c**i
发帖数: 6973 | 1 Ken Liu, TSMC Likely to Advance 28nm Production Schedule. CENS, Oct. 14,
2010.
http://cens.com/cens/html/en/news/news_inner_34054.html
Note:
(a) CENS stands for China Economic News Service 中國經濟通訊社.
http://www.cens.com/censv1/zh/aboutus.jsp
(中經社簡介: 簡稱中經社; 於1974年成立; 聯合報系的關係企業之一)
(b) Altera Corp.
http://www.altera.com/corporate/crp-index.html
(About Altera: Headquartered in Silicon Valley; the leader in innovative
custom logic solutions; Innovative Technology: Working closely with our
manufacturing partner, Taiwan Semiconductor Manufacturing Company (TSMC), we
’re building the industry’s most advanced FPGA, CPLD, and ASIC
technologies.)
My comment:
(a) FPGA: field-programmable gate array
CPLD: complex programmable logic device
ASIC: Application-specific integrated circuit
(b) By comparison, Intel just (yesterday) annonced that its 22-nm Ivy Bridge
processor is slated to come out in the second half of next year, whereas 32
-nm Sandy Bridge is due to appear in 1Q11. For the latter, see
Sandy Bridge (microarchitecture)
http://en.wikipedia.org/wiki/Sandy_Bridge_(microarchitecture) | b**i
发帖数: 257 | 2 I still remember my undergrad teacher saying that quantum effect will
prevent the processing technology going below 40 (?) nm. Anyone can educate/
update me on this? What is the theoretical limit of this?
【在 c**i 的大作中提到】
: Ken Liu, TSMC Likely to Advance 28nm Production Schedule. CENS, Oct. 14,
: 2010.
: http://cens.com/cens/html/en/news/news_inner_34054.html
: Note:
: (a) CENS stands for China Economic News Service 中國經濟通訊社.
: http://www.cens.com/censv1/zh/aboutus.jsp
: (中經社簡介: 簡稱中經社; 於1974年成立; 聯合報系的關係企業之一)
: (b) Altera Corp.
: http://www.altera.com/corporate/crp-index.html
: (About Altera: Headquartered in Silicon Valley; the leader in innovative
| a***e
发帖数: 27968 | 3 for silicon, the limit is around 5nm for quantum effect.
educate/
【在 b**i 的大作中提到】
: I still remember my undergrad teacher saying that quantum effect will
: prevent the processing technology going below 40 (?) nm. Anyone can educate/
: update me on this? What is the theoretical limit of this?
| a***e
发帖数: 27968 | 4 that is because TSMC decided to skip the 32nm node.
TSMC has been on half-node for a while, 65, 50, 45, 40 due to its business
model.
in term of process, no much difference between 28nm and 32nm.
performance wise, intel 32nm is tuned for high performance
while TSMC 28nm is tuned for low power.
HKMG and Strain Silicon, intel is 1.5gen ahead.
【在 c**i 的大作中提到】
: Ken Liu, TSMC Likely to Advance 28nm Production Schedule. CENS, Oct. 14,
: 2010.
: http://cens.com/cens/html/en/news/news_inner_34054.html
: Note:
: (a) CENS stands for China Economic News Service 中國經濟通訊社.
: http://www.cens.com/censv1/zh/aboutus.jsp
: (中經社簡介: 簡稱中經社; 於1974年成立; 聯合報系的關係企業之一)
: (b) Altera Corp.
: http://www.altera.com/corporate/crp-index.html
: (About Altera: Headquartered in Silicon Valley; the leader in innovative
| t****g
发帖数: 35582 | 5 然后呢？就看石墨烯的了？
【在 a***e 的大作中提到】
: for silicon, the limit is around 5nm for quantum effect.
:
: educate/
| c**i
发帖数: 6973 | 6 (1) First off, I am a biologist by training. I did take physics 101 in
college, but never took quantum physics, to my eternal regret.
(2) When I prepared the original posting, I read this though I did not cite
it (but this report is the basis of my statement regrading Intel
announcement--"yesterday").
Brooke Crothers, A peek into the future of Intel processors. CNET, Oct. 19,
2010.
http://news.cnet.com/8301-13924_3-20020078-64.html
(15 nm in the lab)
(3) About your question, first let's hear what the namesake of the law has
to say (he did not explain on record). It is far off (a decade or so away),
and there may be breakthroughs by then.
Manek Dubash, Moore's Law is dead, says Gordon Moore; Key predictor of IT
will end sometime, reckons its progenitor. Techworld, May 13, 2005.
http://news.techworld.com/operating-systems/3477/moores-law-is-dead-says-gordon-moore/
(Mr. Moore: "In terms of size [of transistor] you can see that we're
approaching the size of atoms which is a fundamental barrier, but it'll be
two or three generations before we get that far - but that's as far out as
we've ever been able to see. We have another 10 to 20 years before we reach
a fundamental limit.")
