Limits to Moore's Law Launch New Computing Quests

tringtring alerts us to news that the National Science Foundation has requested $20 million in funding to work on "Science and Engineering Beyond Moore's Law." The PC World article goes on to say that the effort "would fund academic research on technologies, including carbon nanotubes, quantum computing and massively multicore computers, that could improve and replace current transistor technology." tringtring notes that quantum computing has received funding on its own lately, and work on multicore chips has intensified the hunt for parallel programming. Also, improvements are still being made to current transistor mechanics.

Comments

Is this necessary?

[nebaz]

I don't really think a prize is necessary for this technology. Unlike space travel, reearch in chip design have shown to be profitable at the commercial level, and there is also no government monopoly to stifle progress in this area. Whether or not a prize is offered, faster computers and better technology are what we as consumers expect in this area, and what we will pay for.

Re:Is this necessary?

[Eravnrekaree]

Capital investment is most important with things like this. The purpose of the prize however might not be that it is necessary, but rather to accelerate development, speed it up. The benefits of course of improving technology would be significant, more powerful, faster computers.

Re:Is this necessary?

[aztektum]

It isn't a $20mil prize, it's a budget request.

Re:Is this necessary?

[Mike1024]

I don't really think a prize is necessary for this technology.

Who said anything about a prize? The PC World article talks about 'funding for research', i.e. cash given to researchers to develop new technology.

Unlike space travel, reearch in chip design have shown to be profitable at the commercial level, [...] Whether or not a prize is offered, faster computers and better technology are what we as consumers expect in this area, and what we will pay for.

It's true that a lot of commercial effort goes into current chips and the improvement thereof, but there isn't much commercial effort going into areas like quantum computing because the potential rewards are a loooooong way off. Your money is much safer invested in designing a 32-core Core2ThirtyTwo to be made in 3 years, compared to quantum computing, a technology that faces substantial scalability roadblocks and that no-one knows how to design algorithms for.

Most of the current quantum computers which have been demonstrated rely on Nuclear magnetic resonance (NMR), but it is thought this technique will not scale well - it is believed less than 100 qubits would be possible. As of 2006, the largest quantum computer ever demonstrated was 12 qubits (making it capable of such tasks as quickly finding the prime factors of a number... as long as that number is less than 4096.

In summary, promising future technologies often make poor investments because they are (a) experimental and (b) a long way off. So some funding to make research possible wouldn't go amiss.

Just my $0.02.

Re:Is this necessary?

[smaddox]

"Life is like riding a bicycle. To keep your balance you must keep moving."

-Albert Einstein

Re:Is this necessary?

[ScottyH]

you are in the wrong place.

Re:Is this necessary?

[mrxak]

If there actually was a prize, which it doesn't sound like it FTA, it would probably serve more for bragging rights than anything else. Similar to how SpaceShipOne cost $25 million and only got a $10 million prize, the publicity and fame was probably worth it whether the prize was there or not, but the prize actually got everybody's attention.

But, it's just a request for funding, not a prize, so it doesn't matter.

Re:Is this necessary?

[Anonymous Coward]

Sweet, I'd like to formally request, say $10 million, to advise Intel and AMD et al that they are falling behind on following Moore's law and need to develop a 128bit (or maybe more) processors ASAP. A few letters and a handful of meetings should do the trick.

Re:Is this necessary?

[Anonymous Coward]

Bragging rights or not, the $25 million SpaceShipOne came out costing $15 million instead of $25 million.

"If you build it, they will come..."

[syncrotic]

...is a mentality that probably won't work here.

Intel sunk billions into the development of Itanium on the premise that if they make a VLIW architecture, compiler developers will find a way to automatically extract the parallelism necessary to make good use of it. A company with the size, resources, and engineering knowledge of Intel made the mistake of assuming that a fundamental shift in thinking could be driven by money and sheer desire, but it turns out that the problem is not just hard - that would make it solvable given sufficient effort and money - it's actually impossible. Those compiler advances never materialized; you can't draw blood from a stone.

The quest for parallelism in ordinary software might just be similar. Developing tools to make this automated and easy with low overhead is akin to putting a dozen smart people in a room and saying "think up the next big idea that will make me millions." Innovation doesn't work that way; it can't be forced... and money isn't going to make the impossible into the possible.

I think we'll see a move to eight and then maybe even sixteen cores on a consumer-level chip before we see things start going back in the other direction. This will necessary mean a slowdown in the development of processors as CPU manufacturers go back to wringing every last bit of single-threaded performance out of their designs.

Thoughts?

Re:"If you build it, they will come..."

[ZorbaTHut]

While I agree with some of your points, I disagree with your details. There's no proof that compilers can't be made smart enough for that - just because it didn't happen doesn't mean it couldn't. The biggest issues with Itanium were that it was incredibly slow on "non-native" code and incredibly expensive. There was no reason for anyone to buy one without having a compiler built specifically for it, and there was no reason for anyone to spent the effort to write a compiler for it without someone having bought one.

