I wrote a silly little testcase: https://gist.github.com/jcalvinowens/06530e6fc86f3b2c97a7027...

...but both affected znver5 machines I have "pass", so clearly there's more to it than AMD is telling us.

You should be able to get away with "-march=znver5 -mno-rdrnd -mno-rdseed" in CFLAGS if you care.

See also: https://lore.kernel.org/lkml/20251103142800.GFaQi78AgT-qjY30...

again? https://arstechnica.com/gadgets/2019/10/how-a-months-old-amd... ZEN seems to have suspiciously a lot of those rdrand "bugs"