Enterprise SSDs can lose data when the power is switched off for a week

Hrm, I wonder how true that is. I have heard they are not even SAS and just consumer level componentry with 30% OP and 800% markup.

edit- actually poking around, there are indeed SAS SSDs (which are beyond ridiculously priced) and then there are SSDs advertised as "enterprise." The SSD pictured in the article appears to be a SATA. Which type of "enterprise" drives are they likely discussing, anyone know? The SATA "enterprise" drives, SAS drives, or consumer drives?

HjvUA9T.jpg
 
Last edited:
http://www.ibtimes.co.uk/ssds-lose-data-if-left-without-power-just-7-days-1500402


Fortunately, consumer ones will last 2 years without power before losing data.

grain-of-salt.jpg


I am going to dispute that based upon my own experience with my adata SSD

I have a 128GB SSD (an A-Data SX300) that I put into storage in October 2012 with a full windows installation (7) and I powered it on for the first time in MAY 2015... not one bit was corrupted.

I wonder if these are the same people that said CDs, DVDs, and BLU-Rays suffer from bit rot.... lol I got CD's from the 80's that still work...

and just to show that I am not kidding look at the power on hours of this drive and the power cycle count

A-data%20SX300.png
 
Last edited:
I think the thesis of the article might be off base come to think of it. So they are saying Seagate has specified for every 5 degrees Celsius rise, there is a 50% reduction in power-off retention which is interesting and good to know of. What they are not realizing is that most electronics manufacturers base their standard temperature rating at 25ºC. Logically, we would need another 24ºC added to reduce life in theory to 7 days (down from 90 days specified by Seagate supposedly). This may affect 'cold' storage in the middle east but who would leave an SSD in storage at 120ºF?
 
buy a Samsung evo it won't matter it will fade within 6 month hurray for TLC nand ha ha ha
 
Hrm, I wonder how true that is. I have heard they are not even SAS and just consumer level componentry with 30% OP and 800% markup.

edit- actually poking around, there are indeed SAS SSDs (which are beyond ridiculously priced) and then there are SSDs advertised as "enterprise." The SSD pictured in the article appears to be a SATA. Which type of "enterprise" drives are they likely discussing, anyone know? The SATA "enterprise" drives, SAS drives, or consumer drives?

Lets clear the air with some facts...

There are two different JEDEC specifications for calculation of the endurance of an SSD: one for enterprise use and one for non-enterprise use (aka consumer).

As part of these specifications, EOL data retention requirements are listed. These are the data retention requirements once all the endurance of the drive is used (aka TBW/PBW).

These requirements are listed in a grid matrix with power on temperature at write and power off storage temperature for data retention. The actual values in the grid are related to testing/research that was conducted by Intel on the effects of temperature on data retention. AKA, based on the Intel research, the grid matrix values should all be effectively the same once accounting for temperatures.

And these aren't specifications for realities for any given drive, they are the minimum requirements. And once again, they are all also EOL values. The retention rates of the cells decrease over time as the P/E cycles are used and these are the end state values.

Getting back to the edit question you posed, generally unless stated otherwise, one would assume that the drive is rated using the consumer level specification. Certainly all SAS based SSDs are likely using the enterprise spec, and many of the explicitly marketed enterprise grade SATA drives are. Anything using HEMLC (aka High Endurance MLC) is being spec'd using the enterprise specification. This is because HEMLC really isn't a thing, its the same flash as MLC just rated in a different way, trading P/E cycles for retention times...

And in general the SAS SSDs are very expensive, which is because they generally offer steady state performance well beyond anything a consumer drive offers. For instance, when HGST introduced their SAS line in cooperation with Intel, they base OP was 2-3x. In addition, the SAS drives are marketed for full continuous workloads for multiple years and support extremely high endurance at worse case scenarios (upwards of 25 Drive Writes Per Day for 5 years of 4k random and sub 4k random full span writes). The Write I/O numbers for the SAS drives are 4k random full span, in contrast the Write I/O numbers for consumer level drive are 4k random for 8GB span.
 
Lets throw out all digital and go with papyrus scrolls and stone tablets.Digital media has proven itself unreliable... Best call Fred and Barney ha ha
 
I complained about this years ago here and people thought I was crazy. As the NAND process gets smaller, retention becomes more volatile. I wouldn't touch 16nm TLC NAND with a 10' pole. Some things are worth more than a 10% speed increase.
 
As the NAND process gets smaller, retention becomes more volatile. I wouldn't touch 16nm TLC NAND with a 10' pole.

This is one reason why all NAND manufacturers are moving to 3D NAND by 2016. Samsung has moved back to 40nm with their G2 3D NAND.
 
Back
Top