Peter Jason wrote:
On Sun, 13 Jan 2019 16:14:36 -0500, Paul
wrote:

Peter Jason wrote:
I have isolated an SSD because it was very hot to
the touch (others are barely warm) and this has
fixed recurrent "scan/repair disks" flags in the
RHD notifications screen. Is this characteristic
of failing SSDs?
An SSD could be hot, because it did not receive
any cooling air. Some laptop drive bays are almost
completely insulated, and there is no airflow.

Sometimes, a laptop with such a design, ships with a
5400 RPM drive in it. Specifically, because the manufacturer
knows the bay is "power limited" and a higher performance
drive will overheat in there.

*******

The drive could run hot, as a function of the
"level of activity".

M.2 drives (running at up to 2500MB/sec), tend to hit
thermal limits more than conventional SATA SSDs (500MB/sec).

CMOS power dissipation is a function of FCV^2. And F is the
toggle frequency (as an output toggles between 0 and 1 and
back to 0 again.

We can't tell exactly what the SSD drive is doing all
the time, as it has its own processor and background
maintenance processes.

The spec sheet for the drive lists various power specs.
Like, how much power it draws on a write, and how
much power it draws when (nominally) idle.

If you had a clamp-on DC ammeter, you could measure
the current flow on the +5V wire of the SATA power cable.
Maybe a drive I have here, draws around 500mA max (as the ones
I've checked, seem ready to use with USB ports). Using your
clamp-on DC ammeter, you can measure the current flow and
get some idea whether the device is radically outside
of norms or not.

A ceramic capacitor could fail short, or a solder joint could
have a shorted solder connection, and that might
make something hot. That kind of thing has been seen
on DIMMs before, and it's a bit weird, as the DIMMs are
supposed to receive a quick test before they go into the
package. But if the drive is uniformly hot, it could be
that the drive has no air cooling. If the drive only
dissipates 0.15W at idle, it's going to be pretty hard
to raise the casing to 60C that way.

If you have a Kingston SSD, those use SandForce controllers
that compress write data and decompress read data. And that
burns up more power than drives that don't compress. The spec sheet
for the drive, should reflect that extra power usage.

Paul

Thanks, it was a Samsung 750 EVO SSD about 18
months old. Are there super-grade (military
quality) SSDs available with longer life spans?
And are the SSDs that plug into PCE slots any more
reliable? I've always known SSDs to run cool.

The SMART table for the drive has a temperature readout.
You could try that, to quantify the temperature.

A Samsung probably doesn't use data compression like
the Sandforce design does, so there isn't a reason for it
to run quite as hot.

And "hot" is when you burn yourself. When you cannot hold
your finger on electronics for more than 2 seconds, that's
about 60-65C or so. I used to work with boards of ECL logic,
and the entire surface of a running design can be that hot.
(There's no place to lay your scope probe hand.)
But the chips on those could take a lot more heat, more than
CMOS, and as most practitioners would note, those "only
start to work well when they get hot". Whereas the CMOS
everything is made of now, is not the same. The IC packages
have changed too, and fewer ceramic packages are used, and
more plastic for the smaller ones (like the flash) are used.
If you don't have a spec sheet, a temp of 99C is a good upper
limit for modern electronics (the "organic" packaging starts
to degrade above that temp, whereas the silicon can suffer
over a period of time if the temp is 135C junction - parametric
shift). And those kinds of temperatures will raise a blister
for sure. That's the boiling point of water+.

Some SSD housings have thermal tape between the tops
of the IC packages and the metal casing, to enhance
heat transfer. Which is a bit strange when the power
dissipation is in the 2.5W range. Usually companies
try to save money, and let the chips run hotter. (You
see this on TV tuner dongles perhaps, in USB stick format,
an absence of caring about chip temps.)

Paul