Friday, January 07, 2005

Moore's Law progress report

Moore's Law predicts exponential growth in the number of transistors on microchips, with a doubling time of roughly 1-2 years. As discussed here, the prediction continunes to hold for transistors but not necessarily for the overall performance of the chip, as clock speeds have lagged of late. There is some hope that through concurrency (multi-threading of processes), the performance gain can be made up, but this will require fundamental changes in software design.

Moore's Law is at the intersection of physics and economics. We haven't yet reached the point of fundamental physical limitations on chip performance, although we may be only a decade or two away (people have been saying that since Moore formulated his law). But perhaps economic factors are beginning to intrude - the rate at which new fabs are brought on line is determined in part by market demand for bleeding edge performance. Only a very small subset of the market runs applications that need a 4GHz chip.

16 comments:

Anonymous said...

http://www.j-bradford-delong.net/movable_type/2005-3_archives/000132.html

Department of "Huh"?

The Economist writes that "consumers face stark choices" because of a "mobile-energy crisis" in battery technology...

Technological progress in battery technology is slow only by comparison with progress in the information-technology trinity: microprocessors, memory, and hard disks. The fact that the Economist can call an 8% per year growth rate a "crisis" is an index not of slow technological progress but of our extraordinary expectations for improvement.

Anne

Anonymous said...

We have already been pushed to rely more than ever on asset price increases to fund our retirements as defined benefit pension plans have changed to defined contribution plans. Now we are to be asked to consider private investment plans to fund our Social Security needs. However as we become ever more dependant on asset accumulation and price increases, the relatively high prices of stocks and bonds and real estate should warn us that returns going forward are likely to be considerably lower than returns from any period between 1974 and 2000. Nonetheless, we must invest and the topic is critical for our well-being.

Anne

Anonymous said...

http://www.nytimes.com/2005/01/08/business/08skin.html?pagewanted=all&position=

Cosmetics Break the Skin Barrier
By CLAUDIA H. DEUTSCH

rocter & Gamble is about to sell ball bearings - but not the metal kind. Its minuscule mineral spheres are designed to help usher its Olay-brand body lotions deep into the skin.

Freeze 24/7, meanwhile, is pushing the muscle relaxant GABA, or gamma-amino butyric acid, a common ingredient in over-the-counter antianxiety supplements. It is not using GABA to relax minds, however. Instead, the goal is to relax the muscles that cause face wrinkles.

'We knew that if we could find a way to use GABA topically, it would be a killer app,' said Scott E. Gurfein, the founder of the year-old company.

The science of smoothing women's skin is going high tech. And cosmetics companies, whether they serve the masses or the elite, are adopting not just the language of Silicon Valley but many of its most sophisticated techniques.

Researchers for cosmetics companies have spent several years developing chemical bullets to attack wrinkles. But now, the players in this growing industry are turning to the medical and electronics worlds for ways to keep human skin from bouncing those bullets off the body like so many blanks.

'What you are seeing in the skin care world is a mirror of the advancing technology in pharmaceuticals and biotechnology,' said Karyn Grossman, a dermatologist and international spokeswoman for the Prescriptives line of Estée Lauder.

Scientists from far outside the cosmetics world are noticing the change. 'Skin care,' said Neil Gordon, president of the Canadian NanoBusiness Alliance, 'is definitely becoming a big area for nanoscience,' which involves working to manipulate nature at the supersmall level of individual atoms and molecules.

Anne

Anonymous said...

http://bigpicture.typepad.com/comments/2005/01/an_interview_wi.html

An Interview With Jeremy Grantham - Barrons

Asset allocators will be sorely tested in the coming year, says a skilled practitioner of the art:

Duck and cover. That about sums up this investing great's view of how best to cope with what he thinks is likely to be a difficult year in the financial markets. A founder of Boston-based Grantham, Mayo, Van Otterloo, and steward of $80 billion for institutions and the wealthy, Grantham sizes up asset classes and seizes the best opportunities across the broad spectrum of available investments. Now, for the first time in his career, he's faced with what he calls the "asset-allocator's nightmare": asset classes of all stripes are either overpriced or fairly priced. So, what's an investor supposed to do? Grantham offered his advice in an interview conducted in late December. For some of the best advice in the business, please read on:

"Underweight equities globally because they are expensive. Only emerging-market equities are close to fair value, and now could be considered a little overpriced."

Anne

Anonymous said...

http://bigpicture.typepad.com/comments/2005/01/an_interview_wi.html

EXCERPT:

Q: The past few years have been very rough on short sellers. Are you suggesting a new golden period for them?

A: It will probably be a pretty darn good period, and conservative hedge funds are the answer to a maiden's prayer. Even if they turn out to deliver a lower return than you expect going in because of the sheer volume of the competition -- say you expect 10% but get 5% -- that may turn out to be brilliant in the next couple of years. I personally have 50% of my money in such funds, which includes a net short fund but mostly conservative market-neutral funds. Of the balance, I'm very long emerging and international. I've shorted a few S&P 500 contracts against them, just to take some of the sting out of the next few years. Our highest-confidence strategy would be going long emerging markets and short S&P 500 contracts.

