Storage Solutions Samsung Acknowledges the SSD 840 EVO Read Performance Bug

[spr]

Hi,

I stumbled upon this as I wanted to buy EVO

This for potential EVO buyers and user, please take note of this before considering purchase of these drives -> http://www.anandtech.com/show/8550/...40-evo-read-performance-bug-fix-is-on-the-way.

Thanks.

 

[cranky]

Note that this applies to the date when data was written to the drive, not the original date of the data.

If you clone a hard drive, for instance, you still get full read speed even on files over few months old.

It's also not an issue for OS drives because most of the actual file data is usually refreshed almost every boot, but for data storage it may be a problem. Larger drives with older files are more susceptible (in real-life usage scenario).

In practice its not a problem.

 

[Crazy_Eddy]

The odd part is that the bug only seems to affect LBAs that have old data (>1 month) associated with them because freshly written data will read at full speed

Looks like a data retention issue?

 

[cranky]

Pretty sure it's a bug related to TRIM or similar drive refresh algorithm. It seems to be not relocating data correctly, hence the problem.

Note the comments where it is noted certain drive defrag procedures restore full read performance.

 

[Crazy_Eddy]

Someone reported that he fired up an old system that hadn't been used in months and faced this issue. And there are comments about this affecting the old 840 vanilla drive too. Kristian Vatto also made a comment seeming to imply it is a data retention issue :

My theory is that the NAND management algorithms don't take the changes in cell charges into account properly. Electrons leak through the silicon oxide over time, which may result in a change in the voltage state of the cell. That in turn would return an incorrect bit output when read, so the data must be recovered using ECC/parity, which makes it readable but the calculations take time and that translates to poor performance.

link

I admit that TRIM/GC theory does occasionally fly over my head, but I was under the impression that garbage collection wouldn't do anything for static data consolidated into a fully filled block. Or does GC also shuffle fully filled blocks? Maybe the firmware 'fix' will add this option.

On a side note, this bug doesn't appear to have affected the MLC based 840 Pro, so I hope the people blindly recommending the Evo will now know why I urge them to consider an MLC drive

 

[Neotheone]

Samsung has fixed the bug with a patch released around Oct 15, as promised.

However, as someone considering buying his first SSD, I'm wondering if its worth going for a drive which may now be using a tweaked firmware to regularly rewrite the existing data to keep it "fresh", so to speak, in order to avoid performance slowdowns.

Any thoughts on this guys ?

 

[Crazy_Eddy]

Depends on how often they're refreshing the existing data. If lifespan is important to you, then pick an MLC drive. OTOH the argument can be made that you might probably shift to a newer SSD by the time it hits its P/E cycle limit.