Here’s a collection of highlights, selected totally subjectively, from this week’s HPC news stream as reported at insideHPC.com and HPCwire.
>>10 words and a link
First customer ship for SiCortex: 6 TFLOPS, less than 20kW;
First Infiniband-based, scalable Network Attached Storage solution;
Linux Magazine on HPC file systems;
New Globus incubator project aims to make HPC easier;
QLogic and partners demo Fibre Channel over Ethernet solution;
New tech from Hitachi: 4 TB 3.5″ drive by 2011;
Fujitsu debuts new SAN for small to mid-size enterprises;
Animation Studio ROCKS roll announced;
FSU material could give boost to quantum computing;
>>AMD can sell ’em, they just can’t ship ’em
Apparently AMD’s new quad-core Opteron processors are selling like hotcakes. The bad news is they can’t make them fast enough. Damon Poeter at CMP Channel writes that AMD’s partners are generally happy with the new Opteron products, but at least some are having trouble getting a hold of enough parts.
“There are no hardware conflicts and the power draw is as promised. They delivered on their technicals. On these high-performance compute and memory-intensive applications, they’re kicking Intel’s butt,” said Brian Corn, VP of marketing and business development at Waltham, Mass.-based Source Code.
This is a far cry from where we were in June, when partners were publicly questioning whether AMD could get Barcelona’s performance high enough to launch (http://insidehpc.com/2007/06/06/barcelona-demod-at-16-ghz-partners-question-july-launch/).
But the news isn’t all good; when it comes to delivering on actual product, Corn is not so enthusiastic about AMD’s performance.
“We’re extremely disappointed with AMD on a product delivery level…. The real problem seems to be is that AMD doesn’t have any of these things.”
According to some of AMD’s channel partners, the company is giving the Tier 1 system vendors and some favored partners first crack at the new quads, leaving the dregs for the channel.
Did AMD push Barcelona out the door too fast or did it just underestimate the demand? According to Corn, AMD hasn’t offered an explanation for the hold-up.
Read the full story at http://v3.crn.com/white-box/202402138.
>>IBM makes progress on carbon nanotube-based computing
IBM has announced today that it’s been playing around with carbon nanotubes, and has come up with a way to measure the distribution of electrical charges in tubes smaller than 2 nm across. This is an incremental step along the path toward use of carbon nanotubes as semiconductors and wires on chips.
This novel technique, which relies on the interactions between electrons and phonons, provides a detailed understanding of the electrical behavior of carbon nanotubes, a material that shows promise as a building block for much smaller, faster and lower power computer chips compared to today’s conventional silicon transistors.
From the announcement (http://www-03.ibm.com/press/us/en/pressrelease/22441.wss):
“The success of nanoelectronics will largely depend on the ability to prepare well characterized and reproducible nano-structures, such as carbon nanotubes,” said Dr. Phaedon Avouris, IBM Fellow and lead researcher for IBM’s carbon nanotube efforts. “Using this technique, we are now able to see and understand the local electronic behavior of individual carbon nanotubes.”
>>Europe’s leading super online
German publication heise online is reporting that Jülich Supercomputing Centre has installed a 220 teraflop Blue Gene/P system, which is now the most powerful supercomputer in Europe. The article also points out that since the machine uses a mere 500 kilowatts, the new Blue Gene/P is one of the most energy efficient computers in the world. (For comparison, the 500 teraflop Sun supercluster being installed at the Texas Advanced Computing Center will draw 2.4 megawatts — almost five times as much for just twice the FLOPS.)
The heise online article also points to a recent presentation given by Alan Gara, the chief architect for Blue Gene, who foresees a persistent supercomputing energy crisis in the years ahead:
Mr. Gara is convinced that it will be possible at some time between 2015 and 2020 to achieve peak performances of 200 petaflops per second, but that the machines capable of such feats will require 25 to 50 megawatts of energy. And this assessment already takes a 20-fold improvement in energy efficiency for granted.
Read the full story at http://www.heise.de/english/newsticker/news/97535.