Slashdot Mirror
Warning At SC13 That Supercomputing Will Plateau Without a Disruptive Technology
dcblogs writes "At this year's supercomputing conference, SC13, there is worry that supercomputing faces a performance plateau unless a disruptive processing tech emerges. 'We have reached the end of the technological era' of CMOS, said William Gropp, chairman of the SC13 conference and a computer science professor at the University of Illinois at Urbana-Champaign. Gropp likened the supercomputer development terrain today to the advent of CMOS, the foundation of today's standard semiconductor technology. The arrival of CMOS was disruptive, but it fostered an expansive age of computing. The problem is 'we don't have a technology that is ready to be adopted as a replacement for CMOS,' said Gropp. 'We don't have anything at the level of maturity that allows you to bet your company on.' Peter Beckman, a top computer scientist at the Department of Energy's Argonne National Laboratory, and head of an international exascale software effort, said large supercomputer system prices have topped off at about $100 million 'so performance gains are not going to come from getting more expensive machines, because these are already incredibly expensive and powerful. So unless the technology really has some breakthroughs, we are imagining a slowing down.'"
Although carbon nanotube based processors are showing promise (Stanford project page; the group is at SC13 giving a talk about their MIPS CNT processor).
WASHINGTON – The U.S. Environmental Protection Agency (EPA) has made available $2 million in funding for rebates to help public and private construction equipment owners replace or retrofit older diesel construction engines. The rebates will reduce harmful pollution and improve air quality in local areas.
“Exhaust from diesel construction equipment affects children, senior citizens and others in neighborhoods across the country”,” said Janet McCabe, acting assistant administrator for EPA’s Office of Air and Radiation. "These rebates will help equipment owners protect public health and improve air quality near construction sites while updating their fleets.”
Rebates will be offered as part of the Diesel Emission Reduction Act, also known as DERA. This is the second rebate program offered since Congress reauthorized DERA in 2010 to allow rebates in addition to grants and revolving loans. The rebates will support the program’s effort to replace and update existing diesel vehicles, and will target where people are exposed to unhealthy air.
Since 2008, DERA has awarded more than $500 million to grantees across the country to retrofit, replace, or repower more than 50,000 vehicles. By cutting air pollution and preventing thousands of asthma attacks, emergency room visits and premature deaths, these clean diesel projects are projected to generate health benefits worth up to $8.2 billion.
Public and private construction equipment owners in eligible counties that are facing air quality challenges are encouraged to apply for rebates for the replacement or retrofit of construction equipment engines. EPA will accept applications from November 20, 2013, to January 15, 2014 and anticipates awarding the rebates in February 2014.
Construction equipment engines are very durable and can operate for decades. EPA has implemented standards to make diesel engines cleaner, but many older pieces of construction equipment remain in operation and predate these standards. Older diesel engines emit large amounts of pollutants such as nitrogen oxides (NOx) and particulate matter (PM). These pollutants are linked to health problems, including asthma, lung and heart disease, and even premature death. Equipment is readily available that can reduce emissions from these engines.
To learn more about the rebate program, the list of eligible counties, applicant eligibility and selection process, please visit http://www.epa.gov/cleandiesel/dera-rebate-construction.htm
do we need some failed (paid) hypenosys talknician to remind us?
Coding to make best use of resource.
Moving to clockless.
Minimal use processors (custom ASIC).
Live with it. Sometimes you may have to wait Seven. And a Half (what? not till next week?) Million Years for your answer. It may be a tricky problem.
MIPS CNT... how do you pronounce that?
Quantum/optical computing is going from impossible to possible within 1 decade. Is it really a "slowdown" or just a few years of lag time? We'll see.
http://www.washingtonpost.com/opinions/dana-milbank-the-gops-scary-movie-strategy/2013/11/19/41da73a4-5163-11e3-9e2c-e1d01116fd98_story.html
"the ones who join exchanges are likely to be older and sicker, making the insurance pool costlier to insurers. As Larry Levitt, a senior vice president of the Kaiser Family Foundation, explained to me, if costs are more than 3percent higher than anticipated in the first few years of Obamacare, the federal government will have to pick up at least half of the additional expense."
GOOD GRIEF!
This is news? This is not a "Republican Strategy", this is what we call the TRUTH!
And those of us with brains who have looked at this have known this all along!
What a freaking farce.
So what? Much of supercomputing is a tax-supported boondoggle. There are few supercomputers in the private sector. Many things that used to require supercomputers, from rocket flight planning to mould design, can now be done on desktops. Most US nuclear weapons were designed on machines with less than 1 MIPS.
