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

SGI: "We will be back in the visual supercomputing business";
http://insidehpc.com/2007/09/27/sgi-we-will-be-back-in-the-visual-supercomputing-business/

Criminals control four petaflops supercomputer;
http://insidehpc.com/2007/09/25/criminals-control-4-petaflops-supercomputer/

Verari offers new server-storage rack;
http://insidehpc.com/2007/09/26/verari-offers-new-serverstorage-rack/

ByteandSwitch.com hunts for the world's five largest SANs;
http://insidehpc.com/2007/09/26/worlds-5-largest-sans/

First Blue Gene/P plugged in;
http://insidehpc.com/2007/09/26/first-blue-genep-plugged-in/

SGI announces latest network-attached storage family;
http://insidehpc.com/2007/09/27/sgi-announces-latest-network-attached-storage-family/

First demo of quantum computing chip;
http://insidehpc.com/2007/09/27/qubits-on-a-chip/

>>PNNL signs on the line for an 18,000 core Opteron machine from HP

HP and Pacific Northwest National Laboratory announced that PNNL is installing a 163 TFLOPS system focused on research in the environmental molecular sciences. At that size it would have earned the number two slot on the June Top500 list. From the release (http://www.hpcwire.com/hpc/1805360.html):

The system will be a key capability in the Environmental Molecular Sciences Laboratory, a DOE national scientific user facility located at the PNNL in Richland, Wash. As such, the system will be available as a resource to scientists from around the world.

Stats? You bet:

The HP supercomputer architecture runs on HP ProLiant servers and includes an InfiniBand 4x DDR interconnect, 4,620 AMD Opteron processors, 37 terabytes of memory and aggregate disk bandwidth of about 950 gigabytes per second enabled by nearly 21,000 disk drives in HP enterprise virtual arrays. Consisting of 18,480 2.2 gigahertz AMD Opteron processor cores, the supercomputer will have an expected total peak performance of about 163 teraflops.

I believe this machine is the largest Barcelona-based system announced.

>>Woven's growth spurt

This week Woven Systems closed a $20M series B round of venture funding with a tier 1 venture capital firm. Woven plans to use the new-found capital to expand sales of its EFX 1000 line of ethernet fabric switches. Details are available at http://insidehpc.com/2007/09/25/woven-raises-20-million-in-vc-funding/ and http://insidehpc.com/2007/09/25/update-wovens-new-growth-spurt/.

This is a really interesting company with great technology, and it looks like they are really poised for growth: About six weeks ago, Woven added a new CEO with strong growth experience; their 144-port 10 Gb Ethernet layer 2 switch entered general availability a couple weeks ago (these switches can be connected together to build a fabric with 4,000 leaf ports all in layer 2); and now they've got the money they need to build out a sales and support organization. Check out the 4-1-1 on them for more background info: http://insidehpc.com/2007/08/29/the-411-woven-systems/.

The company is recruiting heavily right now. Yhey have a staff of 40 in house, and they have 22 open positions on their Web site, all but five of which are in manufacturing, marketing or sales. The company isn't talking publicly about what exactly it intends to do with the money, but if one was to gauge what company leaders intend to do with their shiny new $20M based on the hiring profile, it looks like most of that money is going into getting product out the door.

And sales is their big challenge. The Woven solution looks like it has a lot of advantages over competing solutions (especially strong is the adaptive congestion control as demonstrated by the Sandia/Chelsio results), but as the new kid, they have to get boots on the ground to start building the mindshare that leads to market share.

Looks like they are now set to give that the old college try.

>>ClearSpeed on the move

ClearSpeed had a raft of business announcements this week that indicate the company is continuing to gain the momentum it will need to innovate over the long term.

First, ClearSpeed has announced a partnership with Sun Microsystems to deliver their Advance accelerator boards for Sun high performance computing solutions. Most recently, the marriage of ClearSpeed and Sun yielded 48 TFLOPS at the Tokyo Institute of Technology. From the release (http://www.clearspeed.com/newsevents/news/pressreleases/ClearSpeed_Sun.php):

"Our partnership with ClearSpeed means that we can offer even better and more efficient performance to our customers," Dr. Simon See, director of HPC Solutions - Systems Practice at Sun commented. "HPC is a major growth area and many of our clients will benefit from the addition of application acceleration technology to their data centers. Sun's expertise in system design, configuration, and support will bring ClearSpeed's technology to them in a low risk, energy-efficient solution."

ClearSpeed also announced this week that Warwick University's Digital Laboratory is using ClearSpeed's e620 boards in a new Sun cluster for real-time rendering of virtual environments (http://insidehpc.com/2007/09/27/warwick-adopts-clearspeed-tech-for-rendering-in-virtual-environments/).

And, finally, the company announced a series of new reseller partnerships that will open China, Taiwan, and South Korea up to ClearSpeed's technology. In China and Taiwan the company is partnering with Super Sonic Technology  (http://www.clearspeed.com/newsevents/news/pressreleases/ClearSpeed_SuperSonic.php); and in South Korea, ClearSpeed is working with Tao Computing (http://www.clearspeed.com/newsevents/news/pressreleases/ClearSpeed_Tao.php).

-----