It's possible that if Itanium had been able to execute x64 or even x86 code at a competitive speed, we'd all be using IA-64 by now (or at least hoping that new programs were recompiled with it.)

Also, I don't actually think we'll have a shift back to single-threaded apps. The fact is that most programs run "fast enough" now, even single-threaded on quadcore systems. The ones that don't (mostly games and some professional software) are frequently relatively easy to multithread. I suspect most programs will stay single-threaded, and the ones that need maximum speed will become extremely multithreaded.

Re:"If you build it, they will come..."

[Colin Smith]

Thoughts? Stop pushing on bits of string. Start pulling them...

Re:"If you build it, they will come..."

[setagllib]

As an anecdote, the single application most important to my work (and hobby) is Eclipse. For the amount of machine assistance it can provide towards productive work, you pay heavily in RAM and CPU power. For a particularly prolonged session on a large project, I can have my entire operating system occupying 100MB of RAM, and Eclipse occupying as much as 800MB, leaving very little room even on a 1GB stick.

It's not that Eclipse has always been this heavy, rather, innovations in machine assistance expand the software to fill the hardware. It's like that in web browsers, document editors, media players, etc.

So as long as we keep expanding the hardware, the software will expand proportionally. For applications that truly require extreme amounts of data processing, they're almost universally so parallel that they can be run on a cluster anyway.

Fortunately the open source space has a broad spectrum from very efficient software all the way up to software which does a lot of clever things for you at the cost of system resources. The commercial space is a bit top-heavy because feature-rich software sells well and sells for a lot, since "one program to do everything" looks great in marketing.

Re:"If you build it, they will come..."

[master_p]

You should have researched the Actor model Erlang and automatic parallelization of purely functional programs before you made that assertion.

Parallelism for ordinary software its already here, it's a matter of time before it is adopted by mainstream applications.

Re:"If you build it, they will come..."

[Anonymous Coward]

The comparison with Itanium is specious.

Itanium was a totally new and unproven architecture. It was expensive, it required new software, a huge amount of compiler research, and it was still only going to get you a relatively small increase in speed. And of course nobody invested enough effort in the compilers and nobody felt like adopting an unproven architecture so the whole thing tanked.

Compare this to multicore processing. I have a desktop PC with four cores in it. I routinely run software which uses them all. In fact I'm running a build in the background right now which is doing a pretty good job of keeping them all occupied.

Multicore processors are the same old architecture, just more of it. They're cheap, they work with existing software and existing compilers, and a lot of software will show nearly linear increases with the number of cores. Of course it all won't, which is why people are researching this so much, but the point is that a lot of the software which I have and run today will happily use more than one core.

These are the key differences. Software was taking advantage of multicore processing from before day one, back when you had to buy two CPUs to get two cores. Whereas for Itanium it all had to be built from scratch. It's vastly easier to get people to make evolutionary improvements, which is exactly what this is. And then Itanium was never going to give you a particularly large increase in performance, particularly when considering performance per dollar or per watt, whereas the performance potential of multicore systems is effectively unlimited.

Re:"If you build it, they will come..."

[epine]

This is even sillier than your reply indicates. By the same logic we could conclude by the failure of the Pentium IV that x86 was doomed. Then Intel came up with Core Duo to show what should have been achieved in the first place (and saved the world many gigawatt years of unnecessary power generation in the meanwhile).

Intel botched their first hack at Itanium. They weren't willing to pony up another couple of billion to get it right the second time. By then their performance war against AMD had set the bar so high on x86 performance, their "pull the rug and own the world" marketing strategy was no longer viable (not even within the Intel boardroom ego chamber).

Intel killed Itanium in the false belief they could go cold turkey on out-of-order (OoO) execution. It's true that OoO scales badly, both in terms of complexity and power consumption, as you broaden the execution pathways. Intel's solution: exterminate OoO. Right from the beginning I thought this was a daft and deadly embrace of determinism.

The sensible solution: constrain the design to a reasonable fixed upper bound on OoO depth. They could have done this by having bundles express groups of *dependent* operations. The OoO ceiling would then be a single bundle unit.

I would have set up bundles to encode at least five operations under ideal conditions. If the operations are dependent, you need to specify fewer total registers (and in fact, commit fewer total registers back to the register file, which can only be an advantage). I'm sure this would complicate validation and maybe there are some other gotchas I haven't consider, but it always seemed obvious to me that this would work better relative to the algorithms I worked with (which have sources of non-determinism you can't eliminate).

You would also add a rule concerning bundle independence. Say the architecture was designed to scale up to four bundles wide, with a peak five operations per bundle. Each of the bundles within the bundle group would be required to be fully independent (shared inputs would be allowed, nothing more).

