I think we both agree that you need to adjust your code if you are going from single core code to multicore (and IME, it is only one core or multiple cores, it is usually easy to write the code so that it adapts to any number of cores available, as long as model is symmetric).
I disagree. It's not important to the majority of programmers who reside in the space of application development or web development. I'd estimate that 99 out of a 100 programmers reside in this space. THere is nothing in the higher level programming paradigms that gives a flying rats' ass about the number of cores. Obviously, this is not the case for OS architects at say Microsoft or the open source community contributing to various flavors of Linux/Unix, but it is true for most programmers.
I even can agree with this - but in that case, you can consider the problem resolved - apache or IIS together with OS will take care about running HTTP requests in parellel. I really do not understand why do you complain about increasing number of cores as some sort of "cheating".
(And here, let me repeat that increasing number of cores is just one way performance is improving, other being increasing IPC of individual cores and yet other one increasing clock speed. I still feel like you are mixing IPC performance with number of cores).
As for the other discussions, what do you feel are the major differences between a single core processor and a single core of a dual core processor without mentioning caching?
Mostly none - within the same design and generation. And you can even mention caching. OK, maybe this is tricky question, there are circuits for sharing memory access, for coordinating memory stores, locking the bus etc which complicate somewhat the OOO core logic, but nothing serious.
Anyway, Celereon G440 (single core Sandy Bridge) will still run 3 times faster than equally clocked P4 running the same code (at binary level).
I'm not saying its cheating, I'm saying its finite, and we're reaching that finite point in those 28 years (you previously used 2040). There is a really interesting article over at Arstechnica: http://arstechnica.com/science/news/201 ... e-atom.ars
It talks about the fact that they've reached the absolute smallest transister, an atom, which is .03NM wide. Current projections have us entering the 11NM size range, down from 16, further down from 28. Assuming we overcome charge bleed with non-silicon materials, I believe we'll hit the theoretical smallest size. From here it's a matter of knowing how many finite cores will fit and how much finite cache size we'll hit, before an entire redesign is required.
ScottL wrote:
It talks about the fact that they've reached the absolute smallest transister, an atom, which is .03NM wide. Current projections have us entering the 11NM size range, down from 16, further down from 28. Assuming we overcome charge bleed with non-silicon materials, I believe we'll hit the theoretical smallest size. From here it's a matter of knowing how many finite cores will fit and how much finite cache size we'll hit, before an entire redesign is required.
OK, now we are getting on common ground. I agree that we have 3-5 relatively simple doubling ahead (10-30x density improvement, 6-15 years ahead), then 2D "printing" will hit quantum levels.
I believe we'll see a redseign in our lifetimes.
So do I (unless there is no more demand for increased performance). But that means moore's law will still work, does not it?
As for major redesigns, I really do think there were many major redesigns in the past, which public has simply missed, and I think it might be the same in the future too. After all, it seems obvious where to go:
But an ~30X improvement in density (based on size reduction of the smallest feature) really means x squared or 900x improvement in the number of transistors on a die of the same size. So let's assume that current tech is the 22nm die used in Nehalem or similar processors. Current Nehalem processors have 6 cores or so. Even negating the impact of speed improvements due to shorter distances, going form 6 cores to 5400 cores or so should provide just a tad bit of uplift.
Gettting back to the real world, the limiting factor is really mempry. Short term scratch pad type stuff occurs in RAM today. A kind of medium ground, measured in microseconds instead of nanoseconds, occurs in flash (which I believe is basically dead 1-3 years out for numerous reasons, foremost of which is my personal contributions to the enterprise storage industry) and lastly spinning rust @ 4-5ms. Things like PCM, and others unmentionable, are driving this down. Now spinning rust may seem like a target, but density is going up by many orders of magnitude while latency decreases by a single order of magnitude evey generation. I think this spells doom for things like tape if they can't keep up, but rotating rust will be with us for a likely 50 to 100 years.
It's more interesting to talk about the future of things like DRAM. I do believe that PCM or memristors or some like tecknology will dispace it. PCM is already starting to in consumer devices like cell phones, and only requires economy of scale to hit mainstream. Memristors (the HP proprietarty thing) is probably 5 years out, but shows promise (Disclaimer: I don't work for HP; contraire, I work for an HP competitor on the storage front). All of this could be cast aside for single molicule transistor tech; except very cheap long term bulk data storage. Think about it. Today companies do things like backup to tape, then ship those tape off to places like Iron Mountain for long term offline storage. If you can make the cost lower and retrieval times faster... it sort of kills tape and burying it underneath a mountain in a mushroom farm ...
JohnFul wrote:But an ~30X improvement in density (based on size reduction of the smallest feature) really means x squared or 900x improvement in the number of transistors on a die of the same size.
Unfortunately, those doublings already count that in (density defined as number of transistors per 2D surface). So we can get really only get about 30x more transistors if we stay with 2D. I would also think there could be 'intelligent design' improvement, say about 2x-3x (after all, AMD Athlon 64 was quite faster CPU than P4 with virtually the same manufacturing process).
Still, it seems like we might need to get 10000x current performance to get something 'thinking'. Might sound a lot, OTOH we have without doubt seen this happen in past 30 years...
