I had similar issues as you say with read being much slower than write using a highpoint controller the fix was to change the default block size of the array to the min the controller supported.

Does anyone have a good feel for a blocksize to use on RAID 1 under FreeBSD ? I know for a system serving pparallel requests the usual advice is to have it as large as possible to try and get a single file onto a single drive, so two in parallel read from a drive each - but for large files and a single read, do you want to make it somewhat smaller to exploit parallelism between the drives for a single request ? I have a pair of drives running as RAID 1 and it is fairly obvious that for a large read it is alternating between then, rather than reading from both in parallel. The stripe size there is 64k, what are other people using ?