You'd probably set up four bundle execution pipelines (each internally with an OoO dispatch queue). There would have to be rules on the maximum rate of forwarding operands from one bundle pipeline to another. Some portions of the register file would be somewhat "bound" to particular bundle pipelines. You'd have to sacrifice register file orthogonality. But it's a false orthogonality to begin with: a recently computed result can only be forwarded so far in a fixed time increment, and you can't afford to provision worst-case forwarding pathways from everywhere to everywhere on a fat uarch.

The Itanium design team was seduced by determinism and orthogonality. Partly this was because x86 instruction encoding is a horror show. With a more sensible ISA, you could have 90% of the advantage of x86 non-orthogonality (code density improvements) at 10% of the complexity of an x86 instruction decoder. I've been saying this for years. Finally, ARM came up with Thumb-2 to demonstrate my point: the best design is a carefully constrained and balanced non-orthogonality.

http://www.arm.com/products/CPUs/archi-thumb2.html

Why didn't ARM do this long ago? The 16/32 decoding mode is maybe 5% as difficult as x86 decoding, and look at the huge advantage it gives you in balanced time/space.

For an Itanium bundle, I would set up the rules for a finite fan out / fan in for every instruction field. In my version of things, a 128 bit bundle would be able (under ideal circumstances) to encode five operations. A bundle decoder must break the bundle into up to five independent instructions.

In my approach, you might have a rule that any bit-field within the bundle can have at most four distinct possible destinations (within the five exploded instructions). Each field within the exploded instructions can obtain bits from (one of) at most four different

Re:"If you build it, they will come..."

[Chandon Seldon]

Itanic failed because the machines had horrible price/performance except in very tiny niches. One of the things that killed Itanic is x86 clusters - aka. parallel programs.

Multicore processors, in contrast, are free. What I mean by that is this: Dual core processors cost basically the same as single core processors at the same core speed. You can still buy single core processors today, but nobody does - there's no reason not to take the free second core. For a variety of reasons, the same thing will be true for quad core processors in a couple years.

By 2011 or so, everyone will have a multicore processor, every application will have multicore support, and looking back the question will seem stupid.

Itanic really was a good architecture - but nobody spent any time on it because nobody actually had the processors. Symmetric multicore x86 is quite a bit less elegant, but it'll get used extremely effectively because *everyone* will have one to play with.

Re:"If you build it, they will come..."

[ykardia]

While I agree with some of your points, I disagree with your details. There's no proof that compilers can't be made smart enough for that - just because it didn't happen doesn't mean it couldn't. While I agree with some of your points, I disagree with your details. There's no proof that compilers can't be made smart enough for that - just because it didn't happen doesn't mean it couldn't. In fact, I think people are working on dealing with things like Nested Data Parallelism (pdf) in compilers right now. I think this will happen in functional languages very, very soon (Haskell, someone below mentioned Erlang). Simpler things, like dealing with flat data parallelism via the compiler (+ a special library) have been possible for a while (see e.g. OpenMP).

Re:"If you build it, they will come..."

[rbanffy]

I think Itanic missed its window of opportunity.

By the time the systems were shipping and there was an mainstream OS (read "Windows") to run applications on it, the AMD64 and multi-core x86 processors were already appearing.

Had HP invested more on HP-UX over the years (making it escape the narrow niche they carved for it), had Linux been more mainstream by that timeframe (read "a decent desktop OS", which it kind of wasn't), had Intel invested a lot of resources making GCC deliver the promised performance on the chips and had they been cheaper, AMD64 and P4HT would have had a heck of a more difficult time in the server market.

There is a lot of "had"s.

I run Linux most of the time. I really don't care what binary architecture my box uses as long as the package repositories support it. It would be really fun to have an s/390-based notebook or a 36-hour-battery multi-core ARM-11 subnotebook, BTW. I have a bunch of Unix workstations and only the trained eye can tell if someone is running CDE on Solaris, AIX or HP-UX. They look and largely feel the same. And they feel the same whether you are sitting on a single processor box or beside a 16 CPU behemoth (wish I had one of those - they sure are interesting). I run OSX and it feels like Unix, despite the Mach microkernel. And it looks the same whether it's on a PowerPC or an Intel Core.

Today, for anyone running Unix-like OSs, binary architecture is largely irrelevant.

The real question is

[Anonymous Coward]

How much experience is this quest worth?

patents?

[Neffirithion]

say they do get these carbon tubing and other stuff that would massively accelerate the technology worlds... Would they have patents on them as well as the 20 million? If so why have the prize? you'll just have to licence the technology from them anyway, so who ever does will be dirt rich + 20 million in pocket... If there is a hole in my thinking... please point it out to me.

