Sign up for our Technology Newsletter 
IBM Still Talking Up SyNAPSE
IBM has unveiled the latest stage in its plans to generate a computer system that copies the human brain, calculating tasks that are relatively easy for humans but difficult for computers.
As part of the firm’s Systems of Neuromorphic Adaptive Plastic Scalable Electronics (SyNAPSE) project, IBM researchers have been working with Cornell University and Inilabs to create the programming language with $53m in funding from the Defense Advanced Research Projects Agency (DARPA).
First unveiled two years ago this month, the technology – which mimics both the size and power of humanity’s most complex organ – looks to solve the problems created by traditional computing models when handling vast amounts of high speed data.
IBM explained the new programming language, perhaps not in layman’s terms, by saying it “breaks the mould of sequential operation underlying today’s von Neumann architectures and computers” and instead “is tailored for a new class of distributed, highly interconnected, asynchronous, parallel, large-scale cognitive computing architectures”.
That, in English, basically means that it could be used to create next generation intelligent sensor networks that are capable of perception, action and cognition, the sorts of mental processes that humans take for granted and perform with ease.
Dr Dharmendra Modha, who heads the programme at IBM Research, expanded on what this might mean for the future, sayng that the time has come to move forward into the next stage of information technology.
“Today, we’re at another turning point in the history of information technology. The era that Backus and his contemporaries helped create, the programmable computing era, is being superseded by the era of cognitive computing.
“Increasingly, computers will gather huge quantities of data, reason over the data, and learn from their interactions with information and people. These new capabilities will help us penetrate complexity and make better decisions about everything from how to manage cities to how to solve confounding business problems.”
The hardware for IBM’s cognitive computers mimic the brain, as they are built around small “neurosynaptic cores”. The cores are modeled on the brain, and feature 256 “neurons” (processors), 256 “axons” (memory) and 64,000 “synapses” (communications between neurons and axons).
IBM suggested that potential uses for this technology could include a pair of glasses which assist the visually impaired when navigating through potentially hazardous environments. Taking in vast amounts of visual and sound data, the augmented reality glasses would highlight obstacles such as kerbs and cars, and steer the user clear of danger.
Other uses could include intelligent microphones that keep track of who is speaking to create an accurate transcript of any conversation.
In the long term, IBM hopes to build a cognitive computer scaled to 100 trillion synapses. This would fit inside a space with a volume of no more than two litres while consuming less than one kilowatt of power.
nVidia’s CUDA 5.5 Available
Nvidia has made its CUDA 5.5 release candidate supporting ARM based processors available for download.
Nvidia has been aggressively pushing its CUDA programming language as a way for developers to exploit the floating point performance of its GPUs. Now the firm has announced the availability of a CUDA 5.5 release candidate, the first version of the language that supports ARM based processors.
Aside from ARM support, Nvidia has improved supported Hyper-Q support and now allows developers to have MPI workload prioritisation. The firm also touted improved performance analysis and improved performance for cross-compilation on x86 processors.
Ian Buck, GM of GPU Computing Software at Nvidia said, “Since developers started using CUDA in 2006, successive generations of better, exponentially faster CUDA GPUs have dramatically boosted the performance of applications on x86-based systems. With support for ARM, the new CUDA release gives developers tremendous flexibility to quickly and easily add GPU acceleration to applications on the broadest range of next-generation HPC platforms.”
Nvidia’s support for ARM processors in CUDA 5.5 is an indication that it will release CUDA enabled Tegra processors in the near future. However outside of the firm’s own Tegra processors, CUDA support is largely useless, as almost all other chip designers have chosen OpenCL as the programming language for their GPUs.
Nvidia did not say when it will release CUDA 5.5, but in the meantime the firm’s release candidate supports Windows, Mac OS X and just about every major Linux distribution.
Are CUDA Applications Limited?
Acceleware said at Nvidia’s GPU Technology Conference (GTC) today that most algorithms that run on GPGPUs are bound by GPU memory size.
Acceleware is partly funded by Nvidia to provide developer training for CUDA to help sell the language to those that are used to traditional C and C++ programming. The firm said that most CUDA algorithms are now limited by GPU local memory size rather than GPU computational performance.