Supercomputers have higher cost/MIPS than larger desktop machines. If you need a cluster, Amazon and others will rent you time on theirs. If you're sharing a supercomputer, and not using hours or days of time on single problems, you don't need one.
We've had Silicon Germanium cpus that can scale to 1000+ GHz for years. Graphene is also another interesting possibility.
The question is that "At what price can you make the power affordable?"
For 99% of people, computers are good enough. For the other 1% they never will be.
If only there was some newer technology to save us! Maybe something using light and quantum super states... idk... crazy talk...
that this is not a complete sentence:
:)
"Although carbon nanotube based processors are showing promise [...]."
Go, speed-editor, go!
my intuition tells me that disruptive technologies are precisely that because people don't anticipate them coming along nor do they anticipate the changes that will follow their introduction. not that people can't see disruptive tech ramping up, but often they don't.
My God can beat up your God. Just kidding...don't take offense. I know there's no God.
I suggest we go back a few levels and back to the 1970's when TTL was being replaced because of it's higher voltages. Remember back when core memory was replaced but before CMOSS? These were the TTL eras made by similar but NOT THE SAME transistors.
1. Silicone bandgap of CMOSS is higher than TTL
2. Gate length is more fabricable. (Fabricate the gates in Mexico; say they were made in USA)
3. Drain has "quantum clogging" problems in TTL but not CMOSS
4. Dopant levels make GaAs less commercially feasible.
5. Wafer sizes still dominated by "silicon" technology. It is not cheaper to go to more e-toxic and alien technologies. Far cheaper to stick with the wafers commercially produced today. GaAs and Indium Phosphate are like communion wafers!!!
6. Investors. We need to keep money at the fore front. Global depression is iminint. Must make cheep and available components with "WHAT WE HAVE" allready!!
TTL. I think its a good idea.
-KD
As a silicon engineer, I beckon you to mod parent up.
Didn't that boat sail with the Cray Y-MP?
All our really big supercomputers today are adding a bunch of individual not-even-Krypto-the-wonderdog CPUs together, and then calling it a supercomputer. Have we reached the limits in that scaling? No.
We have reached the limits in the ability to solve big problems that aren't parallelizable, due to the inability to produce individual CPU machines in the supercomputer range, but like I said, that boat sailed years ago.
This looks like a funding fishing expedition for the carbon nanotube processor research that was highlighted at the conference.
By definition, computers are scalable. Need more performance? Add more processing units/memory.
So what is this moronic dribbler dribbling about? Moore's Law allows for the COST of a given unit of computing to fall, taking into account improvements in the efficiency of mass production- so computing is boosted by more than just engineering improvements in performance per mm2 of 'chip'.
The truth is that too many pointless, and useless 'academics' need to work their narrow, and frequently irrelevant world view in order to continue receiving their pay-cheque. They therefore become sources of pure noise, and zero illumination.
With the growth of GPU processing, and ARM based SoC computers, the field of computing is taking one of its rare significant lurches forward. There is the famous saying that "those that can, do" and this applies just as much to the field of computing. Computer facilities with levels of processing unthinkable even a few years back are being built, many sadly for the use of the 'intelligence' community.
Powerful computing needs powerful infrastructure. The bad old days of near useless 'super computers' that focused purely on the fastest single processing engine itself are long since over. Now you want loads of processing engines, connected to masses of RAM and HDD storage- you know, like the facilities Google builds. But there are idiots like William Gropp, who haven't updated their knowledge of computing in decades (and we all met such idiots with high ranking positions at universities). They will bang on at conferences and seminars like its still 1975.
Computing is about many factors, of which absolute speed in a given CPU core is about the LEAST important. Performance per Watt and performance per Dollar is VASTLY more important. The general programmability of a given super computer system is also very important- powerful hardware is USELESS if using that hardware requires very difficult software.
Saying that existing chip engineering will eventual hit the physical limits of the current materials involved is self-evident, and completely pointless at this time. Suggesting that the existence of this limit means we must MAGIC up some future replacement is like saying we must magically create flying car technology to deal with the limitations of conventional car design. It is the kind of outburst only a truly third-rate, puffed up academic could make.
IF we find a new way to make chips, we will use it. However, throwing temper tantrums will NOT improve the likelihood of this happening. There is very reason to believe that only incremental improvements to current chip engineering will be usefully possible, with similar technology using better materials for conductors, insulators, and semi-conductors giving improvements at any given process process size. At this time, the industry is TOO focused on the need to 'shrink'. FD-SOI technology proves there are ways to squeeze far more performance from conventional chip production without shrinking.
