Determining internal sector size for SSD for possible partitionalignment

  • Thread starter Thread starter Keiron
  • Start date Start date
K

Keiron

Hello all,

I recently purchased a SSD replacement drive (an oldish model (MCL, PATA
etc) for an ageing laptop. It's not one of the better known brands, as
far as I know at least: It's a KingSpec.

http://www.kingspec.com/

I can't determine what internal sector size it uses and so if I need to
align the the partitions for optimal speed. (If that makes sense??)(Also,
I won't be installing Vista or Win7 so I will have to do this manually).

Here is the product datasheet:

<http://www.kingspec.com/downloads/KingSpec_2.5
Inch_PATA_MLC_SSD_Datasheet_V1.1.pdf>

The specific drive model is: KSD-PA25.1-016MJ

Can anyone assist??

Additionally, can anyone shed some light on the significance of the
filesystem type? I.e. type, and node/cluster size. Presumably if the
drive doesn't use 512byte sectors it will use the 4K standard? So if I
use ext4 where you can choose the cluster size(?) I should pick whatever
the drive uses internally for max performance? Or can it potentially use
anything/something else? And lastly, any words of wisdom re: "erase block
size" and how it would effect my choices?

Hope any of this made sense to some one! Any ideas appreciated.

Thanks
 
Keiron said:
Hello all,

I recently purchased a SSD replacement drive (an oldish model (MCL, PATA
etc) for an ageing laptop. It's not one of the better known brands, as
far as I know at least: It's a KingSpec.

http://www.kingspec.com/

I can't determine what internal sector size it uses and so if I need to
align the the partitions for optimal speed. (If that makes sense??)(Also,
I won't be installing Vista or Win7 so I will have to do this manually).

Here is the product datasheet:

<http://www.kingspec.com/downloads/KingSpec_2.5
Inch_PATA_MLC_SSD_Datasheet_V1.1.pdf>

The specific drive model is: KSD-PA25.1-016MJ

Can anyone assist??

Additionally, can anyone shed some light on the significance of the
filesystem type? I.e. type, and node/cluster size. Presumably if the
drive doesn't use 512byte sectors it will use the 4K standard? So if I
use ext4 where you can choose the cluster size(?) I should pick whatever
the drive uses internally for max performance? Or can it potentially use
anything/something else? And lastly, any words of wisdom re: "erase block
size" and how it would effect my choices?

Hope any of this made sense to some one! Any ideas appreciated.

Thanks

http://www.kingspec.com/solid-state-disk-products/ssd-25pata-mlcj.htm

MLC(Test Flash:K9GAG08U0M)

That tells us the type of Samsung chips used inside the drive.
From a copy of K9GAG08U0M.pdf ...

http://www.datasheetdownload.com/download.php?id=674443

Page Program: 4KB
Block Erase: 512KB

As near as I can determine (without owning an SSD myself), the sector size
is 512 bytes. Now, because the K9GAG08U0M doesn't support partial programming,
it would likely be more efficient if the sector size is 4096. But in a quick
search on the web, I don't see any evidence SSDs use "large sectors".

For further tuning info, you might look for Linux specifics.

https://wiki.archlinux.org/index.php/SSD

If a starting sector offset of 2048 is good enough for Windows 7
and SSD, then it should be good enough for Linux. That wastes 1MB of
your 16GB drive. And if your file system has a "cluster size", it is
probably a power_of_two anyway.

HTH,
Paul
 
http://www.kingspec.com/solid-state-disk-products/ssd-25pata-mlcj.htm

MLC(Test Flash:K9GAG08U0M)

That tells us the type of Samsung chips used inside the drive.
From a copy of K9GAG08U0M.pdf ...

http://www.datasheetdownload.com/download.php?id=674443

Page Program: 4KB
Block Erase: 512KB

As near as I can determine (without owning an SSD myself), the sector
size is 512 bytes. Now, because the K9GAG08U0M doesn't support partial
programming, it would likely be more efficient if the sector size is
4096. But in a quick search on the web, I don't see any evidence SSDs
use "large sectors".

For further tuning info, you might look for Linux specifics.

https://wiki.archlinux.org/index.php/SSD

If a starting sector offset of 2048 is good enough for Windows 7 and
SSD, then it should be good enough for Linux. That wastes 1MB of your
16GB drive. And if your file system has a "cluster size", it is probably
a power_of_two anyway.

HTH,
Paul

Thanks Paul, although I'm still confused on the matter: If the page size
is 512bytes (same as the drives interface) then do I need an offset at
all? I thought it was only where the drive uses 4KB internally that the
partition offset had to be set manually? Or am I confusing several issues
here?

I'm confused on the what you've listed as the Page Program, is this not
perhaps a 4KB internal system and so require an offset?

Thanks
 
Keiron said:
Thanks Paul, although I'm still confused on the matter: If the page size
is 512bytes (same as the drives interface) then do I need an offset at
all? I thought it was only where the drive uses 4KB internally that the
partition offset had to be set manually? Or am I confusing several issues
here?

I'm confused on the what you've listed as the Page Program, is this not
perhaps a 4KB internal system and so require an offset?

Thanks

The Block Erase is 512 kilobytes, not 512 bytes. Setting the sector
offset to "2048" is 2048*512 bytes or 1,048,576 bytes. The block erase
for the flash is 512 kilobytes * 1024 bytes/kilobyte = 524288 bytes.
So a Windows 7 style offset of 2048 sectors, is two blocks of flash memory.

"Pages" inside the flash, divide evenly into the block erase size. And
for that flash chip, a page is 4KB. That is room to hold eight sectors
of 512 bytes each. As far as I know, because that is MLC flash, the
4KB quantity, is the smallest quantity that can be updated on the flash
chip.

For a figure describing how it works, see Figure 3 on page 6 here. This figure
is for SLC flash with partial page write capability. The MLC flash chips on your
drive have a larger block size (512KB versus 128KB) and larger page size
(4KB versus 2KB).

http://download.micron.com/pdf/technotes/nand/tn2919.pdf

Paul
 
Back
Top