torturing caps

"a fifth of a wavelength wide"? Wot's the reference?

but

would

issues

No one is saying that anyone does full parametric characterization on any parts. CPUs do in fact have a full logic test (some portions of which may be BIST, but still on a tester) and some sort of speed test done, however.

POST is a red herring.

looks

actually=20

maximum but=20

source

an

production

production

million

space),

at

military

realize if=20

statistical=20

device,

sheets

the

even 20

while.

At a minimum you will get verifiable chain of custody from incoming materials to outgoing product, a byproduct of which is lot tracking verifiability. Other than that, JAN-TX testing is not much different = from regular production with statistical process controls.

costs?=3D20

That is the same old thing about guaranteed performance and typical parameters; the very reason that you have managed to do some interesting things with parts operated further in the margin between typical and guaranteed performance.

?-)

ArF eximer lasers are now used to expose fineline ICs, with 193 nm light. I make the timing controllers for most of them. They are exposing features down to 20 nm or some such. Immersion optics reduces the physical wavelengths some. That's pushing the resists and such pretty hard.

It doesn't look like it's going to be worth it to develop shorter wavelength eximers. The next step may be incoherent EUV, blasting molten tin droplets with lasers or something bizarre like that, 13 nm light maybe. Or e-beam, if anybody can ever get the production rates up.

ASML has shipped a few scanners that have EUV light sources. Reportedly they cost about $110e6 each.

Maybe Moore's Law just stops and we live with it.

John

but

statistical

Are you saying that individual JAN-TX transistors are manufactured and shipped, untested?

John

Yeah, but synchrotron X-ray sources can get down to the 0.1 nm range, and all it takes is an ultrarelativistic synchrocyclotron and a good positron source... there's a half dozen of those in operation right now, just not pressed into manufacturing work.

in

wafer

looks

actually=20

maximum but=20

source

an

production

production

million

of

space),

money at

day

military

realize if=20

statistical=20

device,

couple

sheets

are

are

=20

since the

even 20

while.

from

Presuming that they are actually 100% burned in, that would be a = difficult to support claim.

It ain't so much what you don't know that bites you, but what you think = is so that ain't really so that bites you in the 6.

That applies to both of us.

But i do report my decade+ of test experience as honestly as i am able.