~Neff

Re:patents?

[Jasin Natael]

It's not a prize. It's funding; A budget. This is the older-than-dirt story of, "If you build it, they will come!" vs. "I have a 0.01% chance of succeeding if I try to build it, so who's going to feed my family in the 99.99% probable case that I fail?"

cost per computation / 3-D Chips

[fpgaprogrammer]

Moore's law is an observation about the cost per transistor in a circuit. Making faster computation is all about transistor density and the distance signals must travel. Even after the 2-D transistor density levels off, the race will be on to make cheaper 3-D chips using wafer-bonding methods, giving us a new dimension to increase density and thus speed up computation:
http://mtlweb.mit.edu/researchgroups/icsystems/3dcsg/

And we'll still see the same exponential benefits to GOPs/$ for a long time after 3-D transistor density maxes out. The economics that drive the exponential cost-per-computation trend are more related to volume of demand which offsets high fixed production costs and less related to our ability to actually cram more transistors on a chip.

Re:cost per computation / 3-D Chips

[mean pun]

Even if we keep getting exponential growth of transistors per dollar in the coming years, the question is what to do with them. Arranging them in useful circuits is increasingly difficult because at a certain point adding cache and execution units to a processor just isn't very helpful (hence multi-core). Adding more cores is also not going to help at some point. Moreover, power dissipation can't keep growing proportionally, which means that with increasing transistor counts each transistor will have to dissipate less, which means lowering the average number of switching events per transistor, and how are we going to arrange for that?

Re:cost per computation / 3-D Chips

[fpgaprogrammer]

the hard part is, of course, how do we program it; there are plenty of applications that benefit from parallelization (graphics processing, SDR, FEM). parallelization tends to offer equivalent throughput at a lower rate of switching. we need to review whether high frequency switching is really worth the power when you have trillions of transistors in a cubic centimeter. at todays price for 1 million gate FPGA, a 1 trillion gate FPGA array would cost about $10-20M, I expect this will be down by a factor of 1000 in under 10 years. operating at 100 MHz it would be hard to not have a petaflop of computation. lower frequency requirements lets us get creative with power. power density (temperature) is directly related to the number of switching capacitance in a region. lower frequency circuits and asynchronous circuits can reduce the effect of the most major sources of switching (clock). with adiabatic logic systems you can actually use charge pumping and charge recovery to eliminate capacitive loss during switching but these circuits operate slower. when you combine asynchronous and adiabatic logic you can actually use the REQ-ACK handshake as a charge pump to power on functional units. and if you really want high frequency switching, you'll need to remove thermal energy. one of the early uses of carbon nanotubes in circuits may be as thermal channels. it's also possible to create submersible circuitry using microfluidic ducts to cool wafer-stacked chips.

Re:cost per computation / 3-D Chips

[jelle]

3D stacking of wafers with transistors is not the solution:

If someone were to try it, they better get working on methods to cool those stacks of wafers well, and ways to make the wafers cheaper...

If you make a chip with a stack of, say 10 wafers, you've also had to diffuse 10 wafers, costing, well, the same ten chips of only one wafer... Diffusing doesn't magically get cheaper when you stack the wafers afterwards. I'm sure the 'wafer-bonding' costs some dough too.

And it generates the heat of 10 chips of one wafer...

Now how much does a quad-core single-wafer chip currently dissipate? Isn't that in the range of 90-150 Watts?

Ten of those?

How many Watts does a maker use to boil a pot of water? How much a clothes-iron?

And 18 months later 20 wafers, then 40, then 80?

You get the picture...

Re:cost per computation / 3-D Chips

[lenski]

Check the research into reversible computing:

Widipedia

Control waste heat by managing entropy.

Re:cost per computation / 3-D Chips

[jelle]

I don't know what you're smoking, but.... I'd like to see a working prototype ;-)

Re:cost per computation / 3-D Chips

[Eivind]

Actually, transistors can also become more effective, and have been for decades. If not, you'd be right: Doublt computing-power would mean double power-consumption which would mean double heat-production and spell heaps of trouble.

We're still -far- away from the theoretical limits though.

Flipping a -single- bit MUST consume atleast kT joules, where T is the temperature in Kelvin and k is the Bolzmann-constant of around 10^-23.

So if your cpu runs at 300K (cooling it more won't help because then you'll spend energy for that) you can flip a physical max of sligthly over 10^20 bits. Run at 10Ghz, aproximately 10^10 Hz and you can flip a maximum of 10^10 bits every clock-cycle.

Current CPUs don't come anywhere close to being that efficient. They flip perhaps a million transistors in a clocktick which is a factor of 10000 less than they COULD be doing with a single watt, and they spend not ONE watt but more like 50W to do that.

Still, the limits are visible: We can, theoretically, up cpu-efficiency by a factor of 100.000, but a factor of a million looks, imho, physically impossible.

