Slashdot Mirror

HOWEVER, what gamers want is a decent priced (sub 200 dollar) mainstream i5 with SIX true cores.

6 isn't enough of a jump over 4...

For most home / personal computing (including high end video games) diminishing returns kick in hard past 4 cores. The problem is that in the few cases where tasks can be easily subdivided so as to utilize more than 4 cores, the cores will normally be stuck waiting for memory updates which continues to lag (speed / throughput wise) behind processor compute ability at an increasingly large gap which spans orders of magnitude. Of course the only known way to speed DRAM is to utilize more power, which goes against the general IT development trends (greener computing, more capable mobile).

The processor to memory speed gap is one of the reasons why Intel is investing in novel memory technology (phase change memory, etc.). The recent XPoint memory announcement hinted at potential future usage as "page swap" memory, replacing virtual memory management swapping pages out to disk (mechnical or solid state).

I haven't read all of Intel's releases this week, but one area I'm interested in is seeing how eDRAM (embedded DRAM) aka Crystal Well technology is going to end up being available and utilized across the Skylake line. In memory intensive benchmarks eDRAM has already shown considerably improvement in memory constrained benchmarks in Broadwell mobile processors, wheere it acts as an additional level of cache.

Give me 8 true cores and 16 threads, remove the IGP which I don't need for such a CPU...

Most people don't utilize more than 2 cores for more than 25-33% of the time, so the market for consumer-oriented many core processors just isn't there. People who really need the performance already just buy a Xeon.

Intel's "hyper-threading technology" is one of the biggest disappointments in many years, I wish they would let the branding and feature set die in obscurity like it deserves (IMHO).

Yes, yes, I know, Xeon and Haswell-E, but the reality is that the "need" for 8 core chips won't really happen until more of them hit the desktop market, and what AMD sells as 8 core doesn't count.

Well 8-16 core processors have been around for what, a bit less than a decade now? They won't really happen in the consumer / desktop market, because the market isn't demanding it (with purchasing dollars, not wishful thinking). Look at the very modest take-up of the Haswell-E X99 (LGA2011) 6 to 8 core processors released last year (August-Sept 2014 IIRC).

I love fast computers, personally I have a 6 core i7-5930K, and the performance difference for most home/consumer applications is so trivial that I don't notice a difference over using a 4 core i7-4790K except for in parallel benchmarks.

He's not modeling brains, he's modeling neocortical minicolumns. Theoretically that's not too difficult. I did a schematic for a subset of a minicolumn, a Hebbian cellular assembly. A collection of those in a minicolumn can process some significant chunks of information, given in/out infrastructure.

But, as per http://en.wikipedia.org/wiki/Cortical_column a minicolumn contains about 80 neurons, and there are 50 to 100 minicolumns in a (hyper)column. That's 4000 to 8000, cut it down the middle and say 6000. There are 2 million hypercolumns in a human cortex. According to Markram he's modeling things with a processor representing a neuron. To build a human cortex he'll need 12 billion processors. That's 1.2 million Big Blues. Good luck on getting IBM to fork those over.

But to model it right, the neurons have to be interconnected so that none are more than 6 hops away from any other, with an average of 3 hops. The Connection Machine http://en.wikipedia.org/wiki/Connection_machine would be a better choice to match the connectivity. A fully tricked out CM-1 had 65,536 processors, so the artificial human brain would need 183106 CM-1's with the machines interconnected in hypercubic fashion just as they're wired internally. That's a lot of iron. Worse, the CM-1 used single bit processors. The CM-5 used SPARC RISC processors, and we'll assume for brevity that they can do the job Markram wants done. A fully packed CM-5 could carry 16384 processors http://home.wlu.edu/~whaleyt/classes/parallel/topics/cm.html a quarter the number in a CM-1, so 4 times as many machines -- 732,421 hypercubic networked machines. That's about 21000 times more machines than were built, and those carried no more than 1024 processors. To get this artificial brain going Danny Hillis needs to get his soldering gun warmed up.

