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

HPL gets updated;
http://insidehpc.com/2008/09/24/hpl-gets-an-upgrade/

Tilera revamps design in attempt to scale memory wall;
http://insidehpc.com/2008/09/25/tilera-revamps-chip-in-attempt-to-scale-memory-wall/

Gizmodo spends quality time with bigger-than-it-looks CX1;
http://gizmodo.com/5053076/hands-on-cray-cx1-windows-supercomputer-one-day-itll-make-crysis-cry

insideHPC follows Google trail and comes out with CX1 suspect;
http://insidehpc.com/2008/09/22/the-insidetrack-crays-cx1-partner-unmasked/

Saudi super to be a Blue Gene;
http://insidehpc.com/2008/09/23/saudi-super-to-be-a-blue-gene/

NASA buys 4k core iDataPlex system;
http://insidehpc.com/2008/09/23/nccs-buys-ibm-idataplex/

HPC Server 2008 shrink wrapped;
http://www.microsoft.com/presspass/press/2008/sep08/09-22HPCLaunch08PR.mspx

Microsoft opens new computing research lab in New England;
http://www.hpcwire.com/offthewire/Microsoft_Opens_New_England_Research_Lab.html

SGI revs license terms on OpenGL Sample Implementation, other software;
http://insidehpc.com/2008/09/23/sgi-lifts-opengl-software-kilt-a-little-further/

>>IBM expands On Demand offering, HPC Server test drive and vis

IBM announced two changes to its On Demand compute offering this week.

For the low low price of $99, IBM will let you give Windows HPC Server a test drive:

IBM said today it will offer remote, $99 "test drives" of Microsoft's newly launched supercomputer operating system, Windows HPC Server 2008, via its global network of IBM Computing on Demand facilities.

You can get your test drive in units of 14 or 16 nodes on IBM BladeCenter or System x servers. IBM maintains a broad selection of pay-by-the-hour supercomputing hardware in its Computing on Demand centers, which offer more than 56 terabytes of storage and 13,000 processors.

IBM also announced the addition of visualization services for remote users:

IBM also said that it would make available to on-demand users its high-end 3D visualization engine known as Deep Computing Visualization (DCV). Already used by clients in the automotive, life sciences, aerospace and energy industries, DCV allows for extremely detailed 3D modeling of data. By tapping into DCV on demand, clients can reduce in-house network bottlenecks and offer powerful 3D graphics capabilities to more users. DCV also allows multiple remote users to work on the same data simultaneously.

This is something that many in the community, including me at the day job, are working on, so I'm glad to see this.

>>Just like your datacenter, only in a tent

This week we had word from Nick Carr (whom I've never met but insist on using the familiar form of his name anyway) on a new Microsoft effort to test datacenters deployed in tents. No, seriously.

We have seen datacenters in semitrailers, datacenters in caves, datacenters in Siberia, datacenters in the Las Vegas desert, and datacenters that float in the middle of the ocean. Today we have word, via Data Center Knowledge, that Microsoft has been testing datacenters in tents. (They're calling it In Tents Computing.)

Evidently this is part of a larger effort by engineers at MS to get down to datacenters with a PUE of 1.0 (Power Usage Effectiveness -- a PUE of 1.0 means that all the power put into a datacenter is going into compute; you can read more about it at The Green Grid's site here [warning: PDF]).

One of the dudes attempting this feat of back-to-nature computing has a great blog post about it:

As a former server designer, I know that server vendors "sandbag" their hardware. Sandbagging refers to the practice of knowing you can do more but holding back to hedge your risks; in reality, I believe that manufacturers can take greater risks in their operating environments and still achieve the same reliability levels. Knowing about the chronic sandbagging in the industry, I thought that if I could run some servers in the Building 2 garage or somewhere were the equipment is at least protected from the rain, we could show the world the idea is not that crazy and is worth exploring.

...Inside the tent, we had five HP DL585s running Sandra from November 2007 to June 2008 and we had ZERO failures or 100 percent uptime.

Cool anecdotes: water drips on the rack from the tent without incident, and a windstorm blew a fence onto the rack with no ill effect. The post is a great read, and I'm offering a free case of beer to the first one of you who sets up a cluster of at least eight nodes in the out of doors running LINPACK for a week. Two cases if you have a tropical storm during the experiment.

>>NVIDIA announces 6.5 percent workforce reduction

NVIDIA has announced that it will be the latest addition to a long list of corporate downsizing. At the end of October, it will kindly wave goodbye to 6.5 percent of its workforce, or roughly 360 people. Bummer. According to company representatives, the move comes in the face of strategic growth initiatives. One such initiative is NVIDIA's foray into high performance computing.

"Our action today is difficult, but necessary considering current business realities. Despite our reduction, we will continue to invest in selective high-growth opportunities like our revolutionary CUDA parallel computing technology and our Tegra mobile single-chip computer," said Jen-Hsun Huang, president and CEO of NVIDIA. "We are taking fast action to enhance our competitive position and restore our financial performance. All of us at Nvidia are determined to emerge from these challenges an even stronger company."

The move will also boost what have recently been sagging financials. The company took a financial hit in Q2 when it was forced to replace bad chips used in HP and Dell notebooks. This, to the tune of $196 million. Ouch!

For more info on NVIDIA's cutbacks, read the full article here.

-----

John West is part of the team that summarizes the headlines in HPC news every day at insideHPC.com. You can contact him at john@insidehpc.com.