I own 3 SanDisk 4GB Titanium USB flash drives. SanDisk claims a max write rate of 9 MB/s. Under a variety of tests (FAT32 AND Linux ext3 formats), I do indeed achieve results that are very close in writing 100MB files.

I just purchased another SanDisk Titanium flash drive identical in specs except that it was 8GB capacity. Running all the same tests as with the 4GB versions, I obtained write speeds more than 4 times slower! It thought the unit was defective so I returned it.

I then bought a Corsair Voyager GT, 16GB; various reviews on the web have reported write rates near the USB 2.0 bus limit; ie, 30 MB/s. Once again, I ran many tests with different formats. The results were shocking with write speeds near 1 MB/s; that is almost an order of magnitude worse than my SanDisk 4GB drives.

Clearly, both the new 8GB & 16GB drives are NOT defective. I am absolutely clueless what is happening. USB flash drives 4GB or smaller function at their manufacturer's promised ratings while any drive larger than 4GB falls far short.

Can anyone explain this? Has anybody observed this?

I am running Slackware 13.0 32 bit version with the latest kernels, 2.6.30.5 & 2.6.29.6 on a Lenovo R61e with 2GB ram. In reading logs, proc files, sys, etc, I find no hint of trouble.

Maybe there's some issue with the filesystem?
I think that the first test you should do, is by running "hdparm -tT" against your disks. This should be independent of the particular filesystem.
Then, when making tests by file transfer, first check if the disk has been mounted with some special option.
The "sync" option, for instance, has a terrible effect on disk performance. All writes should be "async" instead.
Also, be sure to always use the same USB port when running these test. Many PCs have mixed USB1.1 and USB2.0 ports.