Wikipedia:Reference desk/Archives/Science/2015 January 16

Welcome to the Wikipedia Science Reference Desk Archives
The page you are currently viewing is an archive page. While you can leave answers for any questions shown below, please ask new questions on one of the current reference desk pages.

January 16

Bigger, better, faster, Moore?

I got interested in this with the earlier question about Moore's law. For a long time I've noticed CPU speeds have been about constant in the 2-3 GHz range. Why do they not get faster? Here's just one example of a not-too-cheap laptop that only has 2.4 GHz. If Moore's law still applies, why is it about the same as my 5 year old laptop (2.1GHz)? I know the law is about number of processors, and the article says that this is not linearly related to speed, but it does say that "There are cases where a roughly 45% increase in processor transistors has translated to roughly 10–20% increase in processing power." That should still mean a lot of speed, so more processors = faster and better, one would think. What's been happening, and is a 2.4GHz machine today that much better than a 2.1GHz machine from 5 or 6 years ago? I'm asking on the science desk because the other question was posted here, but I also feel it's a more general science issue, and not just for techies (it is after all a consumer-type question). IBE (talk) 05:39, 16 January 2015 (UTC)

http://www.tomshardware.com/forum/336310-28-processor-speeds-increasing Greglocock (talk) 06:05, 16 January 2015 (UTC)

Moore's law is about transistor density. CPU speed is peaked for level-0 SRAM single cycle reads. Up until 8-10 years ago (90 nm), Vt (MOS threshold voltage) and operating voltage was scaling with process. Leakage, cell stability, transistor drive, and the bitline development rate all hit a tradeoff wall at about 1 volt and 2 GHz. Normally, when we used to compare frequency, architecture and pipelining were the drivers. When it became SRAM, all single cycle reads were limited by the same cell so that's why ARM and x86 and all the various other architectures seem to peak at the same frequency. The performance gained is that the SRAM still shrinks so SRAM size increases but it isn't getting much faster. Each technology node has Vt/leakage tradeoff but the current consumed is exponential. This causes power delivery and heat issues that are not as good a tradeoff as more SRAM or processors. --DHeyward (talk) 06:33, 16 January 2015 (UTC)

Heat is an issue, but it's less problematic than you might think. Heat is removable with sufficiently-giant heat sinks or liquid nitrogen coolant; if it were the real problem, you'd see at least certain performance-crazed segments of the market running 10 GHz CPUs with cryo-coolers attached. In reality, the limiting factors for the last decade or so has been signal integrity, not thermal load. If you want to really know why we don't build 5 GHz VLSI circuits with today's technology, here are two books you should read:

• Planar Microwave Engineering: a practical guide to theory, measurements, and circuits. CPUs that run digital logic at 3 GHz are really operating their analog parts in the microwave regime. You need to know how the analog electronics actually behave when you build them on silicon with modern processes. At 3 GHz, with the parasitic capacitances that are inherent to real transistors that we can actually build, square-waves look pretty not-square, and ones look an awful lot like zeros. Things get worse when signals have to cross clock domains, or worse yet, leave the substrate across a wire bond.
• Computer Architecture: A Quantitative Approach. No punches pulled, this book runs the numbers on practical and theoretical computer architectures, so that you can understand whether the performance limitations are due to pipeline depth, cache strategy, data hazards, and so on.

