Anyone running a Samsung 840 EVO SSD should read this (new firmware released)

I just stumbled across a discussion that's pretty important, affecting Samsung 840 EVO SSDs.

Apparently, the longer a file remains on the SSD, the slower it is to be read. And by slower, I mean MUCH slower. Discussion on Overclock.net:
LINK

The OCN thread is long, but a particular post in it seems to be probably the most valid theory/summary of the problem:

quote:

---

@fragamennon
Thanks for the infos! It looks more and more that the EVO has a bug.

In the thread with the same topic on a German website the most likely cause of this issue discussed, is a bug in the 'wear leveling / garbage collection' mechanism of the EVO.
https://www.computerbase.de/forum/showthread.php?t=1381604&page=5

EVO bug theory in short:
- Flash cells can only sustain a limited number of write cycles (writing requires erasing the flash cell)
- If all static data/files would stay in the flash cells to which they have been written to originally, the remaining flash cells would have to endure all the write cycles for changing data and would quickly wear out
- That's why all SSD reorganize old data and writes them into different flash cells, in order to spread the necessary erase cycles as evenly as possible over all flash cells. This is called wear leveling

And if the EVO has a bug in the wear leveling which causes the old data to be rearranged in a disadvantageous configuration for reading (e.g. they are all put into the same flash block so that there is no parallel block reading possible), that would explain the results we see here.

Here is an article which explains the GC/wear leveling:
https://www.thessdreview.com/daily-news/latest-buzz/garbage-collection-and-trim-in-ssds-explained-an-ssd-primer/

---

I tested this myself a few minutes ago.

I tested a 2GB sound file in my Watch Dogs game installation folder using Parkdale. I bought my 500GB Evo on August 10, and the file I tested was added to the drive on August 12, so just over 5 weeks ago.

Read: 60MB/s -

I then made a copy of it within the same folder and tested again:

This makes me sad. And Samsung is going to have some furious 840 EVO owners. These have been top performers in basically every test, but unfortunately, reviews are done under file scenarios and not over time.

However, after running MyDefrag v4.3.1 "Flash Memory Disks" option, the original file began to read normally. Defrag processes do a good bit of rewriting to move files around, which would explain the improvement. I'm going to monitor the performance degradation of various static files over the next few weeks, so I have a rough idea of how often I should run this defrag. Since program files and Windows/boot files are seldom/never changed or rewritten, so this can and will affect boot-up/load performance over time unless you use a program to rewrite the files.

The defrag, especially if only done once every few weeks or so, should not significantly reduce the life of your SSD to any span that a normal human tends to keep drives. However, it is quite irritating that this has to be done at all. If it really is just a wear-leveling bug in the controller, Samsung needs to release a firmware update ASAP or they've lost my fanboyism permanently. This does not affect the 840 Pro, however. It's possibly not an issue on any MLC NAND drive.

Until then, ignore any previous instance in any thread where I may have recommended a Samsung 840 EVO.