@EsaT - thanks, that may actually be where my preference for it originated, I couldn't think of any logical reason for it.
I'm pretty bad at letting my drives get too full, probably something I should take more care with. Is space equally an issue for spinny drives or is it fine to ram them to the gills?
Yeah, spinning rust is precisely logical.
First sector address corresponds to first physical location of data in start of physical disk at outermost track.
With same location tracks of different platters making that cylinder.
And because of longer length of that track there than on inner last cylinders, bigger amount of data is read in single rotation.
That also makes same amount of data fit into fewer cylinders requiring less moving of R/W heads compared to end of drive/inner cylinders.
In fact transfer rate from outer cylinders are even up to twice that of the innermost cylinders:
https://www.servethehome.com/hgst-w...iew/hgst-ultrastar-dc-hc510-10tb-hdtune-read/
Hence why you wanted to keep often needed/more performance important data in start of the drive.
And because of mechanical movements involved HDD suffers greatly, if data/file isn't in continuous chunks.
That's why it wasn't good idea to fill HDD.
Having good amount of free room means, that file system can write new data/files in continuous pieces avoiding fragmentation.
Though old FAT didn't care about that and started writing file from the first free sport regardless was there enough space for new file.
For example if you were writing 10 MB file and there were empty spots left by removed 1 MB files, that would have resulted file being split into ten fragment.
That caused need for regular defragmenting.
NTFS is better in that by first looking for single continuous empty spot capable to fitting new file completely.
Though with fuller drive fragmentation started again happening.
Again in SSDs file system sector addresses don't correspond to any single precise location/sequence of consecutive cells in NAND chips.
Controller even keeps juggling data on same sector address of file system all around NAND chips for wear leveling:
Unlike HDD, Flash memory has very limited number of writes it can take.
So if you have static data containing files and other files with frequently changing data/temporary files written and removed that would wear down part of the drive faster.
(while other parts would stay as almost new)
To avoid that controller eventually starts moving static data containing files to different location on NAND chips to use write endurance of the cells evenly.
So it really doesn't matter in which order partitions are.
Every spot has same performance capability and what partition table and file system say as location of data doesn't mean anything for actual location in NAND chips.
But that doesn't mean free space isn't needed to keep optimal SSD operation.
- Wear leveling works best when it has good amount of free space to rearrange data.
- Also reads and writes of NAND flash work differently to HDDs:
Writes happen in bigger chunks than reads.
And unlike in HDD you just can't overwrite existing data meaning NAND cells have to be erased to actually release NAND cells used by old deteled files.
So changing/writing new small file can need first reading out other data around that location and erasing all those cells before writing data back with changed/new file.
- And all modern drives store more than single bit to cell using more than empty and full charge/voltage states.
Downside of that is write speed being slower than writing single bit per cell with either empty or full charge, which is masked by using some amount of free space as single bit cache.
That means in TLC drive 100 MB size file initially takes 300 MB worth of space.
Then when drive idles controller writes that file into TLC form using slower to write 8 charge/voltage levels per cell.
So actual amount of writes needs 400 MB worth of room.
Because of that SSDs always have some unavailable/not shown to user space as working space.
But they can really use more of that:
If drive hasn't had time to erase cells of deleted files and consolidate data with chunks of empty cells being the size of minimum write, amount of work and writes increases fast if new data is written. (leading to higher than necessary wear)
.
