No their results are correct. It roughly halved the compilation time on a newly generated Phoenix project. I'm assuming the savings would be more extensive on projects with multiple native dependencies that have lengthy compilation.

    rm -rf _build/ deps/ && mix deps.get && time MIX_OS_DEPS_COMPILE_PARTITION_COUNT=1 mix deps.compile
    ________________________________________________________
    Executed in   37.75 secs    fish           external
       usr time  103.65 secs   32.00 micros  103.65 secs
       sys time   20.14 secs  999.00 micros   20.14 secs

    rm -rf _build/ deps/ && mix deps.get && time MIX_OS_DEPS_COMPILE_PARTITION_COUNT=5 mix deps.compile
    ________________________________________________________
    Executed in   16.71 secs    fish           external
       usr time    2.39 secs    0.05 millis    2.39 secs
       sys time    0.87 secs    1.01 millis    0.87 secs
    
    rm -rf _build/ deps/ && mix deps.get && time MIX_OS_DEPS_COMPILE_PARTITION_COUNT=10 mix deps.compile
    ________________________________________________________
    Executed in   17.19 secs    fish           external
       usr time    2.41 secs    1.09 millis    2.40 secs
       sys time    0.89 secs    0.04 millis    0.89 secs

Similar result on one of my real projects that's heavier on the Elixir dependencies but that only has 1 additional native dependency (brotli):

    mise use elixir@1.19-otp-26 erlang@26
    
    rm -rf _build/ deps/ && mix deps.get && time MIX_OS_DEPS_COMPILE_PARTITION_COUNT=1 mix deps.compile
    ________________________________________________________
    Executed in   97.93 secs    fish           external
       usr time  149.37 secs    1.45 millis  149.37 secs
       sys time   28.94 secs    1.11 millis   28.94 secs
    
    rm -rf _build/ deps/ && mix deps.get && time MIX_OS_DEPS_COMPILE_PARTITION_COUNT=5 mix deps.compile
    ________________________________________________________
    Executed in   42.19 secs    fish           external
       usr time    2.48 secs    0.77 millis    2.48 secs
       sys time    0.91 secs    1.21 millis    0.91 secs

Oh, interesting. I guess `time` is only reporting the usr/sys time of the main process rather than the child workers when using PARTITION_COUNT higher than 1?