As for loony SF fantasies of 'quantum' and 'optical' computers- well these pie-in-the-sky concepts will keep third rate academics employed for centuries without ever returning anything useful to the real world.
It may not be necessary to be so large if clockless.
To eliminate the wire-like or metallic nanotubes, the Stanford team switched off all the good CNTs. Then they pumped the semiconductor circuit full of electricity. All of that electricity concentrated in the metallic nanotubes, which grew so hot that they burned up and literally vaporized into tiny puffs of carbon dioxide. This sophisticated technique was able to eliminate virtually all of the metallic CNTs in the circuit at once.
Bypassing the misaligned nanotubes required even greater subtlety.
......
... are the biggest problems from where I'm sitting here in the convention center in Denver.
In short, there will need to be a serious collaborative effort between vendors and the scientists (most of whom are not computer scientists) in taking advantage of new technologies. GPUs, Intel MIC, etc. are all great only if you can write code that can exploit these accelerators. When you consider that the vast majority of parallel science codes are MPI only, this is a real problem. It is very much a nontrivial (if even possible) problem to tweak these legacy codes effectively.
Cray holds workshops where scientists can learn about these new topologies and some of the programming tricks to use them. But that is only a tiny step towards effectively utilizing them. I'm not picking on Cray; they're doing what they can do. But I would posit that before the next supercomputer is designed, that it is done with input from the scientists who will be using it. There are a scarce few people with both the deep physics background and the computer science background to do the heavy lifting.
In my opinion we may need to start from the ground up with many codes. But it is a Herculean effort. Why would I want to discard my two million lines of MPI-only F95 code that only ten years ago was serial F77? The current code works "well enough" to get science done.
The power problem - that is outside of my domain. I wish the hardware manufacturers all the luck in the world. It is a very real problem. There will be a limit to the amount of power any future supercomputer is allowed to consume.
Finally, compilers will not save us. They can only do so much. They can't write better code or redesign it. Code translators hold promise, but those are very complex.
A squid eating dough in a polyethylene bag is fast and bulbous, got me?
was going to be gallium arsenide, but it never made it to market.
"To those who are overly cautious, everything is impossible. "
There are plenty of algorithms that benefit from supercomputers. But it turns out that a lot of the justification for funding super computer research has been based on bad math. Check out this paper:
http://www.cs.binghamton.edu/~pmadden/pubs/dispelling-ieeedt-2013.pdf
It turns out that a lot of money has been spent to fund supercomputing research, but the researchers receiving that money were demonstrating the need for this research based on the wrong algorithms. This paper points out several highly parallelizable O(n-squared) algorithms that researchers have used. It seems that these people lack an understanding of basic computational complexity, because there are O(n log n) approaches to the same problems that can run much more quickly, using a lot less energy, on a single-processor desktop computer. But they’re not sexy because they’re not parallelizable.
Perhaps some honest mistakes have been made, it trends towards dishonestly as long as these researchers continue to use provably wrong methods.
The next step has already started.
Micronizing truely massive supercomputers is the next step for "applied sciences". We've gotten used to measuring data centres in power, I recon it will be computing power per cubic foot or something like that. It'll start with drones, then it will move to shipping and long haul robotics. After that it'll move to mining applications. I'm not talking about automating but rather truly autonomous applications that require massive computation for collision avoidance and programmed execution.
At this point it'll be a race to redo the industrial age, albeit with micronized robotics. Again, already started with 3D printing.
Hopefully by then someone figures out how to get off this rock.
That means there are hard limits to technology the NSA is using against us.
Totally off topic, but I ended up getting drunk with a bunch of people that are here in town for SC13 last night. Those boys can drink. But I'm surprised that there wasn't more talk about Tianhe-2 there, and how Chinese is going to kick the US off the top 25 in international supercomputing.
Yes, Moore's Law is just about over. Fortunately all signs point towards graphene transistors actually being workable within a decade. We can make them have a bandgap, we can produce ever larger crystals of pure graphene, we can isolate it from the environment to avoid contamination. We can, in labs, do everything needed to make graphene transistors already. Combining everything effectively and commercially may take a while, but it'll happen and by 2023 you'll be running your Google Glass v10's CPU at several hundred gigahertz with optical interconnects (which can already transition to graphene) with little heat and tons of battery life.
Meetup is the world's largest network of local groups. Meetup makes it easy for anyone to organize a local group or find one of the thousands already meeting up face-to-face. More than 9,000 groups get together in local communities each day, each one with the goal of improving themselves or their communities.For more info , please visit http://www.meetup.com/
D-Wave scaled up superconducting foundry output for their quantum chip, see no reason to not leverage this for conventional superconducting chips.
