Visit additional Tabor Communication Publications
October 04, 2012
I was more than a little perplexed at a recent article I ran across on Tech Radar that suggested high performance will be free, or nearly so, by 2020. You know, like how nuclear power was going to be "too cheap to meter" once the technology became ubiquitous.
The Tech Radar article, penned by self-described tech addict Jeremy Laird, was an interpretation of some presentations at the recent Intel Developer Forum (IDF) in San Francisco. The premise of his piece is based on Intel's plans to keep Moore's Law on track for at least another 10 years, which will put 5nm microprocessors into circulation around the end of the decade. If so, the silicon logic that exists on today's chips will be able to be crammed into just six percent of the space on 2020-era microprocessors.
The almost-free-HPC leap of faith occurs if you now believe that transistor shrinkage corresponds to system price. Even for the chips alone, that's not the case, since the fabs become much more expensive at smaller manufacturing nodes, circuit design testing becomes more difficult, and labor costs tend to rise. So those extra expenses have to be factored in. It is certainly true that peak FLOPS and OPS do become cheaper as you shrink the transistors, just not in a Moore's Law kind of way.
To deliver HPC into the land of the free you also would need to ignore the other components that go into a machine, such as memory, power supplies, interconnects, external storage, and so on. And unfortunately, not all of these technologies are on a Moore's Law trajectory. Laird does admit that the processor is just one element of the machine, but brushes it off. (And, by the way, most of his discussion seems to focus on personal computing devices, not HPC machinery, but the general argument is the same.)
His real concern, though, is how Intel or any chipmaker is going to make money if chips are nearly free. The answer: "sheer volume." Since chips will be so cheap, he writes, "they'll be sticking compute into almost everything." In that sense, he's probably on target. Processors destined for embedded computing are already relatively inexpensive and as they become cheaper, they'll find their way into even more applications.
But even these chips won't be free, and the systems they end up in certainly won't be either. In 2020, your $500 smartphone isn't going to be $30; it's more likely to be $100. (But it will be equipped with a phaser, so there's that.)
For HPC, the chip economics are actually quite different than that of the consumer space. The microprocessors that inhabit supercomputers tend to be more-expensive chips – lower volume, higher margin – and you need lots of them to make up a single system. In any case, compared to the other hardware components that go into the machine, the processor will still represent a significant chunk of the total cost.
But even if high-volume embedded processors infiltrated HPC, that wouldn't lead to free computing either. Just consider the DRAM market, whose suppliers have been selling their high-volume silicon wares at a loss because of oversupply. Despite that, in many cases, the expense of DRAM limits how much memory can be stuffed into HPC systems.
On the other side are those who complain about the high costs of HPC, especially as it relates to future exascale systems. If trend lines hold, a top-tier 2020 supercomputer will cost between $300 and $400 million. That's three to four times the price of Roadrunner, the first petaflop supercomputer, which went into production in 2008. Running that 2020 exaflop system will probably require at least 20 MW of power, costing an additional $20 million per year in the US, and even more in most of Europe.
So much for free. In fact, it's not unreasonable to question whether such systems would even be affordable for individual national labs. And if that's the case, there's less incentive for vendors to even design and build these machines. The top of the TOP500 is already a dicey business, and if that space shrinks too much, system makers will exit the market.
I'm guessing the future of HPC lies somewhere between those two extremes. There's every reason to believe that in 2020 there will be plenty of demand for high performance computing systems – from supercomputers to small clusters. If so, chipmakers and system vendors will find a way to pay for manufacturing and labor costs, while adding in a profit margin that makes sense for their business. So here's my bold predication: In 2020, HPC will be priced somewhere between unaffordable and free.
May 16, 2013 |
When it comes to cloud, long distances mean unacceptably high latencies. Researchers from the University of Bonn in Germany examined those latency issues of doing CFD modeling in the cloud by utilizing a common CFD and its utilization in HPC instance types including both CPU and GPU cores of Amazon EC2.
May 15, 2013 |
Supercomputers at the Department of Energy’s National Energy Research Scientific Computing Center (NERSC) have worked on important computational problems such as collapse of the atomic state, the optimization of chemical catalysts, and now modeling popping bubbles.
May 10, 2013 |
Program provides cash awards up to $10,000 for the best open-source end-user applications deployed on 100G network.
May 09, 2013 |
The Japanese government has revealed its plans to best its previous K Computer efforts with what they hope will be the first exascale system...
May 08, 2013 |
For engineers looking to leverage high-performance computing, the accessibility of a cloud-based approach is a powerful draw, but there are costs that may not be readily apparent.
05/10/2013 | Cleversafe, Cray, DDN, NetApp, & Panasas | From Wall Street to Hollywood, drug discovery to homeland security, companies and organizations of all sizes and stripes are coming face to face with the challenges – and opportunities – afforded by Big Data. Before anyone can utilize these extraordinary data repositories, however, they must first harness and manage their data stores, and do so utilizing technologies that underscore affordability, security, and scalability.
04/15/2013 | Bull | “50% of HPC users say their largest jobs scale to 120 cores or less.” How about yours? Are your codes ready to take advantage of today’s and tomorrow’s ultra-parallel HPC systems? Download this White Paper by Analysts Intersect360 Research to see what Bull and Intel’s Center for Excellence in Parallel Programming can do for your codes.
In this demonstration of SGI DMF ZeroWatt disk solution, Dr. Eng Lim Goh, SGI CTO, discusses a function of SGI DMF software to reduce costs and power consumption in an exascale (Big Data) storage datacenter.
The Cray CS300-AC cluster supercomputer offers energy efficient, air-cooled design based on modular, industry-standard platforms featuring the latest processor and network technologies and a wide range of datacenter cooling requirements.