I have a Samsung SSD 256GB (PB 22-j) and have noticed that write speeds have slowed. It's not a big deal as I store most files on HHD's and SSD is still much faster as an OS drive.
I recently read about TRIM being included in Windows 7 but it's so hard to find out information on it.
Does it come as standard in Windows 7?
Is it something you need to run or is it an unseen app always running in the background?
Does it work for all SSD?
If not, do it work on the PB22-J?
I've even been in contact with Samsung and got no reply. I'm sure someone around here can help.
It only works on some SSDs, like Intel X25-M G2 (and only the G2) and OCZ Vertex. In virtually all cases you need to update the firmware inside the SSD.
I'm pretty sure your Samsung SSD does not support it, Samsung controllers are pretty weak although they offer alot of MB/s they do not perform very well in IOps.
Can you download HDTune Pro and post a screenshot of a "Random Access" benchmark?
------------------------------...man will occasionally stumble over the truth, but usually manages to pick himself up, walk over or around it, and carry on.
Reply to sub mesa
That's not the random access benchmark; did you download the Pro version? It should have a Random Access tab. Sequential speeds are not relevant to system disks; only to data disks for large file storage.
I don't think the current Samsung/JMicron controllers will ever support TRIM, but.. who knows?
------------------------------...man will occasionally stumble over the truth, but usually manages to pick himself up, walk over or around it, and carry on.
Reply to sub mesa
Well this is real performance; not the MB/s which you see everywhere - system disks have little use for high sequential speeds.
IOps is real performance, and random IOps is a reasonable way to simulate more complex I/O patterns used by actual applications like booting, launching applications, etc.
Though SSDs are very good at reading randomly, they have more trouble with writing random patterns. I do think the random write test is 'destructive' though, meaning that it would require to be free of any filesystem. Though i'm not sure how HDTune tests this; if anyone does not have crucial data on the SSD, you can try the random write test. This test should really tell if you got a good SSD or bad SSD.
However, as i've seen some benchmarks of the SSDs and analysed them, i can tell the SSDs with JMicron and Samsung controller are much worse in random write than the Indilinx and especially the Intel controller - which are very good SSD controllers even for random writes.
As a system disk however, random reads are most important. Though sometimes writes happen too and the 'stuttering problem' with early (JMicron-based) SSDs such as the OCZ Core series were because of extremely high latencies for random writes. So a benchmark like this (but then write instead of read) would uncover this.
Only Intel dares to spec the random read/write IOps scores for their SSDs. SSDs with Samsung and JMicron controllers only specify MB/s throughput scores. But throughput really is not that important for an SSD; its important for HDDs which store large files (movies/music/archives) which are read or written in a sequential matter (1-2-3-4-5...) instead of a random pattern.
------------------------------...man will occasionally stumble over the truth, but usually manages to pick himself up, walk over or around it, and carry on.
Reply to sub mesa
Yes well as i said the read IOps are most important i'd say. And ~7000 IOps is alot more than even a barrage of RAID0 15k SAS/SCSI disks can dish out; so its not that your SSD is total crap.
But of course, if people would buy an SSD now, the Intel ones are highly recommended. They have both high random read AND write IOps, whereas your SSD probably only has high random read IOps. It shouldn't be a major issue as that's what it'll be doing most of the time.
------------------------------...man will occasionally stumble over the truth, but usually manages to pick himself up, walk over or around it, and carry on.
Reply to sub mesa