https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=244356
--- Comment #87 from Gary Jennejohn <[email protected]> --- (In reply to Olivier Certner from comment #86) I only used dd for my testing. A file system adds lots of overhead and does not give a clear picture of the raw performance. I used a Kingston DT Rubber 3.0ge 16GB on both FreeBSD and Linux. I used it because it's easy to fill up such a small stick. Unfortunately, the FreeBSD XHCI driver does not correctly attach the stick (probably a bug) whereas USB2 has no problems with it. Under Linux XHCI is happy to use it. This is a rather old stick (9 or 10 years) and Kingston claims that it can only write at a maximum of 30MB/s. A test report I saw showed that the stick can actually get up to about 40MB/s. For the dd test I wrote two different approximately 16GB files to the stick one after the other. I tried both bs=128k and bs=256k, but that had no effect on the transfer rates under FreeBSD or Linux. In any case, whether the stick was filled with 0xFF or all sectors were already dirty didn't change the write transfer rates. Under FreeBSD (USB2) I saw about 33MB/s and under Linux (USB3) I saw about 40MB/s for each of the files. I also used the 128GB (USB3) stick with same dd tests under FreeBSD and also saw on the order of 40MB/s in that test. This stick probably was not full. So, I guess the poor write performance isn't caused by the contoller in the stick. It's probably more likely that the filesystem overhead is the culprit, especially with UFS when the stick is getting full or when it's writing small files into fragments. -- You are receiving this mail because: You are the assignee for the bug. _______________________________________________ [email protected] mailing list https://lists.freebsd.org/mailman/listinfo/freebsd-usb To unsubscribe, send any mail to "[email protected]"
