Flasherly wrote:
On Sun, 02 Aug 2020 03:23:36 -0400, Flasherly
wrote:


No need to go farther -- and the model is QVO (not a misstated "QVC").

Looking a little closer, apart from the greater most, a substantiality
with little apparently to negatively say, are those whom regardless
prefer to deal in the reality of a notable deficiencies given
Samsung's QVO cache implementation design -- in that they concur in
the instantaneousness of the moment I powered this drive up, only to
notice significant speed discrepancies. An aspect which would get
longer, as there are both provisional Samsung software supplements, as
well an actual explanation to limits of any greater expectancy for
perceived performance. None of which affect to mollify what I
suspected -- that the Samsung QVC is a bad, bad doggie -- past, I
should dare say, ostensibly neither to corrupt data or provide
sustained long-term usability upon a premise of care exercised within
reason not to exceed rated limits and specifications.

Want more than to read, then why not just shake it on over, shell up,
and instead buy the EVO.

Initialize the entire drive, then do a read benchmark again.

In Linux, this would be:

sudo dd if=/dev/zero of=/dev/sda bs=65636

Many other cloning methods, don't cause every block
to be written, while "dd" can be made to write every block.

That would write all the sectors up to the end.
And helps reduce the number of errors in each
block that need correction.

And why would you buy a QLC drive in the first place ?

SLC
MLC === includes "MLC-like" ones
TLC ?
QLC ???

I had a TLC drive already, with "un-smooth" performance, and
I took it back to the store.

The SLC or MLC don't use fake "Flash Cache" during writes
and don't have two write phases. Those drives just do
plain writes, at a constant rate.

The higher density flash chips, cannot sustain performance,
so they have to "cheat". Yes, you didn't pay a lot for
the QLC drive, but you also should not be benchmarking
the drive, knowing how poorly the results will turn out.
If you want to bench it, you'll need to write it from
end to end, every sector, first. Then the performance
will match (some) of what it says on the tin.

Paul