Why MySQL 5.7 Still Beats 8.x in Real‑World Benchmarks – A Deep Dive

1. Test

Percona monitors MySQL version distribution and observed that many users remain on 5.7, prompting an extended EOL support plan and a comprehensive performance study.

Assumption

The tests assume a fixed MySQL server configuration so that version differences are the primary variable. Both the default configuration and a set of tuned parameters (see the configuration‑change file) are evaluated.

What tests were run?

Two benchmark suites were used:

sysbench – read/write workload.

TPC‑C‑like – transaction‑heavy workload with joins, GROUP BY, and sorting.

Full test methodology and scripts are available in the linked GitHub repositories (e.g., https://github.com/Tusamarco/benchmarktools).

2. Results

All raw results are stored in the public repository (

https://github.com/Tusamarco/blogs/tree/master/sakila_where_are_you_going

). The article focuses on the sysbench read/write and TPC‑C outcomes because they reflect typical MySQL usage.

Key observations from the charts (small vs. large data sets, default vs. optimized configs):

With the default configuration, MySQL 5.7 outperforms MySQL 8.0.36 and 8.4 in both small and large data sets.

Optimizing parameters allows MySQL 8.0.36 to exceed 5.7’s performance and approach 8.4’s potential.

MySQL 8.4 shows a clear performance drop under the default settings; only after tuning does it become competitive.

Percona Server 8.0.36 is the only version that handles high contention well in the TPC‑C tests.

MariaDB 11.3 consistently lags behind the MySQL variants.

3. What do these tests show?

The data confirms a common user experience: performance tends to degrade as MySQL versions advance, despite new features in 8.x. Quantifying the loss with TPC‑C throughput reveals substantial drops (e.g., up to X % lower TPS for MySQL 8.4 under RR isolation).

Consequently, many users have valid reasons to stay on 5.7, as migrating can risk significant performance penalties.

4. Some thoughts

The author reflects on MySQL’s history, Oracle’s stewardship, and the impact of cloud providers that profit from MySQL without contributing back to the open‑source code. The piece argues for more open development, investment, and possibly moving to alternatives such as PostgreSQL when performance cannot be maintained.

References

Extended EOL support: https://www.percona.com/post-mysql-5-7-eol-support

Configuration changes: https://github.com/Tusamarco/blogs/blob/master/sakila_where_are_you_going/config_changes.txt

sysbench: https://github.com/akopytov/sysbench

TPC‑C Like: https://www.tpc.org/tpcc/

Test plan and details: https://github.com/Tusamarco/benchmarktools/blob/main/docs/plan.md

sysbench script: https://github.com/Tusamarco/benchmarktools/blob/main/software/fill_sysbench_map.sh

TPC‑C script: https://github.com/Tusamarco/benchmarktools/blob/main/software/fill_tpcc_map.sh

All results: https://github.com/Tusamarco/blogs/tree/master/sakila_where_are_you_going

Article on MySQL 8.4 defaults: https://lefred.be/content/mysql-8-4-lts-new-production-ready-defaults-for-innodb/

Can Oracle save MySQL?: https://www.percona.com/blog/can-oracle-save-mysql/