python 3.10 based on benchmarks belonging to specific groups. The table below shows the geometric mean for benchmarks from separate groups for Python 3.11 compared to Python 3.10.
| Group of benchmarks | Python 3.11 to Python 3.10 |
|---|---|
| apps | 1.16x faster |
| asyncio | 1.27x faster |
| math | 1.03x faster |
| regex | 1.11x faster |
| serialize | 1.06x faster |
| startup | 1.03x faster |
| template | 1.14x faster |