This sounds like a lot, but consider that if we keep doubling every 2 years, we'll hit the hard wall in 20 - 30 years. And long before we get close to the hard wall we'll be in a pretty steep terrain.

Unless we go with reversible computing in which case all bets are off and entirely new limits apply.

Re:cost per computation / 3-D Chips

[slash.duncan]

Even if we keep getting exponential growth of transistors per dollar in the coming years, the question is what to do with them. Arranging them in useful circuits is increasingly difficult because at a certain point adding cache and execution units to a processor just isn't very helpful (hence multi-core).

I disagree with at least part of the above.

The problem is that you're not thinking outside the current box we're in. It's not that we have too many transistors and too much "cache", but rather that we have too few, and will continue to have too few for a few Moore's law generations yet.

Consider the effect once that "cache" reaches the half-gig and higher level (and consider that the current cache sizes are per-core, so multiply by the number of cores to get the total per chip size, since that's what we're now talking). Z-RAM http://en.wikipedia.org/wiki/Z-RAM is an interesting technology in this regard. At a half-gig, special applications (I'm assuming operating memory of say 8 gigs is common for general purpose by that point) will begin to make use of this memory as main memory. About 3-4 generations later, on-chip memory capacities will get very reasonable for general purpose use.

Currently, program design and compiler optimization must account for the fact that main memory is off-chip somewhere, running at order of magnitude lower clock rates compared to on-chip CPU speeds. Almost certainly, on-chip memory will still run at fractions of the CPU clock, but getting it on-chip will reverse the trend of the last decade plus of higher and higher CPU to memory speed ratios.

Remember the effect of moving L2 cache from mainboard to CPU daugher-board with the slot CPU generation, and then directly on-chip as transistor budgets allowed with the move back to sockets? Of course, other adaptations were happening at the same time, as main memory got even slower as compared to the CPU. That's where a lot of the out of order and speculative execution stuff that we just take for granted now days came from -- it was forced by the ever increasing speed of the CPU as opposed to RAM, but enabled by the larger/faster L1 and L2 cache sizes.

Now imagine being able to throw most of that away once again, as memory speeds actually catch up several doublers worth because they are now on-chip. That's what we're looking at in a few years, if Moore's law continues to hold and on-chip transistor budgets continue to increase.

The problem currently is that we're in a kind of lukewarm middle ground, increasing cache size doesn't bring so much bang for the transistor budget any more, but the budget is still a few generations too limited to fully integrate main-memory on-chip. As a result, we're devising all these other things to make use of those available transistors with. However, at some point, the transistor budget is going to get big enough to consider integrating main memory, and at that point, I suspect a lot of these other technologies will be at least temporarily dropped.

Yet another angle on the same thing is the fact that a decent share of current chip real estate and transistor budget goes to the whole speculative and out of order execution thing, as forced by the speed differential between CPU compute-cores and main memory. If memory speeds suddenly increase several-fold, the effect of sticking it on-chip, a good share of that speculative logic will no longer be needed as the cost of main memory access, both in direct latency and in branch prediction failure penalty, will be much smaller. Off-hand, I'd guess that'll cut a generation off the transistor budget wait right there, thus bringing it that much closer.

One thing's for sure, there'll be quite a bit of optimization un-learning to go thru once this occurs. Throw in the 1-T Z-RAM tech (assuming that still next-gen tech is actually viable both technically when fused with CPU foundary tech, and that it's even commercially viable at all), and

This may take a while ...

[ScrewMaster]

Well, with any luck they'll get that massively-multicore quantum processor out in time for the release of Duke Nukem Forever. I hear the frame rate will be awesome.

Killer app?

[sakdoctor]

And what would be the killer app that needed all that extra power?
Moore's Law might be linear but who's to say that demand for processing power is also... ...scratch that, Microsoft just released a new operating system. The minimum spec is 640 quantum cores.

Re:Killer app?

[deadline]

"The minimum spec is 640 quantum cores."

Actually, it reads 640 universes. You are guaranteed to get the right answer in one of these.

Re:Killer app?

[mangu]

And what would be the killer app that needed all that extra power?

The short answer is all the applications that run in these computers.

I can think of at least two applications that are often in the news: protein folding and physical simulations of continuous media, like weather and climate, aerodynamics, water, oil, and gas flow in porous rocks, etc.

But I think the future applications for personal supercomputers haven't been invented yet. We don't have the brains to predict what super-human artificial intelligence will be like.

Re:Killer app?

[squeeze69]

640 quantum cores should be enough for anybody.... :-)

Re:Killer app?

[Anonymous Coward]

Think of the holodeck on Star Trek. The AI necessary to drive it aside, consider the amount of processing power needed to simulate a physical universe in sufficient detail that you can't tell it from the real thing. Now that would be the holy grail of, at least, the entertainment industry, which as others have pointed out drives much of the quest for greater processing power.

