"This chip breakthrough will revoutionize the industry!!"
<<Message #138078 from Dave at Jun 25, 2001 1:13 PM
Off Topic
Paul,
I seem to recall that there has been alot of research on SiGe for numerous years. Approximately how long does it take from initial research for a technology to become utilized in industry? >>
I'm not Paul, but I'll comment anyway. Briefly, in bullet form.
* Remember laser pantography? "It's going to change the way chips are made!!" Never got commercialized. Remember molecular beam epitaxy (MBE)? Ditto. Remember silicon compilers? Remember magnetic bubbles? Josephson junctions? Ferroelectric RAMs? Holographic memories? Neural nets? Wafer-scale integration? Aptical foddering? Fifth Generation computers? Synchrotron radiation lithography? I-squared L?
* None of these breakthroughs are core constituents of today's mainstream. Some are niche products. Most are not even that.
* Research labs, universities, grad students, and hucksters produce many alleged breakthroughs each year. The tagline is usually "This will revolutionize...."
* Revolutions require a lot more than laboratory results. This applies to IBM's current SiGe transistors and *ALSO* to Intel's "3 atoms wide" transistors. (A point I commented on a few weeks ago when the touters were touting Intel's "breakthrough.")
* During the 70s, every month or so brought a breathless article in "Electronics" (a fine mag back then, now largely or completely defunct) about some new breakthrough. SOS was of course heavily touted out of the research labs at RCA. I-squared L (I^2L, or Integrated Injection Logic) was routinely predicted to be the "MOS Killer."
* Nearly _none_ of these "revolutions" ever came close to being commercialized. Which is not at all surprising. Commercialization is a creature of many dimensions, including cost, infrastructure, and competition from other approaches.
* The technologies which do manage to be commercialized tend not to be the obscure technologies developed in research labs by companies still funding "basic research."
* Look to what the equipment companies are developing in conjunction with the chip companies--e.g., Applied Materials working with Intel--if you want to see developments likely to be commercialized in the next several years.
* To be sure, new developments of course get commercialized. But usually the ideas are not very "fringey" (fringe ideas).
* Truly exotic materials are usually "not ready for prime time." Replacing just aluminum with aluminum-copper and similar variants took a decade of work by metallurgists. No doubt the first "grad student" experiments, and the first IBM Yorktown Heights experiments, were done in the 60s and 70s. I remember papers on copper interconnects when I first arrived at Intel in '74.
* Bear this in mind when reading breathless stories about how some laboratory experiment means a revolution in chips is about to occur.
It's not a matter of cynicism, it's a matter of realism.
--Tim May |
