View Single Post
  #12  
Old March 24th 21, 02:31 PM posted to alt.comp.os.windows-10,comp.sys.ibm.pc.hardware.storage
Paul[_28_]
external usenet poster
 
Posts: 1,467
Default Why is it not letting me extend the partition?

Yousuf Khan wrote:


On an alternate note, the old drive now has one tiny little bad sector
hole in it, that I'm thinking the drive can deprovision, and carry on
without in the future. Is there something that can allow the drive
electronics to carry on an internal test and remove the bad sectors?

Yousuf Khan


Testing burns wear life.

*******

A sector has three states (for this discussion):

1) Error free (in TLC/QLC era, highly unlikely)

2) Errors present, ECC can correct.

3) Errors present, ECC cannot correct. tiny little bad sector.

If (3) were marked with "write, but do immediate read verify",
this would allow evaluating the material in question, after
it was put in the free pool. The "questionable status" should
follow the block around, until it can be ascertained that it
is (1) or (2) again. If it showed up (3) on a retry, it should
be thrown into the old sock drawer. Any "write attempt", is
an excellent time to be checking credentials of the block.

The procedure should be similar to hard drives, economical
in nature, yet not endangering user data. To do walking-ones
or a GALPAT on the flash block, that would be seriously naughty
and pointless. You could burn out the entire block wear life, then
conclude there is nothing wrong with the block :-)

Seagate has a field on their hard drives, called "CurrentPending".
For the longest while, I took that at face value. However,
that field isn't what it appears. It only seems to increment
when the drive is in serious trouble and has run out of spares
at some level. It's unclear whether there is an "honest"
item in the SMART table, keeping track of items like (3) so
a customer can judge how bad things are.

SMART is generally not completely honest anyway. There's some info,
but they are dishonest so that users do not "cherry pick" drives,
and send back the ones that have a tiny blemish when purchased.

On hard drives, at one time it was considered to be OK for a
drive to leave the factory, with 100,000 errored sectors on it.
That's because the yields were bad, and the science could not
keep up. Now, if SMART was completely honest about your drive,
imagine how you'd freak out if you saw "100,000" in some table.
This is why the scheme is intentionally biased so drive devices
look "perfect" when they leave the factory, when we know there
is metadata inside indicating the drive is not perfect. Especially
with TLC or QLC. SSD drives do not leave the factory with
a state of (1) over 100% of the surface. There is lots of (2),
and more (2) the longer the new drive sits on the shelf. That's
why, if you want to bench a modern SSD, you should write it from
end to end first. This removes the degree of errored-ness on
the surface, before you do your read benchmark test. If the drive
was SLC or MLC, I would not be doing this... It would not need it.

The Corsair Neutron I bought, on first test, I was getting 125 to 130MB/sec
on reads. Dreadful. The performance popped up, after a refresh. I still took
it back to the store for a refund the next morning, because
(maybe) the manufacturer would like some feedback on what
I think of them.

Paul