[Linux-aus] pcie errors

[Adam Nielsen]

> Further to Arjen’s good advice, I’ll support the assertion that WD-40
> is not a thing to be applied to electronics.

Russell wasn't talking about WD-40, the link was to a WD-40 branded
electrical contact cleaner, so it's an entirely different product than
the original WD-40 and it presumably will be fine since it's designed
for cleaning of electrical contacts.

However it's twice the price of Supercheap's option so it might pay to
look around:

https://www.supercheapauto.com.au/p/sca-sca-electrical-contact-parts-cleaner-350g/676267.html

Total Tools also have a good quality option if you want to avoid
Bunnings given their recent controversies, but it costs a few dollars
more than Bunnings:

https://www.totaltools.com.au/21184-crc-350g-co-contact-cleaner-2016

Isopropyl alcohol works fine, but the proper electrical contact
cleaners have a number of other chemicals in them that do a better job
of removing other types of dirt.  I used to use isopropyl but you need
to be able to physically clean the surface once covered in alcohol (e.g.
with a cotton bud) so as that is not very easy for smaller connectors
like USB plugs/sockets or strange shapes, spraying some contact cleaner
in and waiting for it to evaporate does a better job.

> I doubt that memory errors are happening as the system uses ECC RAM
> so any 1 bit error will be corrected and logged and every 2 bit error
> will be logged and flagged.  As I have an odd number of DIMMs
> installed the advanced ECC mode isn't available so if there are 3 or
> 4 bit errors (EG one chip is bad on a DIMM) then they could get past.

Yes you're right, sorry I assumed you were talking consumer hardware.
If you are using something with ECC RAM then it is probably more
accurately reporting PCIe errors as well, so less likely to be a RAM
issue.

> Is hardware store stuff good enough for cleaning DIMM contacts?  A
> quick search turned up WD-40 at Bunnings, it seems that WD-40 is the
> brand and they have a range of different types of cleaning spray in
> the same type of can which is annoying.

Yes I don't know why they kept the WD-40 brand name when it's already a
product name too, that's very confusing.  Any product should be fine as
long as its primary purpose is stated as cleaning electrical contacts.

> Al Maclang suggested isopropyl alcohol.  Would that be the main
> ingredient in those contact cleaners?

I'm sure they have isopropyl but last time I looked at the MSDS for one
it had many other ingredients as well.  Isopropyl is a great solvent
and will dissolve the dirt, but it just sort of sits there and
redeposits again when the isopropyl evaporates.  The contact cleaners
seem to do a better job of shifting the dirt out of the way.  I'm not
sure if it's the chemicals or the fact that they are all pressurised
spray cans, but whatever the reason you do get a better result with a
dedicated contact cleaner.

You do need to use the contact cleaner in a somewhat ventilated area
though.  If you breathe the fumes in you quickly get a headache, it's
strong stuff.  You don't have that problem with isopropyl.

> The z640 systems I have aren't ones with hot-swap PSUs so I'm not
> enthusiastic about swapping those, it's a significant pain to take
> them out.  Has anyone made a device for monitoring PC power?  It
> would be nice to have a PCIe card that connects to a SATA power cable
> and tells you the voltage on each line.  I did a quick check on
> AliExpress and couldn't find any such device.

There are such devices for ATX power supplies, however they aren't that
accurate.  It's usually not the voltage that's the problem, it's the
stability.  If the power supply doesn't do a good job of regulating the
power, then the voltage will go up and down as the electrical load on
the system changes, and that can use the voltages to dip too low, but
only for a split second, too low to spot with a volt meter.

Possibly a multimeter with min/max could react fast enough to capture
the minimum voltage on a given line, but you might need an oscilloscope
to properly see what's happening.

If it's a higher end machine or an enterprise server, typically they
will have very good quality power supplies in them so if it's less than
10 years old then the power supply is probably fine.

It tends to be the lower cost power supplies you find in budget
unbranded computers that cause these sorts of problems.

Cheers,
Adam.