We may never achieve something like the holodeck, but every improvement in processing power allows us to get a little bit closer. Entertainment aside, such technology would no doubt have all kinds revolutionary scientific and industrial applications. Hence there is effectively no limit to the demand for greater computing power.

Re:Killer app?

[ScrewMaster]

We may never achieve something like the holodeck

Because processing power is not what will keep us from developing a Holodeck ... the ability to synthesize matter at will and manipulate it with such detail is a much more difficult task.

However, we may very well achieve some more along the lines of The Matrix or James P. Hogan's star-spanning Visar AI. I'd say it's almost a given that we'll eventually develop some kind of direct neural interface into a computer-generated virtual reality.

Re:Killer app?

[master_p]

How about:

1) parallel search
2) accurate text translation
3) accurate human speech rendering
4) raytracing for 3d graphics
5) advanced physics in 3d applications
6) more dynamic programming languages
7) better video and audio decompression
8) much faster compression
9) ultra fast large WORD document repagination
etc

Re:Killer app?

[Anonymous Coward]

And what would be the killer app that needed all that extra power?

Well, there are a couple, graphics processing for example. Governments in particular however might be interested in two different areas which would profit considerably from massively parallel computing: (semi-)brute force cryptanalysis and simulation (think weapons, in particular, nuclear ones since it's difficult and expensive to do real tests with them).

Re:Killer app?

[HiThere]

You don't need to wait for those. By the time the computers arrive there'll be LOTS of jobs that they could do. E.g., given a genome and a blood sample, predict the best drug to use to treat disease X.

(Yeah, we don't have the database that program would need yet. But it's already being worked on.)

You can also use them to do ray tracking in a changing 3-D environment. (Think realistic games. Lots of people will pay for that one.)

Re:Killer app?

[zippthorne]

Now that would be the holy grail of, at least, the entertainment industry, which as others have pointed out drives much of the quest for greater processing power.

I really don't see how that's possible. The entire US film industry's gross receipts were only $10 billion last year. Microsoft's net income last year was $17 billion, and Microsoft is only ONE company in the tech industry.

Intel, one of the major producers of the actual processing power had a net profit of nearly $7 billion.

I think it's safe to say that the special effects industry far from a principle driver of processing speed advances.

Re:Killer app?

[Anonymous Coward]

And what would be the killer app that needed all that extra power? You must not be too aware of 3D rendering software. You'd be suprised at how long it can take to get a high quality 1024x768 sized image on a typical desktop, let alone try to do multiple frames of animation. Rendering involving raytracing and photon mapping can always use more power.

See here for an extensive list of existing programs that may be able to make use of any extra processing power you can throw at them.

If there was a computer that could render complex raytraced animations in realtime, I'm pretty sure there's a market for that.

video games

[nten]

I'm just making this up as I go along but it sounds plausible. I suspect the Itanium Epic failed because it targeted too small a niche. Supercomputers and larger servers could get better cost/benefit from using desktop processors in larger numbers. First because those processors had more research money behind them so they improved faster than a niche processor, and second because the compilers were already there. IBM learned, and is using that lesson with the CELL in the PS3. Video games and the mass market driving economies of scale is just too powerful a force to be improved upon by something as minor as VLIW. If they had marketed the Itanium the way IBM markets the CELL it might have worked.

It seems trivial that an entertainment escape like a video game should drive technology, but it is. If they find a way to make a quantum, optical, or nanotube computer/console easy to obtain and get 133t framerates with, then it people will beat the path to their door to play madden 2024 in all its glory.

Moore's Law is bullshit.

[Xiph1980]

Moore just happened to make a prognosis that transistordensity would double every 2 years.
It just happened to work out that way. We're about to reach a point where current transistors won't cut in anymore. At such a point we'll either stagnate because we can't make a smaller process than 10 nanometer and we can't find a different functional tech, or we'll make an enormous jump in performance because we'll find something in a different field, be it optics or nano-tubing, that does make processors a lot faster.

Moore's law isn't a law, and should never have been called that way. It's merely a prognosis.
microprocessor technology is driven by the market. If the general consumer thinks their pc is fast enough, manufacturers will focus on energy-efficiency to sell more cpu's, and speed will start to be a secondary concern.

Re:Moore's Law is bullshit.

[Tablizer]

That does not make it "bullshit", but just less than a mathematical or economic certainty. It's an observation that has held up.

Re:Moore's Law is bullshit.

[Anonymous Coward]

My professors always called it "Moore's Observation".

Re:Moore's Law is bullshit.

[Nullav]

If the general consumer thinks their pc is fast enough, manufacturers will focus on energy-efficiency to sell more cpu's, and speed will start to be a secondary concern.