Nimur (talk) 07:19, 16 January 2015 (UTC)

No, heat, power density and IR drop really becomes intractable (and a poorly scaled) problem at the die level. 1 volt @ 130 Watts is 130 amps. Put that into a square centimeter chip with IR drop and inductance and power delivery (and heat removal) become huge, expensive issues. These high-performance CPU's are all flip-chip and die size has to be large when power is high so they have enough power delivery bumps. Wire-bond is out of the question due to inductance and the di/dt requirements. Because frequency is a function of voltage and power is V^2, the resulting increase in power with the increased frequency is effectively a V^3 scaling. So linear increase in frequency comes with cubic increase in power. --DHeyward (talk) 09:19, 16 January 2015 (UTC)

The Core i7-4700MQ CPU in that laptop actually runs at up to 3.4 GHz, temperature permitting. It also has 4 cores supporting 8 logical threads, while your older laptop most likely has 1 or 2 cores supporting 1 or 2 threads. Each core has a sustained maximum throughput of 4 micro-ops per cycle versus earlier generations' 3, and the execution units support 256-bit SIMD registers (AVX) versus your laptop's 128-bit SIMD (SSE). There's probably substantially more on-die cache RAM, and Intel has made various incremental improvements to other aspects of the internal architecture. Putting all of that together, it could easily be twice as fast as your laptop or more on realistic computing tasks. That's not very impressive for 5 years by historical standards, but CPU performance hasn't completely flatlined. (Source for some of the above: Agner Fog's microarchitecture manual.) -- BenRG (talk) 09:44, 16 January 2015 (UTC)

should you mix water back into sour cream, or drain it?

If you open a package of sour cream and there's some water inside, should you drain it (since it's water) or mix it together (on the theory that it was supposed to all be together, that's how much water should have been inside anyway)? — Preceding unsigned comment added by 212.96.61.236 (talk) 07:38, 16 January 2015 (UTC)

I usually drain it, but it's just water. Won't hurt or help much. Can always replace it later if you want, but it'll go rancid before it ever dries out in a fridge. Emulsifiers make sure of that. InedibleHulk (talk) 08:08, 16 January 2015 (UTC)

First, check the expiration date. ←Baseball Bugs ^{What's up, Doc?} carrots→ 08:19, 16 January 2015 (UTC)

I just realized my sour cream company also sells the ingredients. Yours might, too. I'd never thought of sour cream as something to repair instead of replace, but if you're feeling frugal, you might be able to save the mixture, well beyond the best before date. InedibleHulk (talk) 08:30, 16 January 2015 (UTC)

My sour cream company also doesn't believe in real milk. InedibleHulk (talk) 08:38, 16 January 2015 (UTC)

Under that first link under "Flavour" it says, hilariously, "Bland flavour with a hint of dairy." 212.96.61.236 (talk) 08:45, 16 January 2015 (UTC)

The honesty's refreshing. I can't count how many times corn-based ingredient-based products have implicitly promised to blow my face off. According to a guy on the Internet, the missing "amazing explosion" ingredient (and they aren't missing many) is plain old sour cream. Keep it real, Doritos. InedibleHulk (talk) 03:52, 17 January 2015 (UTC)

Watch for the new nitrogen triiodide flavor.  :) (Though of course it is more practical for use in a jawbreaker) Wnt (talk) 19:31, 17 January 2015 (UTC)

Mmmm...legitimately destructive. Those would also give a nice zing to Pop Rocks. InedibleHulk (talk) 22:18, 17 January 2015 (UTC)

Subterranean rivers

Do subterranean rivers generally end up discharging into the sea like most other rivers? I always assumed so but the article doesn't say. If so, do we know where any of these "mouths" are?--Shantavira|^{feed me} 12:37, 16 January 2015 (UTC)

The article names many of the mouths. The water has to go somewhere, after all. However, in some cases like the Mojave River that destination may be an inland delta on a salty lake, or even just to dry up as an intermittent river. Wnt (talk) 15:34, 16 January 2015 (UTC)

Ending at an Endorheic_basin would be another option. SemanticMantis (talk) 15:41, 16 January 2015 (UTC)

Hmmm...this is a very good question!

Surely it can't be just like above-ground rivers draining into above-ground seas because in that case, evaporation from the oceans forms clouds which causes the water to rain on hilltops and mountains to keep the rivers flowing and to stop the seas from eventually overflowing.

