In message <u0s8nj$1bera$1@dont-email.me> at Sat, 8 Apr 2023 18:36:51,
Java Jive <java@evij.com.invalid> writes
[]
>Yes, previously, I've rather been put off SSD drives, because ...
[]
>... which between them give a combined failure rate of at least 15%,
>which I would have guessed was higher than that for conventional HDs,
>but now, trying to remember back systematically as best as I can over
>about 3 to 4 decades, actually I recall 5 early failures in at least
>about 25 HDs, or a maximum of around 20%, so for me SSDs certainly have
>performed no worse, and most probably have performed better, than
>conventional HDs, which I wouldn't have expected to be the case without
>systematically trying to recall the details of the HDs that I've had.
>
_My_ nervousness about SSD drives has been the _manner_ of failure - and
that's probably unfairly based on my experience with USB sticks: my
_feeling_ is that solid-state memory devices fail suddenly with no
warning, whereas spinning drives _tend_ to decline gradually. (Not
always I know: I had one where - I think - the head or heads
spot-welded, so obviously the drive suddenly stopped spinning! [It had
been in a laptop with a heating problem. After all the usual methods
failed, I actually opened it in a clean cabinet we had at work, which is
how I know what happened: I freed it, and got 95-98% of the data off,
though condemned it thereafter.]) But on the whole HDs give advanced
indication of failure: make funny noises, or - more often, I think - no
obvious indication (unless you keep running HDTune), just get slower and
slower as the ECC works harder. (I know someone whose XP - or might have
been '9x - machine was eventually taking a quarter hour to boot! It was
fine once it _had_ booted, unless you did something disc-intensive.)
Then there's the bit about SSDs having a write counter, and suddenly
becoming read-only when it passes a certain point - do they still do
that? - and one product line (Intel I think) which became a brick
(neither read _nor_ write, so you couldn't even rescue the data) at that
point.

Obviously, if you back up properly, none of this should matter, but …
(-:
--
J. P. Gilliver

On 5/3/2023 5:33 PM, J. P. Gilliver wrote:
> In message <u0s8nj$1bera$1@dont-email.me> at Sat, 8 Apr 2023 18:36:51, Java Jive <java@evij.com.invalid> writes
> []
>> Yes, previously, I've rather been put off SSD drives, because ...
> []
>> ... which between them give a combined failure rate of at least 15%, which I would have guessed was higher than that for conventional HDs, but now, trying to remember back systematically as best as I can over about 3 to 4 decades, actually I recall 5 early failures in at least about 25 HDs, or a maximum of around 20%, so for me SSDs certainly have performed no worse, and most probably have performed better, than conventional HDs, which I wouldn't have expected to be the case without systematically trying to recall the details of the HDs that I've had.
>>
> _My_ nervousness about SSD drives has been the _manner_ of failure - and that's probably unfairly based on my experience with USB sticks: my _feeling_ is that solid-state memory devices fail suddenly with no warning, whereas spinning drives _tend_ to decline gradually. (Not always I know: I had one where - I think - the head or heads spot-welded, so obviously the drive suddenly stopped spinning! [It had been in a laptop with a heating problem. After all the usual methods failed, I actually opened it in a clean cabinet we had at work, which is how I know what happened: I freed it, and got 95-98% of the data off, though condemned it thereafter.]) But on the whole HDs give advanced indication of failure: make funny noises, or - more often, I think - no obvious indication (unless you keep running HDTune), just get slower and slower as the ECC works harder. (I know someone whose XP - or might have been '9x - machine was eventually taking a quarter hour to boot! It was fine once it
> _had_ booted, unless you did something disc-intensive.) Then there's the bit about SSDs having a write counter, and suddenly becoming read-only when it passes a certain point - do they still do that? - and one product line (Intel I think) which became a brick (neither read _nor_ write, so you couldn't even rescue the data) at that point.
>
> Obviously, if you back up properly, none of this should matter, but … (-:
SSD drives have three-core ARM processors. There is a whack of
firmware in there, doing maintenance and maintaining "power-safe"
operations (keeping a copy of the translation table). Quite frequently,
when your hand is off the mouse, that three core processor
is doing stuff. The LED does not flash, when the three core processor
is on a rant.
The consumer SSD drives run without using a SuperCap. That's what
"power-safe" means, immediate power failure does not endanger
the "critical data" content. Power failures are also recorded
in SMART, so if you've been mis-treating your SSD, there is
a counter pointing at your misdeed. (I have a SATA to USB
converter that causes the power-failure counter to increment!
Not a builder of confidence, when a flush() was already sent.
This should not be happening. PC SATA SSD operation works fine.)
Early Enterprise drives had a SuperCap and the drive ran off
SuperCap energy, once the primary power feed was observed to
have gone away. This takes some of the pressure off writing
"power-safe" firmware. The SuperCap does not work (necessarily)
at rail voltage, and may use a boost converter to power
circuits on a failure. It only has to run for a second or two.
A few bucks worth of SuperCap would be enough, rather than
one of the $100 ones you could weld with :-)
Some consumer SSD drives, if you examine the PCB, you can see
the pads for the SuperCap (no part installed). The boost chip,
inductor and other bumpf, are also depopulated in the bill of
materials. Shopping on Ebay for a Supercap, isn't enough.
Some SSD drives have DRAM cache, cheaper ones do not. You won't
really find any discussion threads, where there is "evidence
these things exist". Presumably such cache, helps with wear
life and write amplification, best case.
USB sticks ? It's lucky they even have bypass caps.
There is nothing of value in a USB stick. Yes, a dinky
microcontroller is in there. A few USB sticks are "featureful",
but we have to take the word of manufacturer tech support,
and they're not known for information reliability. Most
modern USB sticks, die well before the computed wear life.
Even with no wear leveling, I should be able to write 600
times, and if it fails after 8 writes (dd.exe ==> ISO file),
you have to wonder.
They're just not in the same class.
There is much room for improvement, on USB sticks.
There are a couple sticks with both static and dynamic wear leveling.
But we cannot take some manufacturer tech support dood word for
this, because it could be abject marketing. Only the engineering
department at such a company, would know for sure.
Paul