microarchitect

Maybe it is just me, but overstating anyone's impact to this extent is disrespectful to that person.

I don't claim that the article is not overstating Ramanujam's impact, but you seemed to be arguing that there's no way this could be true, so the article must be overstating his impact. Given that even you don't claim to have the background required to objectively evaluate Ramanujam's impact, this seemed to be a prejudiced comment.

In retrospect, I might have been a bit trigger-happy with the downvote. Sorry about that.

[1] "The meaning of Ramanujam now and for the future", George E. Andrews. http://www.math.psu.edu/andrews/pdf/274.pdf

[2] "The Ramanujam Journal", http://www.springer.com/mathematics/numbers/journal/11139

Quote from the conclusion of [1]:

In his Presidential address to the London Mathematical Society in 1936, G. N. Watson spoke movingly of his emotional response to Ramanujan's achievements. I close by quoting his last few paragraphs [54; p. 80]:

"The study of Ramanujan's work and of the problems to which it gives rise inevitably recalls to mind Lame's remark that, when reading Hermite's papers on modular functions, \on a la chair de poule." I would express my own attitude with more prolixity by saying that such a formula as:

<snipped formula>

gives me a thrill which is indistinguishable from the thrill which I feel when I enter the Sagrestia Nuova of the Capelle Medicee and see before me the austere beauty of the four statues representing Day, Night, Evening, and Dawn which Michelangelo has set over the tombs of Guiliano de Medici and Lorenzo de Medici. Ramanujan's discovery of the mock theta functions makes it obvious that his skill and ingenuity did not desert him at the oncoming of his untimely end. As much as any of his earlier work, the mock theta functions are an achievement sucient to cause his name to be held in lasting remembrance"

To me, it's interesting that a community where fairly easy to build web-applications are called transformative and disruptive also has two out of thirteen comments critical of Ramanujam's impact.

It takes all kinds to make up a society. Some of us want a balanced life; we want to be reasonably good at everything. Some others want to be the best at a certain thing and are willing to sacrifice almost everything else to achieve that goal. To a very crude approximation, those from the former group keep society going while those from the latter push human society forward.

For me, Steve was an interesting curiosity. Not an example to emulate.

On the face of it, there's nothing wrong with being either type, but it does seem like each type cannot fathom why the other behaves the way it does.

I wouldn't be so quick to judge. If you're a woman in India and a strange man asks you to try something weird that could potentially harm your body, I think you have every right to be suspicious of that person.

this fatalistic selfishness isn't a gender based issue. It is, however, one of the corrosive elements of Indian culture that worries me

I agree that Indians are incredibly selfish in certain ways, especially when you look at their actions through the lens of western cultural practices, but this is probably not a good example of that.

The first-order model I was taught in school was that leakage current not considering DIBL used to be proportional to exp(vgs-vt) but thanks to DIBL leakage is now proportional to exp(vgs - vt + eta*vds). The eta here is the DIBL coefficient, which AFAIK is 0.1 or thereabouts.

The only way I see this having a significant effect is through the vt vs. speed trade-off I mentioned above. Lower vt's exponentially increase leakage power, so we push the vt up to combat leakage, but this reduces transistor speed because frequency is roughly proportional to the on-current which is roughly proportional to gate overdrive vdd - vt.

A quick calculation I did with a 90nm model for I have some parameters with vt=0.3, vdd=1.2, subthreshold slope=90mV/decade and dibl coeff=0.1 seems to suggest that leakage would still be < 1/100th of the on current.

The real issue which the OP may or may have been trying to get at is that leakage power eats into the chip's power budget. Since we'd rather not burn our budget on leakage, we reduce leakage by increasing the threshold voltage. But unfortunately, this comes at the cost of frequency.

The other very important issue is that of power-density. There's a famous graph that Shekar Borkar of Intel [1] made showing that if we continued to ignore power dissipation issues like in the past our chips would run ridiculously hot (IOW, they wouldn't work at all because they'd just burn themselves to death).

There are also other issues like the fact that your wires start acting like antennas at 5+ GHz and that reliability concerns like electromigration and dielectric breakdown are getting worse with newer technologies.

A much more recent and reliable (not to mention highly-cited) reference on scaling and power issues is [2].

