[
Anandtech review of the 2016 MacBook]
Read speed is reduced by 31% and write speed by 11%.
I know the 2017 model has different faster storage but reductions will occur, but not necessarily the same percentages. I have seen quite wide range of reported results for the effect of Filevault, including larger reductions.
My apologies for reviving this old thread, but as the 2017 rMB is now available at more affordable prices, more user may face this problem:
With my 2017 m3 I saw read speeds of ~350MB/s and write speeds of ~700MB/s with Blackmagic Disk Speed under High Sierra. I would have expected write speeds to take the bigger hit, but it seems that read speeds do suffer more (and significantly so).
Turning off FileVault, write speeds went up to ~1.050MB/s and read speeds up to ~1.350MB/s. The test was far from being scientific or even running for a statistically relevant amount of time, but I think it gives a good ballpark impression.
I’m not into the technical details, but - assuming the implementation is not flawed in general - reading & decrypting seems to take significantly more CPU power than writing & encrypting. And on the rMB macOS may be CPU-limited for that task. Unfortunately I have disabled FileVault already on the 2017, so I can’t look at the CPU load during testing, but there is an interesting side note: on the 2015 m5 with FileVault enabled, I get write speeds of ~350MB/s and read speeds of ~750MB/s.
So roughly twice the read speeds with FileVault on the older machine with a (supposedly) slower SSD. Could it be that the m3 in the 2017 rMB is lacking some functionality which is present on the m5 from 2015 and that is relevant for FileVault speeds? Can anyone with a 2017 rMB and i5 or i7 run a BlackMagic test with FileVault enabled?