Follow Slashdot stories on Twitter

 



Forgot your password?
typodupeerror

Slashdot videos: Now with more Slashdot!

  • View

  • Discuss

  • Share

We've improved Slashdot's video section; now you can view our video interviews, product close-ups and site visits with all the usual Slashdot options to comment, share, etc. No more walled garden! It's a work in progress -- we hope you'll check it out (Learn more about the recent updates).

×

Comment: Re:Is that really a lot? (Score 1) 259

by ChrisMaple (#49143145) Attached to: Drones Cost $28,000 Per Arrest, On Average

Saying "reform" the immigration system is meaningless, It sounds good to people who don't think, yet it could mean anything from "encourage freeloaders to come here" (which is Obama's official but hidden policy) to "shoot everyone who sneaks across the border."

Sneaking into the country is not a "minor infraction", it's a de facto invasion by an ununiformed enemy.

Comment: Re:Is that really a lot? (Score 1) 259

by ChrisMaple (#49143081) Attached to: Drones Cost $28,000 Per Arrest, On Average

Even if most immigrants are good, that does not mean that on balance the effect of illegal immigration is good. Its very easy for a single person to destroy more than 10 people can create in a lifetime, and that's just the sort of thing that a jihadist who sneaks through our porous borders wants.

Your personal experience is almost meaningless. Do you expect an enemy to tell you he's out to destroy the country?

Comment: Re:Is that really a lot? (Score 1) 259

by ChrisMaple (#49143039) Attached to: Drones Cost $28,000 Per Arrest, On Average

Would you mind defining "social change"? Do you mean promoting the activities of murderous rioters? Providing free money and drugs for people who refuse to work? Free abortions for whores? Jailing CEOs when the Sarbanes-Oxley forms don't give correct results to the penny?

The word "social" at the beginning of any phrase means there's something bad being hidden.

Comment: Consider the Alternative (Score 1) 259

by ChrisMaple (#49142983) Attached to: Drones Cost $28,000 Per Arrest, On Average
When considering the cost of finding and deporting illegal aliens, it must be compared against the cost of failing to find and deport them. Some aspects of illegal aliens are: drunken unlicensed illegal drivers killing pedestrians (no, licensing them does not make it OK), new outbreaks of measles, mumps, and tuberculosis, and the World Trade Center. Still think it's too expensive to eliminate illegal aliens?

Comment: Re:That's computers 10x faster than today (Score 1) 279

by ChrisMaple (#49137839) Attached to: Intel Moving Forward With 10nm, Will Switch Away From Silicon For 7nm

Sorry, it doesn't work that way. We're already well past the point where we can ignore channel leakage considerations, and their interaction with transistor thresholds, supply voltage, and other things. Gate leakage is becoming a problem. Power supply conductors can't be scaled due to migration, they now commonly take up a whole layer. At a guess, I'd say the asymptote for big silicon CPU clock rates is 10 GHz, more than a decade away.

Much of the speedup in the last 2 decades has come from SIMD and multithreading. Multithreading still isn't heavily implemented, so there's a big gain to be obtained there, and the hardware to take advantage of it is more cores, which scaling obviously helps.

More memory on-chip is good, but numerous tests have shown that we're already well into the area of diminishing returns for most applications.

Comment: Re:Leakage (Score 1) 279

by ChrisMaple (#49137539) Attached to: Intel Moving Forward With 10nm, Will Switch Away From Silicon For 7nm
In high speed silicon ICs, SiO2 is no longer "a great gate insulator" because the dielectric constant is too low. Channel resistance is roughly inversely proportional to the dielectric constant of the gate material. This was a big deal a decade ago, but now "high K" material use is routine and hence seldom mentioned.

Comment: Re:InGaAs? (Score 1) 279

by ChrisMaple (#49137457) Attached to: Intel Moving Forward With 10nm, Will Switch Away From Silicon For 7nm
Modern high speed CPUs are dynamic nMOS silicon, with P channel devices sprinkled around for recharging nodes, etc., but not in the signal path. P channel MOSFETS are about 3 times weaker than N channel in silicon, resulting in CMOS being 1/4 the speed of dynamic nMOS. There's a reason that Intel specifies a minimum clock rate on their CPUs, dynamic circuits lose their charge and malfunction below the minimum clock rate..

Comment: Re:amazing (Score 1) 279

by ChrisMaple (#49137153) Attached to: Intel Moving Forward With 10nm, Will Switch Away From Silicon For 7nm

Actual sustained firing rates don't exceed 10/sec. At 100/sec, a neuron risks poisoning itself with waste products.

Your "100 billion independent processing units" are single bit units, in no way equivalent to a CPU, or even an integer math unit.

Just those 2 considerations means reducing your estimate by a factor of 300, from 10 teraflops/s to 300 gigaflops/s. (not far from the power-limited consideration I made a few paragraphs above, 156 gigaflops/s). Power7 with 8 cores can peak close to 800 gigaflops/s, although while consuming much more power.

Comment: Re:amazing (Score 1) 279

by ChrisMaple (#49136795) Attached to: Intel Moving Forward With 10nm, Will Switch Away From Silicon For 7nm

Energy use puts one cap on brain processing. A single neuron firing requires at the very least 4x10^-12 Joules. http://www.nature.com/jcbfm/journal/v21/n10/full/9591146a.html At a brain power of 20 watts, that's 5 trillion firings per second, each firing equivalent to the state-change of a single flipflop feeding 1000 gates. That's ignoring standby/idle power dissipation.

A single firing doesn't mean much. A floating point number is 32 bits, requiring 32 flipflops (neurons). 5e12 / 32 = 156e9 flops/s, i.e. 156 Gflop/s. A very impressive number, if the brain were actually optimized to do floating point math, and didn't have to do anything else. But that's nowhere near 100 trillion.

Comment: Re:amazing (Score 1) 279

by ChrisMaple (#49136287) Attached to: Intel Moving Forward With 10nm, Will Switch Away From Silicon For 7nm

There are obvious fallacies in the brain calculation. One is that nowhere near all neurons are active at once: the brain has specialized components most of which are inactive at any given moment, and memory is a big, low access rate component. Another is that "100 billion neurons...each sending signals to 1000 other neurons" is implying that there's something meaningful happening in 100 trillion places at once, which is just laughable.

Let's say, for instance, that you can recognize 10,000 songs, and mostly follow along with the tune and think of the words before they're sung. When you're listening to one song, you might relate to snippets of another 10 songs. Your brain inactivity index for songs is thus 99.9%, and for many other activities at the same time it will be even closer to 100%. All of the time, most of the brain is inactive.

Comment: Re:Guy is a moron. (Score 0) 126

by ChrisMaple (#49097919) Attached to: Fedcoin Rising?

taxing commerce gives the government incentive to encourage commerce.

This is at best a half-truth, and makes the silly assumption that the government would act rationally. Cigarette taxes are very high to discourage the use of cigarettes (and to rip off consumers.) Greenies promote high gasoline taxes to discourage diving gas-powered cars. Imported goods are taxed to discourage foreign goods (and encourage untaxed domestic producers.)

It's a Laffer Curve phenomenon, not that there's much recognition of that when such taxes are being proposed.

You can not win the game, and you are not allowed to stop playing. -- The Third Law Of Thermodynamics

Working...