View Single Post
  #9  
Old June 12th 18, 08:26 PM posted to alt.comp.hardware.pc-homebuilt
Paul[_28_]
external usenet poster
 
Posts: 1,467
Default Unreally Lucky HDD

Charlie Hoffpauir wrote:
On Tue, 12 Jun 2018 13:41:21 -0400, Larc
wrote:

On Tue, 12 Jun 2018 03:23:01 -0400, Paul wrote:

| There are some shingled Seagates for sale now,
| and you definitely don't want to buy those.
|
| The trick is to find a datasheet with a platter
| count. If the platter count is too low, that's
| a shingled drive. I think you might be able to
| get 2TB on one platter that way. One of the
| other hints, is the shingled drives ship in a
| 0.8" high housing, rather than a regular 1.0"
| high housing.
|
| I don't know if WDC has done this yet or not.

I assume not since all the WD 3.5" internal HDDs are 1.028" high.

Larc

Thanks for this thread! I had not heard of "shingled" drives, ao I
googled it and got the explanation. Seems my 4.0 TB Seagate that I use
for backing up all my data is shingled (based on the fact that it
appears to be about 2/3 the thickness of a normal HDD). So far I
haven't had any problem with it. But if it suddenly craters, I'll have
lost about half my backups.(The other half is split between a 3 TB and
a 1.5 TB drive).


The only reason I "have a hate-on" for shingled drives,
is they're *not* labeled as such. These things
should be clearly labeled. Everyone wins, when
seller and buyer are honest with one another.

I don't like having to use supposition and reading
customers reviews to determine "kinda/maybe" an
item is shingled.

The original shingled drives would read at 200MB/sec
and write at 25MB/sec. Write performance was "uneven".

Shingled drives write 7 tracks in sequence. That's
seven rotations of the spindle, rather than a head
assembly with "7 coils on it".

If you need to change a 4KB sector on a shingled
drive, it's a read-modify-write. You read seven tracks,
changes a paltry 4KB section, and write the seven tracks
back. There is a slightly larger gap between one seven-track
cluster, and the next.

The current generation of shingled drives read at
200MB/sec and write at 200MB/sec. But, it's not clear
what write patterns are going to "bust" the clever
caching methods used to make this possible.

For my own disk usage, I want predictable and
consistent performance. Even in my backup drives.
This is why I cannot offer an endorsement for
shingled drives, as long as they're doing
read-modify-write of seven tracks at a time.
While their HDTune sustained might look fine,
I'm sure there's some write pattern that
will make drives like that "take a flake vacation".

Paul