[fastneutron]

All this, and the nuclear business is still using variants of LCGs from 60 years ago [1] for Monte Carlo simulations.

In this case, their speed, repeatability and “good enough” statistical properties haven’t motivated much to change.

1. https://www.osti.gov/biblio/976209

[berkut]

Which is interesting, because for MC simulations for rendering (path tracing for example) fast and high quality random numbers are very important...

Although there are alternatives like using low discrepancy sequences (Halton, Sobol) which don't explicitly need the random number values generated from a PRNG itself which are used as well...