Below-ground, what is the mechanism to lift the water from subterranean lakes/seas back up to the source of the water for the subterranean rivers? I presume that some water makes it back to the surface from geysers and such - and water flows into the subterranean rivers from the surface to keep them flowing...but it's hard to believe that enough water is lifted to the surface to keep the levels of large underground aquifers from just filling up and causing the underground rivers to stop flowing.

Obviously, for shallow underground water, it can arrive back onto the surface from a spring - but that's not going to work for rivers that are further below ground than the lowest point in the local topography.

SteveBaker (talk) 16:09, 16 January 2015 (UTC)

It all gets cycled (eventually) of course, e.g. subsurface flow, groundwater flow, water cycle, etc. If you want to know how long water stays in a given system, you can look up estimates for Baseflow_residence_time. SemanticMantis (talk) 16:31, 16 January 2015 (UTC)

I can't say for all caves, but for one of the major ones, Mammoth Cave National Park, whose profile looks like this,[1] its underground river flows downhill at a shallow angle and then opens out into the Green River, a tributary of the Ohio. ←Baseball Bugs ^{What's up, Doc?} carrots→ 17:22, 16 January 2015 (UTC)

Some of those underground streams (fountains of the deep) come up under the sea and are known as vruljas, eg at Bay of Mali Ston, Ngaruroro River, Chekka Wonky hole and Pisak. Graeme Bartlett (talk) 02:10, 17 January 2015 (UTC)

Thanks. "Vrulja"! We like that word! And wonky holes, which could only be Australian! Thanks!--Shantavira|^{feed me} 08:50, 17 January 2015 (UTC)

Raspberry Ketone B.Half life

How could I know the Biological Half life of Raspberry ketones?, I wouldn't mind import it from a source you guys consider reliable, to the article. Thx, Ben-Natan (talk) 16:29, 16 January 2015 (UTC)

A quick search of raspberry ketone urine on PubMed turns up PMID 7113261, which compares its metabolism in rats, guinea pigs and rabbits, saying that 90% is excreted within 24 hours and listing metabolites (sounds like the usual: molecule meets oxygen, molecule loses). But it's in a really obscure journal, so a PITA to look up. You could take the log₂10 and suppose that the half-life is 24/3.32 = 7.2 hours, but that's a fraught assumption since there could be many sequential steps the molecule goes through that impose various delays so that it doesn't pee out according to a log table. (I suppose even saying "half-life" already wades into these issues, come to think of it) In any case it would be OR; stick to what the paper says. :) Wnt (talk) 13:34, 17 January 2015 (UTC)

Kg to lunar lb

How much does 22kg weigh on the moon? Apollo_11#Lunar_ascent_and_return says that the astronauts lifted two sample boxes containing more than 22 kilograms (49 lb) of lunar surface material, but because of the moon's lesser gravity, 22kg weighs a lot less than 49lb. 65.210.65.16 (talk) 19:25, 16 January 2015 (UTC)

The Moon's gravity is 0.1654 of Earth's, so a mass of 49lb would "weigh" 49 times 0.1654 on the Moon. Note that the mass remains unchanged, so extra care has to be taken when the mass is moving because it has the same momentum as on Earth. Dbfirs 19:48, 16 January 2015 (UTC)

(ec) Both kg and lb are used here as units of mass. Although the Pound (mass) can be used as a unit of weight, this is also true of the kilogram. See Mass versus weight. Weight is not a very useful quantity when talking about the contents of a sample. However, if you used a spring-based or electronic scale, 22 kg on Earth weighs 0.1654 * 22 = 3.64 kg on the moon, which is equal to 8.02 lb. - Lindert (talk) 19:49, 16 January 2015 (UTC)

Actually Lb is always weight but in standard gravity, it's a constant relation to mass. The term for mass is Slug --DHeyward (talk) 21:22, 16 January 2015 (UTC)

That's not what I was taught! A slug is a rarely-used unit of mass, along with the Pound (mass). Dbfirs 21:27, 16 January 2015 (UTC)

