Odd performance difference of SSD between USB3 and SATA

[riklaunim]

I wanted to put an SSD in the place of ODD in MSI GP70 laptop. I got an SSD drive for it but noticed that not everything is as fast as on another machine with SSD. I did some benchmarks:
- http://openbenchmarking.org/result/1...SO-1403160SO60
You can see SQLite and BlogBench difference in SATA (ODD HDD caddy) and USB3 HDD case connected and booted via USB3. Bootchart also shots that it boots slower on SATA (35 vs 15 sec)

The SSD is GoodRam C50: Phison PS3108 controller, synchronous Toshiba MLC NAND
I've also tested ADATA SP800 small SSD for comparison: SandForce 1222 controller, MLC Asynchronous NAND flash memory (er... isn't that old SandForce incompatible with Haswell?) and it's fast on ODD-SATA:
- http://openbenchmarking.org/result/1...SO-SSDSATATE16 (29.99 */sparta/ dir)

Does this performance drop can be due to controller or NAND Flash type? If so which controllers/NAND types would be good ones?

[riklaunim]

It turned out to be write barrier. When disabled SQLite benchmark need 11.5 secs http://openbenchmarking.org/result/1...SO-BARRIER3039 so typical fast SSD result. It may affect data integrity, but it's a laptop... SandForce SSDs I've tested 1XXX and 2XXX don't have that problem as they are quite fast with barrier (and do SQLite benchmark below 10 sec without it).

[antonygrakham]

I used SSD with both of these types of connections and didn't see any differences concerning with speed. My disk worked fast enough in both cases.