With many small files, the OS overhead, plus opening/closing files can be very significant, as you just found it out.
For DTR tests, you need to use a single big file, or even better, a dedicated test app. In one of my job, we had a windows only device, and we tested its speed by copying files to it from DOS or Unix. And it was about 20% - 25% faster from Unix!!
Yeah, I thought of that too, and that's not it. I copied the VOB from disk to disk and it took about 30 seconds, which is in the same ballpark as the Corsair. Copying the large directory structure from disk to disk took about 85 seconds. That's a huge difference from copying it to the flash drive, yet it involves the file accesses, opening, and closing. That's why I thought the write cache would make a significant difference with the flash drive.
The problem *isn't* the transfer rate. The VOB file proves that the Corsair drive performs to its specification. The problem has to do with the way Windows handles all those files and directories. Perhaps the FAT file system has something to do with it. That's not very easy to test as I don't have a disk that's FAT formatted.
