2020 version VS 2018 version

GROMACS version:
GROMACS modification: No

Dear all,
For a system, I have the results of some long all atoms MD simulation for which I used the 2018 version of GROMACS, now I am going to conduct some other simulations on that system to study a phenomena. Now I have the possibility to also use the 2020 version known to has a better performance.
My question is if the results of the 2018 and 2020 are comparable? And would you recommend to stick to the 2018 version to be in the safe side of comparison between the results of earlier simulations and what I want to do?


Minor versions/patch releases guarantee compatibility. If you change your major version (especially one that differs by two years in time), you should verify that it produces comparable output by repeating any representative simulation(s) before switching a version mid-stream. If I were a reviewer of that study, I’d like to see such a demonstration. There’s no reason to think any of the core features change, but with so many bug fixes and feature enhancements, you never know…