OT Tablets are cheap!

  • Thread starter Thread starter John Doe
  • Start date Start date
Paul said:
It's all about modeling.

You collect field failure data, to determine how a thing fails.
Then, you fit the appropriate statistical model to it. And
try to justify what you're seeing.

For example, the Seagate page has an MTBF calculation. It
assumes a bathtub curve. The calculation would only be
valid, if the actual failures follow that curve. In fact,
the reports I've seen, show hard drives follow a "wearout curve"
and not a classical bathtub. Which means the modeling is wrong,
and the MTBF is meaningless. (The MTBF is used to predict
how many spare hard drives to buy and keep in your
stockroom.) If you've been buying spare hard drives
using the Seagate info, you'll have too few in your
stock room.

Now, what curve does memory follow ? It might be
a bathtub, in an ideal world, but when other issues
are taken into account, the curve might be something
entirely different. And thus, your ability to predict
failures, is compromised.

I never had any memory failures, back in the FPM/EDO
days. (Typical machines here, had eight sticks.)
I've seen more failures in the succeeding generations.
And I'm not seeing anything to suggest there is a bathtub
waiting for me. The failures have been with generic RAM,
and always in the 1.5 year range (1.5 years of daily use).
For branded RAM purchased in the same generation, it's
still working.

I can give you other, non-statistical examples. I
bought 8 sticks of 512MB RAM, generic, without any
markings on the chips. I got the memory for "half price".
I insert three of the eight sticks, in a P4 motherboard.
I test the memory thoroughly. No errors.

I put the motherboard in storage for several years, with the
RAM sitting in the socket.

Later, I pull the PC out of storage (cool and dry locale,
not the garage, not the basement). I test. All three sticks
are showing errors. I take the other five sticks out of their
anti-static tray and I test them. All five sticks
pass error free.

Yes, the event is not statistically significant. But,
look at the symptoms. What are the odds, that all
three sticks just randomly decided to fail while
in storage (not under bias) ? It's a very puzzling
set of circumstances. And fitting a bathtub to crap
like that, would be pointless. This isn't classical
failure behavior at all, and is indicative of corrosion
or electrochemical processes.

Paul
Click to expand...
You should have tried the original three sticks again. Many times, just
reseting the sticks will take care of problems. They need to be moved in the
soctkes.

Henry
 
Henry said:
You should have tried the original three sticks again. Many times, just
reseting the sticks will take care of problems. They need to be moved
in the soctkes.

Henry
Click to expand...

Done and done.

Paul
 
Back
Top