Thank you for your input!

Running fstrim on a mounted filesystem returns “the discard operation is not supported”, and lsblk -D shows no discard capability on the device. Performance is as expected, but this suggests that TRIM is unavailable in this configuration, likely due to the USB–SATA bridge rather than the NAND or controller itself.

For users planning long-term use on Linux systems, especially for sustained writes, this may be a relevant consideration.

Result is still the same. One test is tested using PCIe lanes and the other SATA. SATA III has a theoretical maximum of 6Gb/s or 600MB/s, but we have never seen that reached, regardless of what port you use. Thanks for jumping in.

How reliable are these test results even when certain values seem to be purely random? It makes no sense that the RND4K Q1T1 write speed is higher when using a slower connection.

At least the other values for read and write speed look reasonable and it’s understandable that the sequential read and write speed goes down when you use a slower USB connection. I guess you were using a USB-A port with USB 3.2 Gen 1 which has a max bandwidth of 5 Gbit/s (625 MB/s).

However USB-A is not limited to 5 Gbit/s. It can have a bandwidth up to 10 Gbit/s if your laptop has a USB 3.2 Gen 2 port. So using USB-A does not necessarily lead to slower read and write speeds.