On 2024-02-29, J. P. Gilliver wrote:
> In message <urq8q4$m82g$1@dont-email.me> at Thu, 29 Feb 2024 15:45:08,
>> [...]
>>Nope, I didn't really use em back in the day.
>
> Fairy nuff. (You make me feel old with "back in the day"; when I started
> out with printers, parallel port connections were the norm!)

Centronics connectors were great :)

--
|_|O|_|
|_|_|O| Github: https://github.com/dpurgert
|O|O|O| PGP: DDAB 23FB 19FA 7D85 1CC1 E067 6D65 70E5 4CE7 2860

"Paul" <nospam@needed.invalid> wrote

|
| But the Dell came from a factory too.
| | You're probably thinking that Sparkle Ponies live
| in a Dell factory.
|

:) I figure that Dell doesn't want returns, so they're probably
doing a lot of testing with components, to make sure they're
making a stable choice. I'm making a slightly educated guess
and hoping for the best.

| Sadly, all factories are the same.
| | Dell is unlikely to run its own PCB factory. Like
| HP, they would get a motherboard from Trigem or
| Pegasus or Mitac or Compal or some other ODM.
| | Anandtech had a video once, of the Asus motherboard
| "two minute functional test". This is the test before
| the motherboard is put in the ESD bag. It covers the
| main slots, the CPU socket, some memory slot. Most
| of the two minutes, is the time needed to plug in the
| test items. The actual test runtime is pretty short.
| | The reason the test is short, is it takes 2 minutes :-)
| Times the 3 million to 5 million motherboards per month
| receiving the same test. The result is five hundred tables
| with a person at each table, plugging in shit and testing.
| It is labor intensive. If any part of the process slows
| down... it result in a "need for more tables" :-) I think
| you can see the dynamic tension involved in the topic.
| If they plugged in all the USB ports, they would need
| a thousand tables. That's why we have the joke about
| the test time being short because it's short. Because it's short.
| | Naturally, the overall production loop, includes statistical
| product inspection, which is outside of our little functional
| test table fiasco. Every one of N items are pulled aside for
| thorough examination. If there's a problem, it could mean
| pulling pallets of stuff back from the production area, for
| rework or correction of systematic mistakes.
| | But ultimately, for "fatal" errors, the pile of material
| to be shredded and put in the dumpsters out back, that's
| huge. Most of the product 99 44/100 percent of it, *must*
| go out the door, or the tipping fee and wasted materials
| take all the fun out of it.
| | At some point, things defy logic. If a motherboard retails
| for $40 (which might be a clearance price for all I know),
| there isn't money for test. There should be some price point,
| where only statistical process control ensures product quality.
| This calls for six sigma ICs. Jelly bean TTL logic of years
| ago, that wasn't being tested (only statistical test, if that).
| That's well before someone thought the word "sigma" should
| be involved.
| | While we have expectations of a test strategy, remember that
| every factory is a dirty grubby place filled with morons.
| You can have 99 employees of normal intelligence, and the
| moron running the facility, spoils it for everyone else :-)
| Every plant has a weakest link. Plants have unions. Etc.
| I'm not really expecting thorough testing. What can we expect,
really, given that these are advanced components for relatively
cheap prices? But in all the machines I've built (maybe 15?) only
one has been a lemon. I just figure that as part of the cost.

On 2/29/2024 10:03 PM, Newyana2 wrote:
> "Paul" <nospam@needed.invalid> wrote
>
> |
> | But the Dell came from a factory too.
> |
> | You're probably thinking that Sparkle Ponies live
> | in a Dell factory.
> |
>
> :) I figure that Dell doesn't want returns, so they're probably
> doing a lot of testing with components, to make sure they're
> making a stable choice. I'm making a slightly educated guess
> and hoping for the best.

Dell is a box-assembler.

This means they want "modules" of some sort, tested to some level.

The brute force test you're thinking of ("test all the USB ports
just before computer goes into cardboard box"), is only one of the
ways of attaining equivalent product quality. There are other ways to
do it. And the field defect database, keeps track of how good a
job you're doing. If you are using statistical inference, then
that should be reflected in the defect database.

Even if a guy tests all the ports, before the computer is boxed,
it still doesn't have to work when you get it (infant mortality).

*******

A year ago, I bought three WD Black 1TB hard drives. One of the
three drives, the motor would not spin. Took it back to the
store, the guy in the computer assembly area handles returns,
and he could not get it to spin either. What's wrong with that ?
The motor has to spin, for the drive to be manufactured. It's
an integral part of the process. The crates drives ship in,
are "designed" crates. They're not random cardboard boxes.
They're meant to prevent shocks of more than a certain amount.
It's when a drive comes out of a crate, that when its "shock life" begins.

And that's my first "out-of-the-box" hard drive failure.

Paul