Intel's latest creations are basically x86-themed Transputers, which everyone (other than Intel) has been quite aware was inevitable. The only possible rival was processor-in-memory, but the research there has been dead for too long.
Interconnects are the challenge, but Infiniband is fast and the only reason Lightfleet never got their system working is because they hired managers. I could match Infiniband speeds inside of a month, using a wireless optical interconnect, with a budget similar to the one they started with.
Hard drives - you're doing it wrong. The drive should be battery-backed and have significant amounts of RAM on the controller. By significant, I mean enough to ensure that typical users will not be capable of distinguishing the hard drive from a RAM disk AND to ensure physical writes are delayed long enough that you maximize the number of writes that can be done as a sustained burst. (This reduces drive head movement, simultaneously making writes faster and drives longer-lasting.) Since this requires smart drives, you might as well have an OS on there. No sense wasting your main CPUs on filesystem work that can sensibly be offloaded.
For chips, for chrissakes get with the picture! Wafer-scale integration is the only way to go! There was also recently talk of some new gallium compound-on-silicon approach, with the gallium compound providing the transistors, silicon the connections. Interesting, but the US has outsourced almost all chip manufacture. If they want the new materials, they'll need to train a new generation of engineers and build new plants. If they're going to do that anyway, go wafer-scale. The initial costs will be high, when using the new materials, so you might as well go for broke and make the new stuff just too damn good to not get.
What's left? Ah yes, software. Easy. Pass a law stating that in order to get government funding, schools should teach Occam-pi and not Java. Big government? Too bad. Code quality out there is shite. The only way to fix that is to break bad code. Well, that or put the worst 10% of students likely to turn professional against the wall. However, that would produce a backlash. You could put bags of skittles in their pockets, I suppose.
There. That's the supercomputer industry fixed for another couple of decades. Do I get paid for this?
It's a small world and it smells funny; I'd buy another if it wasn't for the money; Take back what I paid (SoM)
50 years ago, state of the art was a billion dollars, equivalent to $25USD billion today. So they are .004 or 1/250 what they formerly cost. WTG Techies!
I'm not saying that lagging software is a problem: it's not. The problem is that there are so few real needs that justify the top, say, 10 computers. Most of the top500 are large not because they need to be - that is, that they'll be running one large job, but rather because it makes you look cool if you have a big computer/cock.
Most science is done at very modest (relative to top-of-the-list) sizes: say, under a few hundred cores. OK, maybe a few thousand. These days, a thousand cores will take less than 32u, and yes, could stand beside your desk, though you'd need more than one normal office circuit and some pretty decent airflow. I think people lose touch with the fact that our ability to build very big machines, cheaply, filled with extremely fast cores. You read all that whinging about how we hit the clock scaling (dennard) wall around the P4 and life has been hell ever since - bullshit! Today's cores are lots faster, and you get a boatload more of them for the same dollar and watt. And that's if you stick with completely conventional x86_64/openmp/mpi tech, not delving into proprietary stuff like Cuda.
People who watch the top of top500 closely are addicts of hero-numbers and hero-facilities. The fact is you can buy whatever position you want: just pay up. Certainly it's impressive how much effort goes into a top10 facility, but we should always be asking: what whole-machine job is going to run on it? IMO, the sweet spot for HPC is a few tens of racks - easy to find space, easy to manage, can provide enough resources for hundreds of researchers.
3D chips, memristors, spintronics. I am surprised these are not mentioned prominently in this thread. I was hoping to hear about the latest advances in these areas from people in the industry.
3D chips. As materials science and manufacturing precision advances, we will soon have multi-layered (starting at a few layers that Samsung already has, but up to 1000s) or even fully 3D chips with efficient heat dissipation. This would put the components closer together and streamline the close-range interconnects. Also, this increases "computation per rack unit volume", simplifying some space-related aspects of scaling.
Memristors. HP is ready to produce the first memristor chips but delays that for business reasons (how sad is that!) Others are also preparing products. Memristor technology enables a new approach to computing, combining memory and computation in one place. They are also quite fast (competitive with the current RAM) and energy-efficient, which means easier cooling and possible 3D layout.
Spintronics. Probably further in the future, but potentially very high-density and low-power technology actively developed by IBM, Hynix and a bunch of others. This one would push our computation density and power efficiency limits to another level, as it allows performing some computation using magnetic fields, without electrons actually moving in electrical current (excuse me for my layman understanding).
Photonics was already mentioned by others here.