On slide 54, the difference in speed should not be this great. I forgot that I was using a beta version of hashcat, which no longer automatically performs some optimizations in non-benchmark mode, in exchange for support for longer passwords. The previous default optimizations can be enabled with -O.
$ hashcat -m 0 -a 3 bslv.hash -d 1 --quiet --speed-only --machine-readable ?a?a?a?a?a?a?a?a
1:3273837251
$ hashcat -O -m 0 -a 3 bslv.hash -d 1 --quiet --speed-only --machine-readable ?a?a?a?a?a?a?a?a
1:10105820035