My Thoughts on Technology and Jamaica: University of Texas, TACC and Intel to make fastest Supercomputer - 50 Core Processor is Just Swaggin

Friday, September 23, 2011

University of Texas, TACC and Intel to make fastest Supercomputer - 50 Core Processor is Just Swaggin

Yond Cassius has a lean and hungry look;
He thinks too much: such men are dangerous

Shakespeare, Julius Caesar, I, 2

Looks like the news for Intel keeps getting better and better, folks. More and more 2012AD is indeed looking like the Year of Intel and Microsoft as stated in my blog article entitled “Standardizing the Cloud Printer for Tablets and smartphones - Molly Wood rants about the broken printer”. News has reached the ears of this plucky blogger that suggests that Intel is teaming up with TACC (Texas Advanced Computing Center) to build a supercomputer.

What’s unique about this design is that it will be using their as-yet-not-commercially-available 50-core GPU (Graphic Processing Unit) Processor based on the 22nm and 3D Ivy Bridge Processors codenamed “Knights Corner” as stated in the article “Intel 50-core chips headed to Texas supercomputer”, published SEPTEMBER 22, 2011 2:44 PM PDT by Brooke Crothers, CNET News – Nanotech.

This coincides with the research of IBM and 3M into developing an adhesive to glue Processors together in a Vertical Processor Array or Iso-Linear Processor design, which like the above mentioned supercomputer, is also slated to be completed in 2013AD as stated in my blog article entitled “IBM and 3M to develope adhesive for Iso-Linear Processor - Rihanna's Umbrella for Intel's SuperComputer on a Chip”.

Omar Ghattas, who wields the record for the longest title of Director of the Center for Computational Geosciences in the Institute for Computational Engineering and Sciences at the University of Texas at Austin, gushes ecstatic on this supercomputer project's potential, quote: "My group...is excited about the opportunities Stampede offers to greatly accelerate our work in quantifying uncertainties in computer models of dynamics of polar ice sheets, global seismic wave propagation, and whole-earth plate tectonics,"

The specs of this dream Stampede supercomputer machine are as follows:

1.     Several Thousand Dell Zeus Servers, each running two (2) 8-Core Intel Xenon Sandy Bridge Processors
2.     32 gigabytes of RAM memory.
3.     Several hundred Servers running Knight's Corner 50-Core Intel 32nm Ivy Bridge Processors

But the essence of the excitement that seems to have been glossed over in this report is the fact that this supercomputer, a joint project between TACC, University of Texas and Intel codenamed Stampede (‘cause it’s Texas y’all) is that the speed that they are aiming for is in the 10 petaflop region.

Say what!!

That’s right folks, you heard me correct, as I didn’t misspeak. Their finished Stampede 50 core GPU supercomputer will steam roll over China’s “Tianhe-1A” supercomputer, also built using GPU but supplied from Nvidia as described in my blog article entitled “Nvidia and the GPU Supercomputer - Go Speedracer”.

For the record, this is the same Nvidia that has plans to bring Quad-Core Kal-El Processors to tablets and smartphones as stated in my Geezam Blog article entitled “Nvidia and the Quad-Core Kal-El CPU – Android leaves A5 Dual-Core Processor in the Dust”.

China’s “Tianhe-1A” supercomputer is currently the No. 1 Supercomputer since November 2010AD with a speed of 2.57 petaflop/s as per the ratings of the statistics as stated in the article “China grabs supercomputing Leadership spot in latest Ranking of World's Top 500 Supercomputers”, published Thursday, November 11 2010 at 22:42, Top 500.

It bested the U.S. Department of Energy’s (DOE) Oak Ridge Leadership Computing Facility’s Cray XT5 “Jaguar”, located in Tennessee which is now relegated to 1.75 petaflop/s running the Linpack benchmark which is used in determining and rating supercomputer speeds and performance. Just to be precise, a “flop” is a float point operation and represents a mathematical calculation.

In my blog article entitled “Nvidia and the GPU Supercomputer - Go Speedracer”, I echoed the sentiments of NCSA (National Center for Supercomputing Applications) Director Thom Dunning, who effectively all but said that the next fastest supercomputer would be a GPU of a similar build as the Nvidia machine running on a Linux Distribution stated in the article “NCSA Director: GPU is future of supercomputing”, published October 29, 2010 4:31 PM PDT by Brooke Crothers, CNET News Nanotech - The Circuits Blog.

My blog article also pointed out the fact that the next supercomputer could also be built in the Caribbean with the collaboration between the various Universities in Jamaica, as the “Tianhe-1A” supercomputer, built by the People’s Republic of China’s National University of Defense Technology located in Tianjin achieved this using multi-core Nvidia GPU’s running on a Linux Distribution in November of 2010AD as stated in the article “China unseats U.S in supercomputer rankings”, published November 14, 2010 10:00 AM PST by Erica Ogg, CNET News - Circuit Breaker.



So now that TACC, University of Texas and Intel may be debuting a 50-core processor GPU based supercomputer to challenge that title, Speed Racer (2008) Style, it heralds the coming of age of the GPU in supercomputer design as well as yet another exciting chapter in the race to have the fastest Supercomputer title back in the Western Hemisphere. This all as prognosticated in my blog article entitled “Nvidia and the GPU Supercomputer - Go Speedracer”.

The consortium of TACC, University of Texas and Intel, in a world where anyone can now build a supercomputer and supercomputing power of a lower order may be coming to regular consumers via the research of IBM and 3M as stated in my blog article entitled “IBM and 3M to develope adhesive for Iso-Linear Processor - Rihanna's Umbrella for Intel's SuperComputer on a Chip” may be W.T.F (Wayne Marshall. Tifa. Future Fambo) Just Swaggin.

No comments: