Write bottleneck encountered in SSD benchmarks

[andy318]

I just got a Samsung PB22-J drive with the latest VBM19D1Q firmware. In different benchmark programs, sequential read speed approaches the value in the specification (220MB/sec) but write speed is much lower than the specified 200MB/sec.
Take a look at the ATTO benchmark screenshot, seems like there is some bottleneck which is slowing down the writes. Any ideas, anyone?

ATTO Benchmark

 

[andy318]

Forgot to mention that i'm using the SSD in my Dell Inspiron XPS M1330 laptop.

 

[pjkenned]

That may just be the write speed of the drive. Typically drives have lower write speeds than read speeds. 50MB/s off on one benchmark is pretty big, but not unheard of. I've seen HD Tune and ATTO differ by 200-300 MB/s (that's difference mind you) on 8 drive Raid 5 arrays.

 

[jimhsu]

That to me screams "fragmentation". I don't know how the Samsung performs when fragmented, but all SSDs without TRIM (basically, everything) can potentially perform this way.

Here's a recent ATTO bench for a heavily used, 90% full X25-M G2: http://www.diyeconomist.net/img/2009-09-29_0850.png

 

[andy318]

Not sure how fragmentation would affect write time because SSD does not really get affected by it. If you mean, that a used drive will have slower writes because an erase needs to be done first, then it shouldn't apply in this case because the drive is brand new and the firmware it comes with (VBM19D1Q) supports TRIM. I'm running Windows 7 and ran the "fsutil behavior query disabledeletenotify" command and it returned the value to indicate that TRIM is supported

 

[andy318]

Never mind. I changed my power plan from Balanced to High Performance and re-ran Crystal & ATTO and the figures are amazing now.

Crystal DiskMark

ATTO Benchmark