Both AMD and Nvidia provide general purpose GPU (GPGPU) accelerator parts that provide significantly faster computational processing than traditional CPUs, however they have only between 6GB and 8GB of local memory that constrains the size of the dataset the GPU can process. While developers can push more data from system main memory, the latency cost negates the raw performance benefit of the GPU.
Kelly Goss, training program manager at Acceleware, said that “most algorithms are memory bound rather than GPU bound” and “maximising memory usage is key” to optimising GPGPU performance.
She further said that developers need to understand and take advantage of the memory hierarchy of Nvidia’s Kepler GPU and look at ways of reducing the number of memory accesses for every line of GPU computing.
The point Goss was making is that GPU computing is relatively cheap in terms of clock cycles relative to the time it takes to fetch data from local memory, let alone loading GPU memory from system main memory.
Goss, talking to a room full of developers, proceeded to outline some of the performance characteristics of the memory hierarchy in Nvidia’s Kepler GPU architecture, showing the level of detail that CUDA programmers need to pay attention to if they want to extract the full performance potential from Nvidia’s GPGPU computing architecture.
Given Goss’s observation that algorithms running on Nvidia’s GPGPUs are often constrained by local memory size rather than by the GPU itself, the firm might want to look at simplifying the tiers of memory involved and increasing the amount of GPU local memory so that CUDA software developers can process larger datasets.
nVidia Speaks On Performance Issue
Nvidia has said that most of the outlandish performance increase figures touted by GPGPU vendors was down to poor original code rather than sheer brute force computing power provided by GPUs.
Both AMD and Nvidia have been using real-world code examples and projects to promote the performance of their respective GPGPU accelerators for years, but now it seems some of the eye popping figures including speed ups of 100x or 200x were not down to just the computing power of GPGPUs. Sumit Gupta, GM of Nvidia’s Tesla business said that such figures were generally down to starting with unoptimized CPU code.
During Intel’s Xeon Phi pre-launch press conference call, the firm cast doubt on some of the orders of magnitude speed up claims that had been bandied about for years. Now Gupta told The INQUIRER that while those large speed ups did happen, it was possible because of poorly optimized code to begin with, thus the bar was set very low.
Gupta said, “Most of the time when you saw the 100x, 200x and larger numbers those came from universities. Nvidia may have taken university work and shown it and it has an 100x on it, but really most of those gains came from academic work. Typically we find when you investigate why someone got 100x [speed up] is because they didn’t have good CPU code to begin with. When you investigate why they didn’t have good CPU code you find that typically they are domain scientist’s not computer science guys – biologists, chemists, physics – and they wrote some C code and it wasn’t good on the CPU. It turns out most of those people find it easier to code in CUDA C or CUDA Fortran than they do to use MPI or Pthreads to go to multi-core CPUs, so CUDA programming for a GPU is easier than multi-core CPU programming.”
Future PCs Will Be Constant Learners
Comments Off on Future PCs Will Be Constant Learners
Tomorrow’s computers will constantly improve their understanding of the data they work with, which in turn will aid them in providing users with more appropriate information, predicted the software mastermind behind IBM’s Watson system.
Computers in the future “will learn through interacting with us. They will not necessarily require us to sit down and explicitly program them, but through continuous interaction with humans they will start to understand the kind of data and the kind of computation we need,” said IBM Fellow David Ferrucci, who was IBM’s principal investigator for Watson technologies. Ferrucci spoke at the IBM Smarter Computing Executive Forum, held Wednesday in New York.
“This notion of learning through collaboration and interaction is where we think computing is going,” he said.
IBM’s Watson project was an exercise for the company in how to build machines that can better anticipate user needs.
IBM researchers spent four years developing Watson, a supercomputer designed specifically to compete in the TV quiz show “Jeopardy,” a contest that took place last year. On “Jeopardy,” contestants are asked a range of questions across a wide variety of topic areas.
Watson did win at its “Jeopardy” match. Now IBM thinks the Watson computing model can have a wide range of uses.
Big Blue Still The Patent King
As technology companies start to stockpile patents so that they can see off their rivals IFI Claims Patent Services, a company that maintains global patent databases, has clocked the outfits with the most weapons in any patent war.
