October 30, 2012
From germanium to silicon and finally down to carbon. Back in the 1940s, scientists at Bell Labs purified germanium, a heavier element in the carbon/silicon family, used to make the first transistors. For the last several decades, scientists have been making smaller and smaller transistors out of silicon, doubling the transistor density every twelve to eighteen months, in accordance with Moore's Law.
Computing advances rely on the continued exponential growth of transistors per chip. However, there is a finite amount that silicon transistors can shrink, and there are signs that the trend has already slowed. While transistor density continues to increase, "clock speed," or the speed at which the transistors can turn on and off, has not. As such, processing in parallel to overcome this limitation has gained increased importance.
Carbon nanotubes would hypothetically kill two birds with one stone. Earlier this year, IBM presented a paper that demonstrated the ability of carbon to act as a transistor across just ten nanometers – half the size of silicon. Further, semi-conducting carbon provides a friendlier environment for electron movement. From a computing standpoint, that means data can be transported at higher speeds and that transistors can be more easily switched. This ultimately increases efficiency of the device by a factor of five to ten.
The issue with carbon nanotubes that prevented their consistently being turned into transistors is that they consist of both semi-conductive and metallic material. While the semi-conductive carbon nanotubes allow electrons to shift along them easier, the metallic side does not. As such, great pains have to be taken to sift out the metallic. Further, like their silicon brethren, carbon nanotubes have to be placed delicately and in a controlled fashion on the chip.
If those obstacles can be overcome, the technology has the potential to surpass silicon as the preferred transistor material. "The motivation to work on carbon nanotube transistors is that at extremely small nanoscale dimensions, they outperform transistors made from any other material," said Supratik Guha, director of Physical Sciences at IBM Research.
According to him, carbon could outperform conventional silicon by a factor of five. "However," he said, "there are challenges to address, such as ultra-high purity of the carbon nanotubes and deliberate placement at the nanoscale."
To address those challenges, IBM came up with a chemical method to attract the semi-conducting material. By immersing a substrate consisting of hafnium oxide and silicon oxide into a liquid solution of carbon nanotubes, the semi-conductors attach themselves to the hafnium oxide. As a result, IBM was able to place ten thousand carbon-based transistors onto a wafer using standard semi-conductor processes.
Of course, today's silicon-bearing chips harbor millions of transistors, a number that is expected to rise into the billions within the next couple of years. So carbon nanotubes still have a long way to go.
With that being said, IBM beat previous efforts, which were only able to mount a few hundred working carbon transistors. The advancements from here will come from being able to separate out the metallic from the semi-conducting, a process which is continues to be refined. According to Guha, IBM is fairly confident that by the end of the decade, they will be able to ensure 99.99 percent purity, leading to a point where carbon may take over as the element of choice for next-generation computing.
10/30/2013 | Cray, DDN, Mellanox, NetApp, ScaleMP, Supermicro, Xyratex | Creating data is easy… the challenge is getting it to the right place to make use of it. This paper discusses fresh solutions that can directly increase I/O efficiency, and the applications of these solutions to current, and new technology infrastructures.
10/01/2013 | IBM | A new trend is developing in the HPC space that is also affecting enterprise computing productivity with the arrival of “ultra-dense” hyper-scale servers.
Ken Claffey, SVP and General Manager at Xyratex, presents ClusterStor at the Vendor Showdown at ISC13 in Leipzig, Germany.
Join HPCwire Editor Nicole Hemsoth and Dr. David Bader from Georgia Tech as they take center stage on opening night at Atlanta's first Big Data Kick Off Week, filmed in front of a live audience. Nicole and David look at the evolution of HPC, today's big data challenges, discuss real world solutions, and reveal their predictions. Exactly what does the future holds for HPC?