Others have said some of this: random devices are not apples to apples as they may have different behavior
test urandom separately by dd from urandom to /dev/null test by reading and sending to /dev/null, bs=1m (not what you want to make fast, but interesting data point) It is possible that something in NetBSD should be improved. People here are not trying to claim that's not the case - just to make progress understanding. Also, on my apu2, using an internal mSATA SSD (exactly the same as yours), reading 1 GB from it and discarding it under NetBSD 8 amd64 (not what you are doing) gets me: $ dd if=/dev/rwd0d of=/dev/null bs=1m count=1024 1024+0 records in 1024+0 records out 1073741824 bytes transferred in 3.359 secs (319661156 bytes/sec) which looks good. (I have not tried to use an SD card on mine.)
