• Re: RPi desktop on a Acer Aspire TC-780 - any gotya's ?

    From The Natural Philosopher@3:633/10 to All on Fri Nov 20 15:45:02 2026
    On 20/11/2026 15:38, R.Wieser wrote:
    Theo,

    If you keep some amount like 10% of the SSD zeroed then the
    controller can use those blocks to improve the wear lifetime
    ('over provisioning').

    Do you mean unpartitioned, or would something special (zeroed?) be needed ?

    And thanks for the suggestion.

    Regards,
    Rudy Wieser


    I've been using swap on SSD on this machine for several years. No issues
    at all

    I think one key thing is use a much bugger drive than you (think you)
    need. I have 250GB here and use about 130GB


    --
    "What do you think about Gay Marriage?"
    "I don't."
    "Don't what?"
    "Think about Gay Marriage."



    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Theo@3:633/10 to All on Fri Nov 20 16:16:58 2026
    In comp.sys.raspberry-pi R.Wieser <address@is.invalid> wrote:
    Theo,

    If you keep some amount like 10% of the SSD zeroed then the
    controller can use those blocks to improve the wear lifetime
    ('over provisioning').

    Do you mean unpartitioned, or would something special (zeroed?) be needed ?

    'Empty'. When you delete a file, typically filesystems tell the device that the blocks are no longer used (this is called 'TRIM') and it can put them
    back into the empty block pool. The more empty blocks there are, the more breathing space the SSD has to use when parts of the flash wear out.

    If you zero blocks, the SSD can also decide they're 'empty'. If you leave parts unpartitioned but still containing data from the last time it was
    used, the SSD is still contractually obliged to keep the data in case you
    need it another time and so it can't reallocate them to cover wear.

    Just making a big file and then deleting it is usually enough to trigger
    TRIM of the blocks, but also zeroing the SSD and then leaving 10% (or
    whatever) unpartitioned works too.

    This can also help with 'performance consistency': if you want to write but
    the SSD has to do a lot of juggling to find an empty block, maybe erasing
    and shuffling some other blocks, then you can get erratic performance under heavy load. But that's only really an issue with big write workloads like database servers, not domestic desktop use.

    Theo

    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Richard Kettlewell@3:633/10 to All on Fri Nov 20 17:39:44 2026
    "R.Wieser" <address@is.invalid> writes:
    Theo,
    'Empty'. When you delete a file, typically filesystems tell the
    device that the blocks are no longer used

    Off by default in ext4, I believe.

    Ah, you mean I should never fully use the whole partition ("Used
    space" vs "Free space"). With 128GB to work with I'm not going to
    worry about that. :-)

    If you zero blocks, the SSD can also decide they're 'empty'.

    I don't understand : when a block is not part of any file chain than its unused. The wear-leveling process could commandeer such blocks at any
    time - even though that could twart a later "undelete" attempt.

    The drive doesn?t know which (logical) blocks are not part of a file,
    unless the OS tells it.

    --
    https://www.greenend.org.uk/rjk/

    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From R.Wieser@3:633/10 to All on Fri Nov 20 16:38:22 2026
    Theo,

    If you keep some amount like 10% of the SSD zeroed then the
    controller can use those blocks to improve the wear lifetime
    ('over provisioning').

    Do you mean unpartitioned, or would something special (zeroed?) be needed ?

    And thanks for the suggestion.

    Regards,
    Rudy Wieser



    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From R.Wieser@3:633/10 to All on Fri Nov 20 18:29:41 2026
    Theo,

    'Empty'. When you delete a file, typically filesystems tell the
    device that the blocks are no longer used

    Ah, you mean I should never fully use the whole partition ("Used space" vs "Free space"). With 128GB to work with I'm not going to worry about that.
    :-)

    If you zero blocks, the SSD can also decide they're 'empty'.

    I don't understand : when a block is not part of any file chain than its unused. The wear-leveling process could commandeer such blocks at any
    time - even though that could twart a later "undelete" attempt.

    Hmmm ... Does Linux even have an "undelete" method ?

    Also, files can contain lots of zeroes. How will the wear-leveling
    mechanism know that such in-use blocks filled with zeroes should not be
    taken ?

    If you leave parts unpartitioned but still containing data from
    the last time it was used, the SSD is still contractually obliged
    to keep the data in case you need it another time

    That doesn't make any sense. The data in blocks not part of any partition
    (or file!) needs to be kept ? For what reason ?

    but also zeroing the SSD and then leaving 10% (or whatever)
    unpartitioned works too.

    Again, I do not understand. The blocks in the unpartitioned area might be zeroed out, but the moment the wear-leveling mechanism swaps such a block
    out with one inside a partition that swapped-out block now in the unpartitioned area is going to contain data.

    Regards,
    Rudy Wieser



    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)