The CM-5 has a 900 m^2 footprint www-csag.ucsd.edu/individual/achien/cs433/papers/jpdc95.ps , so the requisite collection would cover 255 square miles plus access space and a massive amount of cable space for the interconnection.

In the absence of machine specs, I'll use the estimate of power consumption per compuational nose used in a Los Alamos paper comparing their options when they were shopping for some heavy iron. That estimate produces a probable power requirement for the CM-5 array modeling an entire human cortex at 12 to 24 gigawatts.

I don't disagree too strenuously that a design could be done in 10 years. But the collection of machines, even given advances in technology? Make it 100 to 200.

Then comes the final shot fired: why the hell would anyone cripple some perfectly good hardware by forcing it to act in such a capricious, error prone fashion as a human brain? Why not also buy an airplane and just taxi it down the streets rather than using a car? We already know that stem cells can be used to repair brain damage. Better to apply the appropriate technology to the problem.

Try this: http://www.cs.wlu.edu/~levy/software/pecon/ I wrote it and have run it on Linux and Mac OS X. It's a free, bare-bones alternative to something that Mathworks sells for a lot more money: http://www.mathworks.com/products/distriben/ Of course, if you want *real* high-performance, you won't run Matlab, but lots of people have tons of Matlab code and many idle processors, like the original poster.

In regards to the General Theory- As the creation of a single mind, it is undoubtedly the highest intellectual achievement of humanity. (Boorse, Motz, and Weaver; The Atomic Scientists)

http://alsos.wlu.edu/information.asp?id2=1857&past =2

Fermi I was a Liquid-Metal Fast-Breeder reactor (LMFBR) that had part of its core melt down when a cooling pipe was obstructed by a metal plate that broke off.

Since the cooling medium was liquid sodium, which "reacts" with air or water, any breach of the reactor would likely be followed by a fire or explosion.

And since it was a fast-breeder it contained weapons-grade plutonium. Any explosion could be followed by low-level nuclear detonation.

How likely this is is subject to debate, so don't take it as a sure thing.

And we make our containment building pretty tough so there's no guarantee that the reactor would be breeched either.

But since Detroit was nearby (downwind?) hence the book and the song...

Proponents of nuclear power will tell you that only two rods out of its 100 melted. What they don't say is that several rods warped, making them hard(er) to remove (there are no control rods in a LMFBR so you moderate it by removing fuel rods) and these remaining fuel rods were in danger of melting as well.

It was a pretty close call.

Keep in mind that Chernobyl "only" lost some of its rods, nor did it have a nuclear detonation either. In any case, you wouldn't want to live next door :)

http://www.nuke.hun.edu.tr/english/links/lmfbr.htm l

Ben Stein History

In 1973 and 1974, he was a speech writer and lawyer for Richard Nixon at The White House and then Gerald Ford. (He did NOT write the line, "I am not a crook.")

Computer science people have a name for everything. Some tasks are easily parallelizable, and others are not.

There is something called "Amdahl's Law" which is all about this topic. Basicly, you can predict the percent speedup based on how parallelizable the task itself is. This law, of course, assumes that the OS and multitasking is completely overhead-free, so it is sort of an upper bound on how fast something can go. But it is still interesting.

THIS is simply the first result form a google of "Amdahl Law".

If the UK and USSR had surrendered early, the USA would've still won by 1948. That would be when the atomic bombs fall on Berlin.

Or when the atomic bombs fell on Washington.

Why you will always be wrong. Basically, clusters make use of existing networking technologies (higher overhead) and supercomputers (SIMD, MIMD, whatever) are designed to make that overhead as low as possible.

Thus, by definition, for HPC applications that can't be parallelized enough to overcome the communications overhead, super computers will always beat out clusters and have a place.

The programmer and time are not fungible. We cannot simple expect to complete a project that takes 1 man 18 months with 18 men in 1 month. As you add more men the time improvements become less and less.

In other words, programmers tend to run afoul of Amdahl's Law. ;-)

Actually, Amdahl's Law would probably be a good way of calculating the maximum effective team size. Unfortunately, it can be very difficult to ascertain a value for the "work" needed on a project. Not to mention the "human factor" of programmers who are faster, less experienced programmers, and "cowboy coders" who refuse to check any of their work into version control.

