[clug] Effect of operating temps on life of integrated circuits

David Tulloh david at tulloh.id.au
Sun Jan 24 01:26:58 MST 2010


steve jenkin wrote:
> I've been trying to track down any estimates/rules-of-thumb about the
> effect of increased junction temperature on expected lifetime of IC's
> like CPU's and RAM - and was hoping the Collective Wisdom of CLUG (and
> some better Google-Fu than mine) could come up with quotable figures.
>   
This is real dark art stuff.  Having spent (literally) months working 
with a team trying to diagnose an IC failure it's not easy to figure out 
(we never really did).  There's a lot of stuff that could go wrong here 
and it's not predictable unless you work in the area (by which I mean 
being an Intel IC guy). 

As the articles said, temperature basically increases the rate of 
chemical reactions.  Temperature cycling induces physical stress based 
on things moving with the heat.

Chemical reactions that could occur that I am aware of are bond 
migrations and silicon whiskers.
Whiskers are the big threat from the new ROHS (no lead) mandates, 
essentially a thin whisker grows out from the joint causing a short if 
it hits another joint.
The bond migrations occur at the joins of the silicon and the outside 
world.  I think they cold weld the copper wires onto the silicon pads, 
if this is done too strongly the copper migrates into the silicon and 
the pad can rip off, too little and the copper wires get too thin and 
break.  The migration speed would be affected by temperature, the 
ripping off by temperature cycling.

You probably get migrations across PN junctions with temperature, which 
is what I think you were initially referring too.  That seems possible 
but most of the IC failures that I'm aware of came from the 
manufacturing of the package rather than breakdowns within the silicon 
itself.


My advice.  Forget about trying to determine your failure rate from the 
mechanisms of failure.  There's too many and you won't get enough 
information to properly determine the failure rate.  Rather just get 
figures of the failure rates of CPUs in general and assume they roughly 
hold.  Overclockers forums could give anecdotal information.


David


More information about the linux mailing list