Samsung SSD 840 susceptible to flash read speed degradation

[leafer]

Might be old news to some, but this problem affects the 840 EVOs and regular 840s. They're also butt-hurt expensive in this country as well and I was saving up for a Samsung until I came upon a few articles.

 

This one has a video discussion of the problem.

 

http://www.pcper.com/reviews/Editorial/Samsung-840-840-EVO-susceptible-flash-read-speed-degradation-over-time

 

Anyways, it appears Sandisk and Intel make some of the most reliable SSDs out there.

Edited March 16, 2015 by leafer

[BitMaster]

Don't mix up 2 different models !

 

The 840 EVO does have problems that even the latest firmware patch afaik could not fully fix.

Also, those 840's DONT have 10 year warranty.

 

The 850 Pro does NOT suffer that problem and has a 10 year warranty.

 

 

So if you buy a Samsung now, get a 850 Pro.

 

Put 2 of them in Raid-1 if you need reliability. A single drive is always walking on the edge, no matter how reliable the series is, it fails or doesn't.

 

 

Bit

Edited March 16, 2015 by BitMaster

[leafer]

Ahh. I listened to the vid again and indeed he said the pro SSDs don't suffer this problem. Thanks for pointing that out, Bit.

[T_A]

If you are really interested in endurance of SSD drive , you really need to read this article about SSD drive comparison for maximum writes before errors and before complete drive failure.

 

The SSD Endurance Experiment: They're all dead

 

of course all drives will die in the end , but some (like Samsung) showed amazing numbers that would take real world users many many years to reach (by that time you will upgrade several SSD`s atleast)

 

the 840 Pro reached 2.4 petabytes (!!!!) before dieing.

Edited March 16, 2015 by T_A

[leafer]

Here's a test I saw yesterday.

 

http://ssdendurancetest.com/

 

Whoa, that 840 pro costs 94$ more than same capacity sandisk.

Edited March 16, 2015 by leafer

[T_A]

That test you put up has 840 EVO 120GB , not PRO (not sure if you meant it showed a PRO)

 

anyway my 250 EVO did in around 10 months 2.2TB and its my OS and Games drive so reaching even that 80 TB for 120GB in the test seems unlikely (and 250 will last atleast twice)

[BitMaster]

With different words, most SSD's sold today live longer than any HDD ever would if it had to write such immense data over years.

 

Mechanical wear and tear on such drives is enough to make it fail in a few years heavy usage.

 

From 16 250GB Samsung HHD drives I bought for a Raid System like ⅓ of them failed in less than 3 years, luckily they were all Raid-6 !

 

From 6 x 2TB drives ( enterprise storage ) 3 of them failed during warranty ( Seagate and WD ).

Also all Raid-6 so we didn't loose any data.

 

 

HDD's are anything but reliable if seen as a single device.

 

The best insurance against data-loss is a Raid system with multiple redundancy ( Raid-6 ) a Battery Backup Unit and a decent Raid Controller like Adaptec or LSI etc..

 

For my gaming rig I may opt for a Raid-0, forget about redundancy, make daily backups and accept risk while gaining performance. It's a either or.

 

Bit

[T_A]

BitMaster pretty much summed it up, with all the bad rep SSD reliability got , people are clueless that HDD are not even half as reliable even under normal use

(i wont even go into External USB drives and Laptops uses)

[BitMaster]

It's sad bad true.

 

Over the past 5 years I have seen more failed drives than ever before, in percentage.

 

I love Raid-6 and Adaptec, believe me, that never ever let me down when things got tough.

 

Bit

[SkateZilla]

The 850s are fabulous,

[leafer]

All right I think I over reacted after glancing through these reviews. These tests made it seem as if the lesser brands would die prematurely when in fact, an average user would never move that crazy amount of data on his SSD. So, I should just go for the cheapest one regardless of brand.

[Sgt_Baker]

x2 To what BitMaster says about standard HDDs. When you're in the business of purchasing them in bulk, you happen across a surprising number of failures.

 

Incidentally, Google published some interesting material in this field a while back, which was based on their experience with HDDs in the gazillon servers they run.

[BitMaster]

I feel the need to add some wisdom, gained through pain and agony:

 

DO NOT USE the embedded Intel Raid functions if you really need that data on the drive(s).

Many many ways of rescuing your data are voided if you bet on a software horse rather than a hardware driven horse.

 

NO linux will recognise the logical drive = no ubuntu way of saving your data quickly !!!

MOST rescue cd's will also NOT see your logical drive but 2 independent drives !!!

Make sure that your high end version of Acronis, Norton etc.. can see your logical drive before you go wild ! Many of them have real issues with sw driven raids.

 

Reconstructing a Raid-5 consisting of 3 x 500GB took over 3 days !!!!!!!!!!!!!!!!!

 

We ( or the customer that would not listen ) had to wait THREE days until he was able to use that PC again. During reconstruction the PC was so slow that you could not use it. Even better, after 3 days it tilted and it all was in vain.

 

Whereas, on a decent raid controller, you would not even notice that it is in degraded state ( 1 or 2 drives failed ) and you would also not recognise that it rebuilds in the background. Just keep on working and in 30-180min usually your logical drive is redundant again. Smooth...

 

The really ONLY thing those Intels are good for is a merciless RIAD-0 for DCS. Forget about recovery, do a plain backup to have access to cold files AFTER you manually installed the OS again. Accept the risk and gain some performance. Anything else is dangerous and a marketing lie.

 

Bit