Samsung 850 EVO SSD Linux Benchmarks

[Licaon]

in raid the 840 evos kind of deal as long as you secure erase every now and then.

Except they don't support 840EVOs in RAID only Pros, whatever that means, so I ignored them. I got a pair of Kingston HyperX 3K 120Gb instead (yeah yeah I know about their V300 issue) and they work great in RAID 0. I also got some 3Ks for different systems, non-RAID, either those or Intel 530s, whatever was cheeaper at the time, all working great so far.

[chrisb]

compression. it's probably zero-data.

Yup...

533 	/*
534 	* This routine opens, writes the amount of (zero filled) data to a file.
535 	* It chunks IO requests into the specified buffer size. The data is just zeroed,
536 	* nothing in the kernel inspects the contents of the buffer on its way to disk.
537 	*/

It would be trivial to replace the

memset(io_buffer, 0, io_buffer_size);

with some non-compressible data

[mercutio]

I've heard about possible performance bugs in the firmware, but not anything about data loss, full or partial. I generally consider Samsung to be among the few reliable SSD drives (along with Intel and Crucial) and all my SSDs are Samsung (840 Pros for my primary desktop and laptop, 840 Evos in secondary roles, because I do trust the Pro models more), and have never once had a problem. I've got one old Crucial SSD, the first I ever owned, and that's been reliable too. Oh, and a crucial mSata drive that's been good too.

I'd believe data loss during power-cycling, which would better be avoided but is to be expected since only Intel stands apart in this regard, but if you have sources showing data loss from a Samsung Drive I'd be interested to see it, and whether reports stem from Pro or Evo drives. Googling Samsung SSD data loss didn't turn up any hits on the first couple pages.

the 840 evo performance drop is pretty chronic. not as bad as the 840 non-evo, which is going at 10mb/sec now

i've shifted to 850 pro for desktop, as the 840 evo was just getting too bad. in raid the 840 evos kind of deal as long as you secure erase every now and then.

[mercutio]

Any speculation as to why the OCZ Vertex3 performs so well in the FS-Mark tests? Its more than 2x better than the competitors in the graphs, but it's really a mediocre mid-range drive.

compression. it's probably zero-data.

[mercutio]

the 250gb should benchmark quite a bit higher for write-focused tests.

in normal usage the write speed should be fine for the 120gb model though.

ime the 840evo 120gb's are pretty bad. but even when they're bad they're still much more usable than a hard-disk. i had to secure erase one to get performance back even with the firmware restoration fix and freshly imaged.

secure erasing is a pita, i had to hotplug the drive to do it. but it fixed the performance.

where linux really seemed to fall down is it just doesn't seem to throw enough simultaneous requests at disks at once, so iops performance goes down.

[flubba86]

Any speculation as to why the OCZ Vertex3 performs so well in the FS-Mark tests? Its more than 2x better than the competitors in the graphs, but it's really a mediocre mid-range drive.

[ravyne]

I can only link a german source: http://www.heise.de/newsticker/meldu...e-2560890.html
tl;dr: firmware update for 850 PRO could turn the SSD into a brick.

I do have samsung SSDs in my computers, too (bought three when they were cheap), but all things considered, Samsung sucks at software. Always have, always will. I'm not really surprised by the firmware issues.

Thanks for the link, I appreciate it.

I guess such problems don't really concern me too much. This sounds like a faulty firmware that just bricked the drive, which is far less evil than a firmware that appears to be working but is silently losing or corrupting files. If I'm flashing firmware on an important drive, I'll make sure to have a back up. Having to RMA the drive would be a major pain, I admit, but is much less catastrophic than actually losing data.

But I suppose I'm fairly cautious against drive failures of any kind, anyways, I mostly keep my data on a mirrored volume; just app installs on the SSD, data files that need the performance are on the SSD but backed up regularly to the mirrored volume, (and replicated there more frequently for changes in-between).

[davidbepo]

everything went fine since buying this ssd, its so fast and power-saving
thanks for benchmarking it

[rohcQaH]

I've heard about possible performance bugs in the firmware, but not anything about data loss, full or partial.

I can only link a german source: http://www.heise.de/newsticker/meldu...e-2560890.html
tl;dr: firmware update for 850 PRO could turn the SSD into a brick.

I do have samsung SSDs in my computers, too (bought three when they were cheap), but all things considered, Samsung sucks at software. Always have, always will. I'm not really surprised by the firmware issues.

[ravyne]

Samsung's always been reliable for me.

Considering that Samsung has been in the news a few times during the past weeks about trouble with their firmware (including the possibility of total data loss), I won't be basing my purchasing decisions on benchmarks alone.

I've heard about possible performance bugs in the firmware, but not anything about data loss, full or partial. I generally consider Samsung to be among the few reliable SSD drives (along with Intel and Crucial) and all my SSDs are Samsung (840 Pros for my primary desktop and laptop, 840 Evos in secondary roles, because I do trust the Pro models more), and have never once had a problem. I've got one old Crucial SSD, the first I ever owned, and that's been reliable too. Oh, and a crucial mSata drive that's been good too.

I'd believe data loss during power-cycling, which would better be avoided but is to be expected since only Intel stands apart in this regard, but if you have sources showing data loss from a Samsung Drive I'd be interested to see it, and whether reports stem from Pro or Evo drives. Googling Samsung SSD data loss didn't turn up any hits on the first couple pages.