[1] I found a copy here: http://www.nanowerk.com/spotlight/id1762_1.jpg [2] http://www-vlsi.stanford.edu/papers/mh_iedm_05.pdf

I don't doubt your claims about the amount of the work the two courses at MIT require. However, you're forgetting that for some of those 140 hours you're being lectured by the top professors in the area. You're unlikely to have a question that they can't answer. Other parts of those 140 hours are spent talking with some of the smartest students on the planet. Some more of those 140 hours are spent solving problems carefully designed to maximize understanding. If you can't figure something out, you have fellow students, TAs and professors to harass. Somebody studying alone at home doesn't have this environment.

And then you need to take into account the fact that if you're undergrad at MIT you're already smarter than 95% or so of the population. Basically, the takeaway here is that your 70/140 hour number applies to something like 0.1% of the population under the assumption that they're studying CLRS at home.

You need to see this in context. This is advice for CS undergrads looking for jobs. This advice is being given out on reddit. Anybody who needs this advice (and CS undergrads from MIT certainly don't) will not be able to work through CLRS in a week.

The lower bound for men is zero.

Maybe it is zero for men who expect their wives to do all the parenting. It's certainly not zero for everyone else.

There's a difference between these three statements:

1. Women and men are different - obviously true and too vague to be meaningful.

2. Women have breasts while men do not - obviously true and mostly (and hopefully) irrelevant to this discussion.

3. Women work fewer hours than men - A specific claim about women that may be used by some to justify their prejudices against women. I don't doubt that you made the claim without malice, but unfortunately these are the kinds of remarks that reinforce sexist stereotypes. For this reason, I wish people wouldn't make such claims unless there is watertight evidence backing these claims.

Even if this were true, it doesn't mean that men get more stuff done. I' be interested in a study that measures the difference in job performance for equal compensation levels between men and women. Even that study might be biased because there's some evidence that if a woman and a man perform at exactly the same level, men get better ratings, but it would be better than simply measuring hours "worked".

I also think throwing in statements that disparage women with the caveat that "I don't have the facts to back this up" is irresponsible and contributes to the sexism problem.

Anyway, to me the arguments here about babies and startups are eerily reminiscent of the arguments against women becoming professors. Stuff like: women want to have babies, how will they advise their students when they're pregnant? They can't bring in grant money for a whole two years! Blah blah blah. Think of our academic standards.

Fortunately that argument has been thoroughly debunked by the large number of successful female professors who are also mothers. Frankly, I don't buy the argument that having a pregnant co-founder hurts a startup. If women can get tenure at a top-10 university while raising a kid, I'm pretty sure they can also keep your startup going while raising a kid.

EDIT: Heh, I've been downvoted. Why am I not surprised?

I'm extremely skeptical about your claim. Just reading through the chapters will take the better part of a day. Working through the exercises and thinking about what each of these algorithms will take much longer.

I doubt you can pick up a solid understanding of just the first 15 or so chapters of CLRS in a few months, let alone days.

If the authors actually tried to measure Linpack scores on their cluster, they'd find that peak performance would be much worse than the 12 teraflops they're claiming as "peak" performance. This link [2] seems to indicate that one S2050 along with a fairly fast CPU gives you about 0.7 teraflops. Even if they get perfect scaling across GPUs, which they won't, and if they could solve all the heating issues with their design, they won't get more than about 0.7*6.5 = 4.55 TFLOPS peak on linpack.

In any case, many (39 out of 500 to be precise) of the most powerful supercomputers in the world also use GPUs to accelerate certain applications so there's simply no way this thing would be faster than those computers.

[1] http://www.top500.org/lists/2011/11/press-release [2] http://hpl-calculator.sourceforge.net/Howto-HPL-GPU.pdf

I think this is rather ridiculous. We don't decide what goes into scholarly journals based on a public poll, I don't see why we should do that for an encyclopedia.

EDIT: If one accepts the premise that even wikipedia requires curation and cannot simply include all submitted articles, then it follows that one of the tasks of the editors is making value-judgments about what knowledge is interesting and important to wikipedia's readers. I assume the guidelines about notability, conflict-of-interest and whatnot are attempting to set forth policies that assist in making these value-judgments.

You appear to be arguing that the only basis for making this value-judgement is recognizability to a layperson. I'm having a hard time understanding why this is a suitable metric.

I doubt that having lots of information about porn stars will benefit wikipedia's readers. It might entertain them, but I don't think it has any scholarly benefits. At the same time, I don't see how removing information about scientists benefits wikipedia's readers. I can see that in some extreme cases having lots of articles about little-known scientists might clutter wikipedia. However, I don't see how holding articles about published scientists to a higher standard than articles about pornstars benefits either wikipedia or its readers.