Sounds good to me.

Re:Moore's Law is bullshit.

[Anonymous Coward]

But Moore's Prognosis doesn't rhyme!

Re:Moore's Law is bullshit.

[iocat]

Or maybe we'll go back to respecting efficiently written code, as opposed to tolerating things like chat programs that require ungodly amounts of RAM (Trillian, I'm looking at you), or OSes that require GBs of RAM (ahem, MS) and waste trillions of cycles due to shoddy programming, just because we have computers that can run 1000s of times faster than they did in the 80s... but provide essentially the same user experience and speed of execution.

Re:Moore's Law is bullshit.

[mdwh2]

My professors always called it "Moore's Observation".

A law is an observation. That's what it means. There seems to be a misconception that "law" means "fundament law of nature which is 100% proven to be true", but that's not correct.

(True, you might say it's not a scientific law in that it's not an observation about scientific matters, but instead, an observation about technology, economics and other factors, but it's still a law. You might as well moan that it's not a legal law - big deal.)

Re:Moore's Law is bullshit.

[mdwh2]

Moore's law isn't a law, and should never have been called that way. It's merely a prognosis.

Actually, I believe he based it on observed past behaviour, so even though he may have intended it to also be a prediction, calling it a law is fine.

I'm also confused by "merely" - I'd argue that saying it is a prognosis carries the implication that it will hold in future, whilst "law" implies that, just like other laws, it is merely a generalisation of observed behaviour.

Re:Moore's Law is bullshit.

[Nicolay77]

I have to add: Moore's law is not about density in a physics way, but about transistor density per a fixed amount of money. That's it, the cost of a fixed number of transistors will halve each two years.

So it's more like an economic law.

About your argument 'prognosis vs law', well, almost anything in economics is more a prognosis than a law, but whatever, nobody cares.

The Hunt

[nick_davison]

and work on multicore chips has intensified the hunt for parallel programming. Try down the back of the couch. Whenever born again Christians come to the door and ask me if I've found Jesus, that's where I tell them he was all along. Everything ends up there eventually.

Breakin' da law, breakin' da law...

[Anonymous Coward]

So, is Moores Law a law or the quota the industry need to meet?

Officer : "Sir, I'll have to arrest you for breaking Moore's Law"

Intel exec : "Oh noes!"

Diamond is a virtual's best friend

[liquiddark]

Don't worry. In a few years it won't just be for high-maintenance ladies anymore. Instead, you'll be able to melt through the surface of the earth while running spreadsheet calculations at lightning speed.

Re:Diamond is a virtual's best friend

[Anonymous Coward]

Maybe its just me, but I never did read your article. I was too busy clearing my schedule break-out style on the advertisement. Thanks!

never?

[Anonymous Coward]

never gonna happen, we have more pressing needs.

Re:Killer app? Second Life

[lenski]

Physical implementation cannot take advantage of Moore's Observation (great phrase, by the way). This is why my new quad phenom with 3 Gbytes RAM and 500 Gbytes of disk costs $750 but my VW Passat wagon cost me $25000 (and that's getting a great deal on it in 2002!)

You are right that it will be like the matrix or Hogan's story (sounds like an interesting story, I'll have to check it out).

A fully virtualized environment benefits directly from precisely the same exponential improvements that have occurred and will continue to occur in information processing technologies.

I would bet a fair chunk of change that the first entity that passes the Turing Test will be found in a virtual world. And in the virtual world, "holodeck" is exactly what it is, no new development necessary. Other than that slick neural interface. I don't see that happning, though...

I'd bet that we will implement Kruzweil's plan, which is to "download ourselves" to a virtual environment in order to get those benefits. I liken it to a transporter replication process, where the replicant (did I say that?) ends up in the virtual world and the original copy remains in meatspace. Or should the still perfectly viable living meat-based original person be terminated after the replication is complete, following the rule that "There Can Be Only One"?

Re:Killer app? Second Life

[ScrewMaster]

Just FYI, the books you want to read are Inherit The Stars, The Gentle Giants of Ganymede and Giant's Star, all by Hogan. He later wrote some additional books in the series. Yes, they're pretty good reads, if you like reasonably hard sci-fi. I think you can get them on Amazon.

We *will* see single threaded apps... well sort of

[UberMongoose]

"Also, I don't actually think we'll have a shift back to single-threaded apps"

Not completely true. We might still see single threaded at the conceptual level with languages supporting latent parallelism, even though the program flow is conceptually single threaded. Think parallel "for" loops and futures. The burden of actually distributing the parallel execution would be the language runtime's responsibility. This way you have code that acts and behaves like it is single threaded but actually scales on processors as more cores get added.