(i) I wish to remind you that since 2005 "two or three generatios" have
actually been accomplished in this period, as Moore's Law had predicted.
(ii) This interview was cited in
Moore's law
http://en.wikipedia.org/wiki/Moore's_law
(iiii) This is his last prediction. Why? Read on.
Jonathan Fildes, Meeting the man behind Moore's Law. BBC<
http://news.bbc.co.uk/2/hi/technology/7080646.stm
Quote:
"In the article in Electronics Magazine, he predicted that the number of
transistors on a silicon chip would double every year for ten years. He
later revised the forecast to doubling every two years or so, as the initial
breakneck speed of development and shrinkage waned.
"So what does he think will happen in the next 40 years? 'I'm through with
making predictions,' he chuckles. 'Get it right once and quit.'
(4) However, one need not be a quantum physicist to comprehend the theory
that Moore's law will hit a wall sometimes.
Dana Blankenhorn, Moore’s Law reaches its limit with quantum dot amplifier.
Smart Planet, May 25, 2010.
http://www.smartplanet.com/technology/blog/thinking-tech/moores-law-reaches-its-limit-with-quantum-dot-amplifier/4219/
("But there has to be a limit. The atomic scale. You can’t make a circuit
smaller than an atom. The limit has been reached.")
* Not really, in my opinion as a layperson. See the right table in
Atom
http://en.wikipedia.org/wiki/Atom
(Diameter range: 62 pm (He) to 520 pm (Cs))
(5) Attempts have been made to circumvent the limitations.
(a) US military to attack Moore's Law for future computers. BBC, Aug. 10,
2010.
http://www.bbc.co.uk/news/technology-10924841
Note:
(i) The report did not talk about solutions. Come on, this is military
secrets.
(ii) quintillion: base 10 and exponent 18, or 10 to the 18
(iii) The BBC report mentioned "current top supercomputers which manage just
over one petaflop - 1000 trilliion calculations per second," which is 10 to
the 15.
* peta-
http://en.wikipedia.org/wiki/Peta-
Click "exa" in this web page.
* FLOPS: FLoating point OPerations per Second
(iii) The BBC report also state, "The limitations of that approach are the
mushrooming power, management and structural issues that crop up as
components shrink."
To which one should add cost of building a foundry, which balloons too with
successive generations.
(b) Bill Dally, Life After Moore's Law; It's time for the computing industry
to take the leap into parallel processing. Forbes, Apr. 29, 2010.
http://www.forbes.com/2010/04/29/moores-law-computing-processing-opinions-contributors-bill-dally.html
(the writer is Nvidia VP)
(c) Addy Dugdale, Quantum Computing Gets a Step Closer With Seven-Atom
Transistor. The Fast Company, May 24, 2010.
http://www.fastcompany.com/1651617/quantum-computing-gets-a-step-closer-with-seven-atom-transistor
educate/
【在 b**i 的大作中提到】
: I still remember my undergrad teacher saying that quantum effect will
: prevent the processing technology going below 40 (?) nm. Anyone can educate/
: update me on this? What is the theoretical limit of this?
| c**i
发帖数: 6973 | 7 My original posting did not imply either is better. The report about TSMC
does not say if 28-nm fabrication will be commercial/mass production. And
you are aware of the fact that TSMC encountered difficulties in 45-nm, and
the then CEO lost his job.*
* On Oct. 6, 2010 Nvidia CEO Jen-Hsun Huang denied rumor and said Nvidia "
Won't Have Globalfoundries Make Tegra, [but] Sticks to TSMC." (headline of a
report).
【在 a***e 的大作中提到】
: that is because TSMC decided to skip the 32nm node.
: TSMC has been on half-node for a while, 65, 50, 45, 40 due to its business
: model.
: in term of process, no much difference between 28nm and 32nm.
: performance wise, intel 32nm is tuned for high performance
: while TSMC 28nm is tuned for low power.
: HKMG and Strain Silicon, intel is 1.5gen ahead.
| c**i
发帖数: 6973 | 8 I just corrected the link, with BBC report on US miliatry
http://www.bbc.co.uk/news/technology-10924841
* * *
【在 c**i 的大作中提到】
: (1) First off, I am a biologist by training. I did take physics 101 in
: college, but never took quantum physics, to my eternal regret.
: (2) When I prepared the original posting, I read this though I did not cite
: it (but this report is the basis of my statement regrading Intel
: announcement--"yesterday").
: Brooke Crothers, A peek into the future of Intel processors. CNET, Oct. 19,
: 2010.