Whether one is taught about pounds as a unit of weight depends on where one studies and what discipline. As a physics student in Canada, we were taught that pounds were strictly a unit of weight, and slugs are the appropriate "imperial" unit of mass. American engineering students, on the other hand, seem to be taught how to work with pounds as a unit of both mass and weight. I have run across kg as a unit of force in the U.S., which to me is an abomination.--Srleffler (talk) 03:25, 18 January 2015 (UTC)

In the UK in the 60s/70s, we were taught that a pound is a unit of mass, and its associated force is a poundal.--Phil Holmes (talk) 11:43, 18 January 2015 (UTC)

Yes, I remember meeting the slug as an amusing aside in the 1960s, but the poundal was much more common as a unit of force in the Foot–pound–second system (where the pound has to be mass). It seems that Canada used a different system which our Template:GravEngAbs calls a "British Gravitational System", but the link is circular and the text says that it is used by American engineers. I'm baffled! Dbfirs 13:32, 18 January 2015 (UTC)

The kilogram is a unit of mass. A mass of 22 kg will still have a mass of 22 kg if it is relocated from the Earth to the Moon. The force of gravity, at the surface of the Earth, is the mass times the acceleration due to gravity, 22 × 9.807 ≈ 215.8 newtons. On the moon the force on the object due to gravity is 22 × 1.622 ≈ 35.68 newtons.

It would be foolish to further contaminate the Moon with US customary units so I will not do so. Jc3s5h (talk) 21:51, 16 January 2015 (UTC)

If you're suggesting that the metric system is the realm of lunacy, I won't argue. Now, here's a poser: How many newtons does one fig newton weigh? ←Baseball Bugs ^{What's up, Doc?} carrots→ 22:12, 16 January 2015 (UTC)

If ever I visit the Moon, I shall make a point of taking some Imperial units with me, especially a 100-year-old pound mass that I happen to own. Fortunately for Jc3s5h, I'm not likely to make that journey. Dbfirs 22:22, 16 January 2015 (UTC)

What software would you use to model something spreading in water (river, lake, sea)?

What software would you use to model something spreading in water (river, lake, sea)? — Preceding unsigned comment added by 31.4.152.13 (talk) 23:45, 16 January 2015 (UTC)

The look of it? Maybe Blender. The actual physics? Maybe Simulink. 75.75.42.89 (talk) 00:38, 17 January 2015 (UTC)

How about reading Schlumberger's Water Services software overview? They sell the world's best aquifer management, monitoring, and simulation software. Nimur (talk) 00:50, 17 January 2015 (UTC)

Thanks you both. I'd take a look at both options for modeling the physical process. My interest is not to make a visual model with tools like Blender, but to understand what will happen next (after an oil spill or other type of contamination of a water resource.--31.4.153.226 (talk) 01:16, 17 January 2015 (UTC)

Here's a full book available at no cost from the Bureau of Ocean Energy Management: Review of the State-Of-The-Art on Modeling Interactions Between Spilled Oil and Shorelines for the Development of Algorithms for Oil Spill Risk Analysis Modeling. Fascinating topic! The software involved in fluid flow simulations pushes the state-of-the-art in scientific computing, so this is not necessarily going to be an easy topic for a hobbyist; the softwares that exist are frequently custom-made, and aimed at a small community of highly-skilled expert users. Nimur (talk) 03:25, 18 January 2015 (UTC)

Note that if you're talking about something biological, like an algal bloom, then quite different software would be needed. StuRat (talk) 06:27, 17 January 2015 (UTC)

It really depends on what your specific goals are. The mathematical analysis will basically be done with a type of Convection–diffusion_equation. The commercial products will still have that at the core. Here is some code and instructions that will let you model various simple cases in MATLAB - [2]. SemanticMantis (talk) 17:27, 18 January 2015 (UTC)