$20 million is nothing

[ChrisGilliard]

If you can come up with a technology that continues Moore's beyond the limits of silicon, you will make a lot of money and everyone knows this. That is why Intel, HP, Ibm, etc are investing billions in this kind of research. I don't have any complaints against the nsf funding this research but it's a little insignificant compared to the rest of the market funded research going on and I think this money could be better spent on other things that the free market is not already funding.

Re:Moore's Law is really about price-performance

[billstewart]

Sure, Moore expressed his observations in terms of transistor density, and the speed factor of 1-2 years has varied a bit over the last few decades, but what it's really about is price-performance of technology in a positive-reinforcement market. If you want to sell more chips, you either have to make them faster or cheaper or both, unless you're the only player in an underserved market.

So the expensive fast chips get faster to sell to customers with the need for speed, and the production technology gets refined to make more chips cheaper at a given speed, so the currently-fast speeds get cheaper, and the currently-cheap chips get faster, but on the other hand you do spend more capital on each new generation of fab plant.

And as the chips get faster, the software makers use up the available speed, and as the software makes machines slower (but more useful, or more friendly, or more popular), the customers want faster chips or bigger memories or bigger disks or all of the above.

The big threats to Moore's Law right now aren't so much that we're running into the edge of silicon technology, but that Microsoft Vista is sufficiently unsatisfactory that people aren't buying it unless it ships on their new laptops, so there's less demand for faster machines, and also that gamers are playing more MMORPGs, where faster CPUs and graphics chips don't make as much difference in game capability as they do with standalone games (but even so, a cutting-edge graphics card costs more than a business-class desktop computer.)

On the other hand, virtualization (which is pretty much the reinvention of time-sharing) is pushing the business sector toward doing new and exciting technology for clustering storage, and at least creating some demand for RAM, and using up some of those multi-core CPUs even though they're buying fewer of them. And we're starting to hit environments where the cost of electricity for cooling and power exceeds the cost of the CPU itself, so price-performance is starting to get measured in watts/bogomips, rather than just dollars/bogomips.

What about Single Core Performance

[nateman1352]

Everyone in the computing industry these days seems to be mesmerized by multiple core computers. What about improving single core performance? There has been many papers published back in the 1970s showing that multiple cores are really not effective at improving performance of a single application assuming that it has more than 1% non-parallizable code. Moreover as the number of cores increase it has been shown that thread management starts consuming the majority of the system's time, not the actual application. So even though it looks busier its not actually doing anything useful.

You know there is still people working on increasing single core performance. What about Professor John McDonald at RPI, one of the industry's smartest and most regarded researchers, hes attempting to build a 32GHz single core right now.

Re:What about Single Core Performance

[GordonCopestake]

Reminds me of PATA and SATA. We are moving to SATA away from PATA because it's faster to just run through all the instructions on one go (Single Core) than attempt to bring all the different answers together at the end of the IO operation (PATA). I would rather have a single 32Ghz CPU than a 1Ghz CPU with 32 cores, and I can bet it will run all my real world apps a hell of a lot faster.

Automatic Parallelism is a falicy...

[FlyingGuy]

Being able to automate the task of sending off threads to various cores is pretty much and impossibility. The level of exceptions to any set of rules that allow the compiler or even a run time environment with managed code would be so large that the MCP would be in a constant busy state just figuring out if it was possible to send various threads of to n numbers of cores. much less keeping all the threads synced and sorting out the wait times for various threads on various CPU's to finally all be finished and allowing the main thread ( if there even will be a single main thread ) to collect the results of all the other threads that were arbitrarily launched by the MCP work devision algorithm.

I think that for the foreseeable seriously powerful parallel processing will remain the domain of those coders who can take a set of problems and hand code them into the various threads and ensure the synchronization does not fall apart.

What happened to GaAs FET and germanium?

[Beliskner]

What happened to Gallium Arsenide technology? It's supposed to be 10 times faster than silicon

And what about silicon germanium?

Re:Killer app? Second Life

[Eivind]

They cannot directly. But did you ever stop to consider what you really paid for when you bougth the vw for $25K ? It's basically all work. The hours of the people putting the car together. The hours of the people making the parts. The hours of the people minin the metals. The hours of the people drilling for the oil used to provide the energy for mining the metals and so on.

Increased mechanisation and computer-power means a single individual can do more and more in the same time, which again leads to the price in relative terms becoming less and less. (relative as in compared to the cost of one work-day)

Influences goods that consist of a lot of work and a little bit of raw-materials more than goods that are the other way around offcourse, but the trend is still there: we're getting more and more productive in other areas than electronics too, the progress is just much slower. Doubling is more like 20 years than 20 months.

Perhaps it should have been called 'Moores Code'?

[Anonymous Coward]

In a pirates of the carribean sense. As the code is more a guideline than a rule.

Re:Killer app? Second Life

[ScrewMaster]

Offtopic, my foot.