: http://news.cnet.com/8301-13924_3-20020078-64.html
: (15 nm in the lab)
: (3) About your question, first let's hear what the namesake of the law has
| c**i
发帖数: 6973 | 9 This is HKMG.
http://www.intel.com/technology/silicon/high-k.htm
But strained silicon (from IBM or Intel): all I can find in the web are old,
about eight years old.
* * * HKMG and Strain Silicon, intel is 1.5gen ahead.
【在 a***e 的大作中提到】
: that is because TSMC decided to skip the 32nm node.
: TSMC has been on half-node for a while, 65, 50, 45, 40 due to its business
: model.
: in term of process, no much difference between 28nm and 32nm.
: performance wise, intel 32nm is tuned for high performance
: while TSMC 28nm is tuned for low power.
: HKMG and Strain Silicon, intel is 1.5gen ahead.
| a***e
发帖数: 27968 | 10 intel put that in production at 90nm, that is almost 10years ago.
tsmc put it in recently
old,
【在 c**i 的大作中提到】
: This is HKMG.
: http://www.intel.com/technology/silicon/high-k.htm
: But strained silicon (from IBM or Intel): all I can find in the web are old,
: about eight years old.
:
: * * * HKMG and Strain Silicon, intel is 1.5gen ahead.
| a*******i
发帖数: 11664 | 11 graphene是肯定没戏的
【在 t****g 的大作中提到】
: 然后呢？就看石墨烯的了？
| c**i
发帖数: 6973 | 12 I am not an expert on semiconductor.
So This is what I will say.
It is correct that TSMC is sending out 28-nm with high-k metal gate (HKMG)
technology (in 4Q10).
However, other statements in your posting are incorrect, as far as I can
tell.
(1) Bogdan Botezatu, Intel and TSMC Go for Advanced High-k Technology.
Softpedia, Dec. 13, 2007.
http://news.softpedia.com/news/Intel-And-TSMC-Go-For-Advanced-High-k-Technology-73742.shtml
("The International Electron Devices Meeting surely was a good place for
Intel to boast their newest technological breakthrough. The company have
described their 45-nanometer logic technology, the first to integrate in
high-volume manufacturing process high-k/metal gate transistors. They were
not the only ones however to achieve spectacular results, as Taiwan
Semiconductor Manufacturing Co. (TSMC) also reported successful results, but
on a 32-nanometer architecture")
(2) Bohr, MT, Chau RS, Ghani T and Mistry K, The High-k Solution;
Microprocessors coming out this fall are the result of the first big
redesign in CMOS transistors since the late 1960s.
http://spectrum.ieee.org/semiconductors/design/the-highk-solution/0
(read only paragraph immediately following the heading "Of course, we weren'
t alone")
The graphic in this web page is too small, but other documents within and
without Intel indicates Intel shipped out 45-nm chip (code-named Penryn) of
HKMG technology in 4Q07. I disagree with the following report to the extent
that Intel has shipped TWO generations of high-k. In my opinion, only one
generation, at 45-nm (32-nm is expected in 1Q11).
(3) Mark LaPedus, IBM 'fab club' denies problems with high-k. Sept. 27, 2010.
http://www.eetimes.com/electronics-news/4208861/IBM--fab-club--denies-problems-with-high-k-semiconductor
Quote:
"Intel Corp., Taiwan Semiconductor Manufacturing Co. Ltd. (TSMC) and others
have moved towards ''gate-last'' technology. Intel has been shipping
processors based on high-k since the 45-nm node. Intel has shipped two
generations of high-k. TSMC has yet to ship its technology.
"In contrast, IBM's 'fab club' is using a 'gate-first' technology. But to
date, IBM's technology partners [including Infineon, NEC, Globalfoundries,
Samsung, STMicroelectronics and Toshiba] have yet to ship high-k--at least
in chips in mass volumes.
【在 a***e 的大作中提到】
: intel put that in production at 90nm, that is almost 10years ago.
: tsmc put it in recently
:
: old,
| c**i
发帖数: 6973 | 13 What is k?
High-k dielectric
http://en.wikipedia.org/wiki/High-k_dielectric
(dielectric constant κ)
【在 c**i 的大作中提到】
: I am not an expert on semiconductor.
: So This is what I will say.
: It is correct that TSMC is sending out 28-nm with high-k metal gate (HKMG)
: technology (in 4Q10).
: However, other statements in your posting are incorrect, as far as I can
: tell.
: (1) Bogdan Botezatu, Intel and TSMC Go for Advanced High-k Technology.
: Softpedia, Dec. 13, 2007.
: http://news.softpedia.com/news/Intel-And-TSMC-Go-For-Advanced-High-k-Technology-73742.shtml
: ("The International Electron Devices Meeting surely was a good place for
|
|