Sorry, but that wasn't much of an apology for a racist slur.

First a bit of background. I come from India where they practice reservation which I think would be considered much "worse" than affirmative action over here. The way the system works is that a certain percentage of seats are reserved for people from certain (usually lower castes). So for my engineering examination, I wrote a common exam conducted by the government along with about 80,000 other students from my state. Among the 80,000 about 35k or so did well enough to be given a "rank". The rank means exactly what you think it means.

Now, students are allowed to choose the college they want to go to in the order of their ranks. However, reservation means that people like me, who don't "have a reservation" have to choose seats from the general quota. The general quota was, believe it or not, 20% of all seats. Out of my class of 140 EE students, we had 29 students who came through the general quota. About 15 more seats were available to students who were willing to do without the government subsidy.

Side note: Because my education was subsidized by the state government, I paid about $300 dollars a year in fees. The ones who went without the subsidy paid about $1200 dollars a year. All

All the remaining 100 or so seats were reserved for people from one or another of the reserved castes. For people in the general quota, the lowest (highest numerically) rank admitted to my department was 114. For a person coming from a reserved quota, the lowest rank admitted was 29xx.

At the time I felt this was grossly unfair. I busted my ass to be ranked in the top 100 among 80,000 candidates to get into this place and these guys were admitting students who weren't even in the top 2000 simply because they were born into the right family.

Over time, I've come to realize that I enjoyed a great many privileges these other kids did not. I had sporadic access to the internet, I had parents who were reasonably supportive of my studies, I knew English really well. I didn't get the 90th rank simply because I was awesome. I got it because I was lucky enough to be born with (some) privilege in a third-world country.

Many of my classmates from disadvantaged backgrounds have gone on to become competent engineers. I realize now that they fully deserved their chance.

[1] http://www.eetimes.com/ContentEETimes/Images/Design/Prog%20L....

Threshold voltage is not really an effective knob

Why is it not an effective knob? Most modern designs include sleep transistors in an attempt to not leak when a circuit is inactive. These would not work unless we could engineer high-vt transistors.

Maybe you're referring to some papers that used to come out a few years ago which suggested that leakage power will dominate total power. As I said above, this is unlikely to happen. It doesn't make sense to operate at a combination of supply voltage (Vdd) and threshold voltage (Vt) where leakage dominates total power. I think these papers misunderstood the fact that threshold voltage and hence leakage itself is a knob that the device manufacturing folks can control.

Active power is the one that's related to the frequency (P ~= CV^2f). Leakage power will "leak" even if the transistor is not switching.

If you're implying that leakage power doesn't affect frequency, you are wrong. Transistor speed depends on the gate overdrive which, for modern velocity-saturated devices is proportional to Vdd-Vt. Leakage power itself is proportional to exp(-Vt). There is a clear trade-off here between how fast you run your chip and how much it will leak.

This is kinda nitpicking, but I'm not sure leakage will ever dominate active power. We still have the ability to reduce leakage if we want, we just have to give up frequency for it. In the past we didn't have to play this trade-off but even now I don't think it ever makes sense to run your chip so fast that leakage is more than dynamic power.

I do agree that for any given node, Intel is still going to ahead of the rest. It'll be interesting to see how much this helps them.

The other big technical issue is the end of Dennard scaling. For most of the last three decades, scaling CMOS processes bought you three things: more transistors, higher frequency and lower power. Things are different now. We can't really scale frequency any more because we've run into the power wall. We used to get lower power at the same frequency by scaling the supply voltage, but this also required us to scale the threshold voltage (a device parameter). Unfortunately we can't scale the threshold voltage willy-nilly like in the past because leakage power increases for lower threshold voltages and is now a significant contributor to total power. We still get more transistors per unit area, but it's not clear whether the economic costs of building up new fabs and switching to a new process are offset by the benefits of having more transistors to play with.

The bottomline is that it's not clear whether Intel's biggest competitive advantage, that of having a manufacturing process superior to everyone else, is still that much of an advantage.

PS. One thing I find truly amazing is that Dennard predicted that we'd run into all these problems back in his landmark paper in 1974!

What does it mean to say that a government "governs the least"? Does it make the fewest number of laws? Does employ the fewest number of people? Does it spend the least money? All of the above? Some of the above? Why are these good?

If any country can do it, it's the US. The universities attract the smartest students from all over the world, and have the best faculty in the world by a large margin. The government simply needs to figure out how to harness all the brainpower to solve real-world problems.