Recently a "nuclear scientist" (the guy was a nut) was rather disparaging of Computer Sciences in an open forum. (Made the standard comments about "not a REAL science", blah, blah). He then went on to claim that in his younger days he had built the earliest computer cluster and that the software he wrote was so powerful that the cluster could have scaled up to infinity.

Suffice it to say, it felt really good to throw Amdahl's Law at him and call *bullshit*. He didn't talk to me after that. :-)

Not Computer Science, but the University Computing dept. Most will have recommendations for incoming students and information on connectivity issues. For example, my old school has such a page - http://computing.wlu.edu/students/guide/index.html . Also, you can often purchase computers through your school, with the major advantage being that the school provides tech support for free, and sometimes you can get a deal.
Other than that, I'd say no a palm and yes to a notebook.

I wonder if there were any Russians scientists on board the ISS who said, "I told you so, comrade, I told you so."

The Gallarus Oratory on the Dingle Peninsula in Ireland has stood for 1300 years, and it's constructed of stacked stones with no mortar. The method of stacking provides stability and keeps the inside dry.

It is ironic that water in space exhibits properties similiar to what was speculated to derive from the delusional Soviet discovery of polywater in the late 60's. See here and here.

I wonder if there were any Russians scientists on board the ISS who said, "I told you so, comrade, I told you so."

If you don't plan to handle anything much larger than about 30-50cm in size, I would suggest a clean glove box instead of a room.

I used glove boxes from time to time when I was studying chemisty. You can see a picture of one here.

The ones I was using, where generaly the size of a table, and about 3' high. You accessed your work inside using big rubber gloves, that where attached to the box so that the atmosphere in the box is seperate from the one outside. The front and some of the sides where glass. The glove box had one or more air-locks so you could take things in and out. (More than one lock, because smaller ones cycle quicker).

Chemistry glove boxes are usualy used for experements involving chemicals that are sensitive to oxygen or water vapour, so the atmophere in the box would generaly be dry nitrogen, at positive pressure, with no specal provision to avoid dust. Glove boxes are also used ocasionaly for radioactive compounds, where the box protects the laboratory atmosphere from the experenent, rather than the other way arround.

One of the chemistry departents I studied at had some glove boxes that they had made in-house, and I don't think it would be especaly difficult for you to make one. (They said theirs cost them about 10K to make, including labour).

The main body of the home made box, was made from acrylic pannels bolted together and sealed with silicone. The gloves are avalable from chemistry suppliers for about 10 pounds/pair. The airlock was made from welded sheet steel. The atmosphere in this box was maintained by flushing through with dry nitrogen (boiled off from liquid N2).

Obvously if you plan to use your glove box for electronics, there is no need for an intert atmosphere, but you will have to take precautions to avoid dust, presumably via an air filtration system, which I don't think would be to hard to design. Also, if you are serous about preventing bactera or suchlike from getting into it, I would include a UV lighting system to kill them, that you either switch on when you are not using the box, or filter out through the windows.

I hope this helps.

Personally, I find it interesting that this algorithm was developed by the same guy who wrote Alice's Adventures in Wonderland. A guy I teach with showed it to me a couple months ago, and I'm planning on using it in class soon to teach some programming concepts.

First they ignore you, then they laugh at you, then they fight you, then you win. -- Gandhi

Based on the description of the article, I looked up some things. What can I say? Somebody modded me down, so I'm at 49, and I'm incomplete without that karma point.

Amdahl's law

Amdahl's law

On chip multiprocessing

Simultaneous multithreading

The i860 was put out almost ten years ago and was originally intended for accelerating graphics applications. (At least that was what I was told. :-) It was also used in the Intel IPSC/3 (sometimes called the IPSC/860). Talk about old school! This is the stuff used in the old Paragons and Delta machines!

Does anybody out there know the relationship to the i860 and the chipset discussed in the article???