|
Friday July 31, 2009
|
It may seem that the age of the supercomputer a la Cray-1 has long passed. But that's not entirely true--we just don't hear about them as often. For example, scientists at the U.S. Department of Energy's Argonne National Laboratory are visualizing supernova, protein structures, and other complex phenomena using supercomputers.
The scientists there are using a technique called software-based parallel volume rendering, which interprets the billions of data points collected from MRIs, X-rays, or research simulations. They're also working on equations that could search for sudden density changes--for example, separating bone from muscle data--in order to generate complex visualizations.
Using parallel computing, such as with Argonne's Blue Gene/P supercomputer, scientists can create images using the computer's 160,000 cores. (Try that on a Core i7.) The above image is a rendering of a supernova--specifically, the mechanisms behind a star's violent collapse, with different colors and transparencies depicting different values of entropy. (Image credit: ANL)
|
|
|
Mixx It
Digg It
Stumble
More...
July 31, 2009 5:30 PM
>Try that on a Core i7
I would love to see comparison Argonne's Blue Gene/P supercomputer volume rendering performance with Fovia HDVR running on dual W5580. Really, if you are confident to go public with results you are welcome. You may contact me stefanbanev at yahoo.com and I may assist to get in contact with Fovia or you may go directly; I'm sure they would love to take the challenge.
Good luck,
Stefan
August 2, 2009 2:53 PM
>"We were able to scale up to large problem sizes
>of over 80 billion voxels per time step and
>generated images up to 16 megapixels,"
It is ~4K cube data and 4K projection plane rendered by Argonne's Blue Gene/P 160000 computing cores... well, it is fine if this "accomplishment" is founded through private founds and it is huge waste if it is paid by tax money. You may do this job interactively on 4 sockets (16 x i-7 cores) machine to spend tax money more efficiently and likely get better performance, you should ask help from private business folks who knows how to do job cost-effectively.
August 2, 2009 6:16 PM
I would love to see the Abermarle727 postcompted on a JR7998. The intergalatic peripheral would be amazing! This could then be interspersed with a photo-synthetic hyperbole which would cause a magnetic renasonce not seen since Mega-Warp II.
August 3, 2009 10:52 PM
@homorenderus - Large projects like this are usually public/private partnerships. And I think you're jumping to conclusions based on political beliefs rather than fact - I think if you investigate the technology behind some of these supercomputers, you'll understand that the sheer number of and clock speed of these supercomputers is not always a complete indicator of their performance, when compared to the way they interact and are programmed to process data.
August 4, 2009 3:21 PM
>if you investigate the technology
>behind some of these supercomputers,
Are you kidding, what technology, how to compensate lack of neural tissue by number of cores? Look, there is 4K cube of voxel data to be rendered interactively at high quality. It is very specific and concrete procedure and there is technology to do it with 16 cores so, to do it with 160000 is a total waste; I see no problem to spend dumb private money for that, it is actually a healthy redistribution of wealth but I'm definitely not drilled if any of my tax money wasted for such "accomplishments" .