Some of the trends an high-lights of the past year. Hex core chips have become mainstream now. The 32 nm chips fabrication is maturing. In the TOP500, clusters continue to gain over MPP systems and GPGPU did further evolve. Programming interfaces for it, did further develop and receive industry wide support. In network interconnect, Infinfiband and Gbit Ethernet are in a race for supremacy. It is to early to tell the outcome of that race. Supercomputing during the past year did also enable crucial scientific applications breakthroughs.

A big surprise was teh Chinese Nebula system, that made it to the second position in the TOP500.

On the chip level, Intel, IBM, and NVidia didunveil new chips. An interesting example is the IBM Power based Blue waters node, that was on display on the show at ISC10. The FPGA programming language CUDA and attached processor exite new generations of students, Sterilng said. Which is something HPC needs.

Hovwer, what happened to Itanium, seems like we have seen the demise of the Intel Itanium? for the HPC market. Several processors have beend discontinued. Hence, in memoriam for the Sun Rock and the Intel Larabee GPGPU chip.

Some new interesting machines are in Vhina the Tianhe-1 that is used for industrial an scientific applications; and in Russia the Lomonosov Russia is playing more and more a role in the provision of industrial grade supercomputers. Other interesting systems are Juropa in Germany and Shaheen in Saudi Arabia.

In system software, there were several advances including the release from PGi of version 10.5 that includes CUDA Fortan.

Oracle/Sun has released Lustre version 2.0 beta 1 but it is still unclear what the license status of Lustre will be.

In technology, an important advancement is graphene. Advancements in Technology CaTech researchers demonstrated a nanowire at 28nm scale. IBM got a graphen3transistor at 100 billion cycles/second.

In quantum computing new algorithms at MIT might be used for solving linear equations. Trillion solutions to a trillion variable problem are available fast, but reading out the data negates all won time. So it is really great until unless you want the answer.

Cloud computing will be important for the HPC community. For throughput and for rapid deployment. But not for low latency computing, Thomas Sterling said.

Green computing still attracts interest. When you do more green computing is that good? A lot of green computing still is a lot of computing.

Sterling considers HPC Education still as very important. So he mentions some initiatves frm LSU, the Great Lakes consortium, and the 10th LANL supercomputing challenge

Of course Exaflop/s was also a main issue during the past year. The first major effort, EISP - Exascale International Software Project - is a major effort, including people from all continents, except Antartica.