Prosthetics Breakthrough Might Fuse Nerves With Fake Limbs
Dirk, alongside colleagues at Sandia National Laboratories, the University of New Mexico and the MD Anderson Cancer Center, set out to develop a synthetic substance that could act as a scaffold — that is, an artificial structure that can support tissue growth — successfully merging severed nerves with robotic limbs.
Of course, researchers have already made plenty of efforts to directly integrate nerves and prosthetics. But, according to Dirk, they typically “didn’t use technology that was compatible with nerve fibers,” which are tightly bundled and flexible. “Nerves need to grow and move around; they’re not going to integrate well with a stiff interface.”
WASHINGTON (Reuters) - A four-legged robot known as the Cheetah lived up to its name on Monday, setting a new land speed record for legged robots by running at 18 mph (29 kph) on a treadmill at a laboratory in Massachusetts, its developer said.
DARPA is unveiling a portable laser weapons system, HELLADS, which seems like something out of a sci-fi movie. The new laser application, created by General Atomics Aeronautical Systems with a custom power system from Saft Batteries, will help change the way the American military fights future wars. Current military laser systems are bulky contraptions which are mainly the size of a passenger jet, while the proposed DARPA weapon can fit on the back of a flatbed truck. The 150-kilowatt, solid state laser weapon is strong enough to take down drones or other aerial targets; a prototype is expected to be available by the end of 2012.
The NSA Is Building the Country’s Biggest Spy Center (Watch What You Say)
Under construction by contractors with top-secret clearances, the blandly named Utah Data Center is being built for the National Security Agency. A project of immense secrecy, it is the final piece in a complex puzzle assembled over the past decade. Its purpose: to intercept, decipher, analyze, and store vast swaths of the world’s communications as they zap down from satellites and zip through the underground and undersea cables of international, foreign, and domestic networks. The heavily fortified $2 billion center should be up and running in September 2013. Flowing through its servers and routers and stored in near-bottomless databases will be all forms of communication, including the complete contents of private emails, cell phone calls, and Google searches, as well as all sorts of personal data trails—parking receipts, travel itineraries, bookstore purchases, and other digital “pocket litter.” It is, in some measure, the realization of the “total information awareness” program created during the first term of the Bush administration—an effort that was killed by Congress in 2003 after it caused an outcry over its potential for invading Americans’ privacy.
Yup, a humanoid robot, straight up Terminator style. Except instead of killing people this one is being designed to potentially save them by walking and crawling through narrow smoky passageways and using its sensors to figure out where to throw grenades loaded with fire retardent. Yup, firefighting grenades.
Re: the NSA project, what are they going to do if https, ssl, or similar encrypted communications become the norm?
Re: the fire fighting robot, A fire is a lot easier to identify than an enemy soldier, and fire retardant grenades far less serious if mistakenly aimed than a rifle.
hanelyp wrote:Re: the NSA project, what are they going to do if https, ssl, or similar encrypted communications become the norm?
Re: the fire fighting robot, A fire is a lot easier to identify than an enemy soldier, and fire retardant grenades far less serious if mistakenly aimed than a rifle.
In this thread, many of my comments are pseudo-tongue in Cheek. This is a subject of dark humor for myself and others.
Obviously fire fighting robots are only going to be an accidental threat to humans, but the point is that the technology keeps evolving and the abilities of these machines keep improving.
Assuming an ordinary projection, they will at some point be able to outperform humans in feats of strength and speed. At some point, who ever controls these machines will have a weapon with which they can quickly dispose of humans that don't agree with their policies;
Terminator Robots under the control of a human master.
‘What all the wise men promised has not happened, and what all the damned fools said would happen has come to pass.’
— Lord Melbourne —
I personally don't worry if NSA or ECHELON track what internet sites I visit. If the list of sites gets interesting enough to be switched from filters to humans then they can't sift through it without reading it, good for them.
hanelyp wrote:Re: the NSA project, what are they going to do if https, ssl, or similar encrypted communications become the norm?
Re: the fire fighting robot, A fire is a lot easier to identify than an enemy soldier, and fire retardant grenades far less serious if mistakenly aimed than a rifle.
In this thread, many of my comments are pseudo-tongue in Cheek. This is a subject of dark humor for myself and others.
Obviously fire fighting robots are only going to be an accidental threat to humans, but the point is that the technology keeps evolving and the abilities of these machines keep improving.
Assuming an ordinary projection, they will at some point be able to outperform humans in feats of strength and speed. At some point, who ever controls these machines will have a weapon with which they can quickly dispose of humans that don't agree with their policies;
Terminator Robots under the control of a human master.
And thats the thing..even if "true" AI remains elusive, an army of killer robots ranging from insect size to "terminator" size or larger under the control of our rulers. I mean nuclear war would devastate the planet, this obviously restrains to some degree even our craziest leaders. But these killer machines could selectively target people, leaving infrastructure, property the evnvironment alone.
williatw wrote:
I mean nuclear war would devastate the planet, this obviously restrains to some degree even our craziest leaders. But these killer machines could selectively target people, leaving infrastructure, property the evnvironment alone.
I am not quite sure how is that different from current status. Replace "killer machines" with "special forces" and the statement still holds true.