More than 224,505 utility patents were awarded in the U.S. last year, jumping two percent over the previous year’s record-breaking tally of 219,614 patents. IBM has always had the most patents, probably because it has been around the longest. The company was granted 6,180 utility patents, up nearly five percent from 2010. Samsung was the number two 4,894 patents, followed by Canon at 2,821 patents, Panasonic with 2,559 and Toshiba with 2,483 utility patents.
Microsoft, which held on to the third spot in 2010, is in the sixth place with 2,311 utility patents granted last year, According to IFI CEO Mike Baycroft global companies, and especially Asian ones, are collecting U.S patents at a dizzying pace, and now Asian firms hold eight of the top 10 slots in the 2011 ranking.
IBM’s Watson Shows Up For Work
IBM’s Watson supercomputer is about to start work evaluating evidence-based cancer treatment options that can be delivered to the doctors in a matter of seconds for assessment.
IBM and WellPoint, which is Blue Cross Blue Shield’s largest health plan, are developing applications that will essentially turn the Watson computer into an adviser for oncologists at Cedars-Sinai’s Samuel Oschin Comprehensive Cancer Institute in Los Angeles, according to Steve Gold, director of worldwide marketing for IBM Watson Solutions.
Cedars-Sinai’s historical data on cancer as well as its current clinical records will be ingested into an iteration of IBM’s Watson that will reside at WellPoint’s headquarters. The computer will act as a medical data repository on multiple types of cancer. WellPoint will then work with Cedars-Sinai physicians to design and develop applications as well as validate their capabilities.
Dr. M. William Audeh, medical director of the cancer institute, will work closely with WellPoint’s clinical experts to provide advice on how the Watson may be best used in clinical practice to support increased understanding of the evolving body of knowledge on cancer, including emerging therapies not widely known by physicians.
IBM announced earlier this year that healthcare would be the first commercial application for the computer, which defeated two human champions on the popular television game show Jeopardy! in February.
China’s Supercomputer Uses Homegrown Chips
Comments Off on China’s Supercomputer Uses Homegrown Chips
China has built its latest supercomputer based entirely on homegrown microprocessors, a major move towards breaking the country’s reliance on Western technology for high-performance computing.
China’s National Supercomputer Center in Jinan debuted the computer last Thursday, according to a report from the country’s state-run press. The supercomputer uses 8,704 “Shenwei 1600″ microprocessors, which were developed by a design center in Shanghai, called the National High Performance Integrated Circuit Design Center.
Details of the microprocessors and the design center were not immediately available.
The supercomputer has a theoretical peak speed of 1.07 petaflops (quadrillion floating-point calculations per second), and a sustained performance of 0.79 petaflops when measured with the Linpack benchmark. This could place it at number 13 in the world’s top 500 supercomputing list. Photos of the chips used and the supercomputer’s data center can be found here.
China’s Shandong Academy of Sciences built the computer. Officials of the academy could not be immediately reached for comment on Monday.
A report from The New York Times said the supercomputer’s name in English was the Sunway BlueLight MPP.
Japan Takes 1st Place On Supercomputer List
Comments Off on Japan Takes 1st Place On Supercomputer List
A Japanese computer has earned the number one spot on the Top 500 supercomputer list, ending China’s short reign of just six months. At 8.16 petaflops (quadrillion floating-point calculations per second), the K computer is more powerful than the next five systems combined.
The K computer’s performance was measured using 68,544 SPARC64 VIIIfx CPUs each with eight cores, for a total of 548,352 cores, almost twice as many as any other system on the Top500 list. The computer is still being put together, and when it enters service in November 2012 will have more than 80,000 SPARC64 VIIIfx CPUs according to its manufacturer, Fujitsu.
Japan’s ascension to the top means that the Chinese Tianhe-1A supercomputer, which took the number 1 position in November last year, is now in second spot with its 2.57 petaflops. But China continues to grow the number of systems it has on the list, up from 42 to 62 systems. The change at the top also means that Jaguar, built for the U.S. Department of Energy (DOE), is bumped down to third place.