What's new

Improving SSD performance

Most flash have a 10-year data retention spec - which means that when the flash is new data that is written will not degrade below the raw bit error spec for 10 years when stored unpowered at 30 or 40 deg C. This time decreases at higher temperature (because the leakage currents are higher) and as the flash gets closer to its endurance limit (because the signal to noise ratio of the cell is degraded due to trapped charge in the oxide - so it takes less leakage to cause an error). There are also read and program disturb effects that come into play.

Some flash does require that data be moved/refreshed after about 100K reads, due to read disturb effects. Many SSDs also do background scanning of data to move/refresh data that is close to the correction capability of its ECC.
 

leeshor

Well-Known Member
Performance issues, degradation, are not what I'm talking about. Some people feel that if the drive isn't refreshed that data loss will happen. Hogwash.
 

Nicola

Member
Performance issues, degradation, are not what I'm talking about. Some people feel that if the drive isn't refreshed that data loss will happen. Hogwash.
If you don't tackle the degradation of the signal, what else could happen apart from data corruption?
 

GreyFox7

Super Moderator
Staff member
Most flash have a 10-year data retention spec - which means that when the flash is new data that is written will not degrade below the raw bit error spec for 10 years when stored unpowered at 30 or 40 deg C. This time decreases at higher temperature (because the leakage currents are higher) and as the flash gets closer to its endurance limit (because the signal to noise ratio of the cell is degraded due to trapped charge in the oxide - so it takes less leakage to cause an error). There are also read and program disturb effects that come into play.

Some flash does require that data be moved/refreshed after about 100K reads, due to read disturb effects. Many SSDs also do background scanning of data to move/refresh data that is close to the correction capability of its ECC.
In the olden days core memory had to be refreshed after every read :) but we didn't require the system operators to be involved. :D
These details should be handled by the device firmware. it's interesting but not something users should be mucking around with in general.
 

leeshor

Well-Known Member
If you don't tackle the degradation of the signal, what else could happen apart from data corruption?
This discussion is about a specific bug that causes read degradation. My comments were targeted at the notion that if the drive, any drive, is not refreshed now and then that data loss will occur. This notion harkens back to the early days of purely magnetic "hard" drives but carried over to what they once called bubble memory, now SSD.

I never said this issue isn't a problem. I'm ONLY addressing the comments regarding data loss from not refreshing or lack of use of all the data on the drive in general.
 

Nicola

Member
This discussion is about a specific bug that causes read degradation. My comments were targeted at the notion that if the drive, any drive, is not refreshed now and then that data loss will occur. This notion harkens back to the early days of purely magnetic "hard" drives but carried over to what they once called bubble memory, now SSD.

I never said this issue isn't a problem. I'm ONLY addressing the comments regarding data loss from not refreshing or lack of use of all the data on the drive in general.
Unfortunately this drive, if not refreshed now and then, loses (or MIGHT lose) data. Mostly every SSD has that but it is worse for the 840EVO being it a TLC, which is more sensitive to this. A good firmware should refresh the data once it notices corruption.
What you're missing, still, is the reason behind the read degradation.
If you check, the problem is that the data is already corrupted and what kicks in is the ECC. (I believe you know the purpose of ECC, especially what the E stands for)
Since the ECC processor has not been sized to handle the amount of corrupted data, a slowdown is noticed.
Funny thing is, the ECC processor uses a lot of power and generates some heat. Few months ago I even started googling for "840 EVO heat" because mine was becoming quite hot (could feel it through the case of my laptop). And, of course, I had no idea of this whole thing.

You can find a hint even on Samsung site: http://www.samsung.com/global/business/semiconductor/minisite/SSD/global/html/why/MlcNandFlash.html

and a nice experiment here: http://www.pcper.com/reviews/Editor...-read-speed-degradation-over-time/Update-Temp
 
Last edited:
@leeshor: Agreed. Drive firmware should take care of this and the user should never have to "refresh" the data of any type of storage device (with maybe the exception of 100 year tape archives).
 

razy60

Member
I read in reviews (before buying) the SP3 had a Samsung SSD on board. I actually have a Hynix SSD in my I5/256/8Gb SP3. Upgraded to the latest firmware version (thanks to a Dell tool, Dell uses the same SSD in some laptops).
which tool did you use and link if possible please.
 

ynohtna

Member
I guess i have this problem! Been feeling everything is sluggish these days. Even typing this message is ever so lightly delayed. How long does the disk fresh process take? Just trying to plan the fix.


sp3hdtunetest.PNG
 
Top