13 Jun
2006
13 Jun
'06
6:43 p.m.
Antonio Carlini wrote:
I thought this was effectively more significant on the
smaller geometries. I.e. that the older devices made with
larger features were more immune (well, not "immune" as
much as "takes longer to dig through a 1 micron feature
than it does a .25 micron one!). So, folks with their shiney
new 4GHz machines can expect them to fail much quicker than
an older 100MHz box.
[Aside -- I wonder if the number of *instructions executed*
is, coincidentally, a constant? If so, folks running those
machines burning billions of opcodes waiting for the next
keystroke are getting the electronic "shaft"! :> ]
And, powered up devices greatly accelerate the process.
Has anyone any
"real" data on EM effects?
Or is this effectively urban legendry?
I have read in several places (industry papers and the like)
that hot-carrier injection does cause localised damage that
eventually leads to failure. But even if the effects have been exagerated, solid
state
Diffusion is real and will kill a chip even if it is never
powered up. What I don't know is whether the problem will
show up in 10 years, 100 years or 1000 years.