These results looks pretty odd. Did the test have trim enabled? Which scheduler was used? Was SATA set to IDE or AHCI? noatime set?
i got some odd results with my raidz zfs 3xsamsung 840 evo on these tests too, i didn't upload my results in the end, as i don't think it's a proper way to benchmark real world usage at the same time.
the samsung 840 evo's only do about 135mb/sec write speed when you write more than 3gb of data, that these tests do.
that said, i still haven't figure out best way to benchmark real world usage. i found something to test read iops in parallel, but all ssd's i've come across seem to such at sequential 4k read iop's, and that's more like what most linux systems are actually doing.
curiously, i also foudn when using md that there was a difference in performance between direct and buffer cache speeds, but there was a much bigger difference in cpu usage - to the tune of an order of magnitude difference in dd cpu usage with using buffered i/o versus direct, and a much smaller difference with raw speeds.
Guys, as an owner of 250GB version of Samsung 840 EVO, I did quite thorough research of my drive on the web before I bought it and the result is that is is one of the fastest SSD on the marked. This is clearly visible from numerous benchmarks all over the web, so I am afraid there is some serious problem with your tests. Are your filesystems alligned? Is 6Gb SATA really working? Do you have AHCI enabled and working? Do you have latest FW in all drives? there is no way that Samsung 840 EVO is so much slower than OCZ Vertex 3, actualy it is usually faster, sometimes significantly:
Guys, as an owner of 250GB version of Samsung 840 EVO, I did quite thorough research of my drive on the web before I bought it and the result is that is is one of the fastest SSD on the marked. This is clearly visible from numerous benchmarks all over the web, so I am afraid there is some serious problem with your tests.
I own an Evo 250GB version as well, and yes, it flys on Linux despite the inability to use the software enhancements. I have also owned both the Vertex II 60GB and Vertex III 90GB respectively (as well as a Crucial M4 128GB and a Mushkin Chronos Enhanced 60GB) and have a collection of synthetic benchmarks for all (Using the benchmark feature in Gnome Disk Utility). The Samsung rates the best overall of the bunch although I have not had it long enough to know if performance degrades over time (Both Vertex drives had to be Secure Erased every couple of months to restore full performance as they would drop by 20% with regular use). Interestingly, the benchmark differences are essentially insignificant for read speeds (varying test to test between 545MB and 559MB, excepting of course the Vertex 2 which is SATA II if I recall correctly). The write speeds vary more widely.
All that being said, having looked at the scores in this article, I come to two possible conclusions.
Either the performance difference between the 120GB and 250GB is staggeringly different* and the 120GB Evo is a complete piece of garbage or something in the Phoronix benchmark process is horribly askew.
* As mentioned earlier in this thread, the 120GB is clearly lacking compared to its larger kin but I am hard pressed to believe that it is only slightly faster then spinning platters.
Here are the results of my Toshiba Q Series Pro: http://openbenchmarking.org/result/1...hm=y&obr_sgm=y
Note that my processor is not as strong as the one Michael used, and it's not a clean install (with programs running during tests), so any processor-bound tests will show poorer performance in comparison.
But overall it seems it beats the Samsung 840 EVO 120 GB and the Kingston drive as well. For the price (same as the Samsung drive) that's really not bad at all.