Our seven-year forecast for asset classes expects a real return of around 6% from emerging markets. We expect to add three or four points to that from active management for a total return of 9%. We expect the S&P 500, on the other hand, to deliver minus-2%. That's an 11-point spread. And emerging markets have never looked higher-quality than now. Their reserves are fabulous. Their currencies, consequently, all look pretty good. Their growth rates are fabulous. The back page of the Economist lists the 24 emerging countries. If you were to take the gross-domestic-product growth rate of the European Union for the last year and add it to that page, it would rank dead last on a list that includes Egypt, Israel, Turkey and Slovakia. They are all beneficiaries of China, have a terrific fundamental strength at the moment and a good qualitative position. Still in all, they may go bump in the night. But on a three-year basis, I think they are going to hammer the S&P again.

Anne

Anonymous said...

The Federal Reserve has been criticized for failing to try to adjust asset prices with monetary policy. We might think through whether the Fed should interfere with asset prices other than in the bond market. My leaning is against any such specific price interference, but after all every Fed move effects asset prices in various ways so I am open to argument.

Anne

Anonymous said...

The problem then is we came out of a fierce bear market with still high price earning ratios for the S&P. To cope with the recession around the bear market the Fed lowered interest rates sparking elevated prices for bonds and real estate. We now have expensive assets through and through. Whether commodity prices are are elevated is not clear to me, but there is little investment value in the American economy unless traditional valuations are of no account.

Anne

Anonymous said...

The economy is growing, and we are beyond a jobless recovery, though not beyond one in which wages and benefits lag in growth. Care must be taken by the Federal Reserve that the tightening sequence does not slow us and further interfere with job creation or wage increases.

Anne

Anonymous said...

The idea that we must fret about a possible financing problem for Social Security in 38 to 48 years is absurd, when there are immediate financing problems to deal with. But, the wish is to undo the concept of Social Security and the press is largely convinced Social Security is failing awfully so the public is increasingly convinced.

Anne

Anonymous said...

Reporters on Social Security on PBS and NPR take it as a matter of fact that the program is in crisis. Have we really done so poor a job teaching such reporters how research is to be done? Grumble, grumble.

Anne

Anonymous said...

Now, you and I are suspicious of hedge fund returns. Besides, hedge funds are no middle class retirement vehicle. What sort of investments are middle class retirement vehicles in what may be a low return environment. These last 5 years long term bond funds at Vanguard were superb. What now, we wonder?

Anne

Anonymous said...

This weekend I am reading Don Quixote, slowly slowly, for the Don was born 400 years ago and Edith Grossman has a wonderful new translation. Shall we tilt at windmills together? Now, here is a book. I shall take many days, for I do not wish to come to the end.

Anonymous said...

Anne, not Sancho Panza.

Anonymous said...

Knight of Doleful Countenance Gets Little Love at Home
By RENWICK McLEAN - New York Times

MADRID - This is the city where Miguel de Cervantes did most of his writing, published "Don Quixote," died and was buried.

But his tomb is closed to the public, his house no longer stands and the shop where "Don Quixote of La Mancha" was first printed is marked only by a plaque.

Cervantes, always an outsider in Spain during his lifetime, is in some ways still struggling to fit in here, even as the country plans big celebrations for the 400th anniversary of the publication of "Don Quixote" in January.

"Cervantes is the most important Spanish writer," said Antonio Muñoz Molina, director of the New York office of the Instituto Cervantes, which promotes the teaching and study of Spanish. "But he is not the most representative of the Spanish. His irony, his sense of humor - they are too subtle to seem Spanish." The gulf between Cervantes and his people may help explain his hollow presence in the streets of Madrid.

Anne

Anonymous said...

Steve,

Can you tell this Anne person to stop making off-topic posts? What does any of this have to do with Moore's Law?!?

MadJock said...

To summarise the article – CPUs are hitting a limit that’s slowing down individual processor speeds, and that the historical rapid increase in speed of CPUs to take care of poor coding is over. Further improvements will be aimed at hyperthreading, multi-core CPUs and more cache. Software developers need to consider this, and that it will likely be several years before the industry adjusts accordingly.

I agree with a fair amount of what he says, although I don’t think the ‘free lunch’ is over, just taking a pause. Intel are trying to sort themselves out after 5 years of terrible decisions (principally the pushing of MHz over performance in the P4, RAMBUS etc), and should be back on track sometime next year. As a consequence, AMD aren’t feeling the need to release ever faster processors so it’s ‘quiet’ just now. Most new features in PCs are in things like 64-bitness(pretty major in itself), wireless networking, faster/better memory architectures, storage, low power, and interoperability with consumer devices such as cameras and TVs. Hyperthreading, as implemented by Intel, is a dead-duck and will be made redundant by multi-core CPUs unless it’s implemented like in high-end systems (unlikely due to cost). More and more systems will be multi-CPU (Sony’s Cell architecture is based around this).

Something else that pushes multi-core is the packaging - you may shrink the CPU but you've got to get 100s of wires off the thing, and that sets a minimum footprint. Once you drop below that minimum footprint, you may as well use the space for something else, like a second CPU.

Moore's Law is holding - the number of transistors still may be doubling, but you're getting something else for it rather than just raw speed.

Blog Archive

Labels