Slashdot Mirror
Simulating Supernovae with Graphics Cards
astroboy writes "As graphics cards get more powerful, Los Alamos and Utah scientists have developed a package, Scout, to use those usually-languishing FLOPs to do simulations, and to visualize of them on the on the run. As an example, they have released
movie of part of the evolution of a core-collapse supernovae"
BitTorrent for the movie, in case of Slashdotting: here
quidquid latine dictum sit altum videtur.
the PDF http://xbmodder.us/scout.pdf the torrent for the .mov:
http://xbmodder.us/Scout.mov.torrent
The torrent for avi (divx4)
http://xbmodder.us/Scout.avi.torrent
Usually, I think that New Scientist is pretty accurate as far as laymen-science articles go, but they've let a big mistake slip be.
From the article:
"The Scout programming language, developed at Los Alamos National Laboratory (LANL) in California, US, lets scientists run complex calculations on a computer's graphics processing unit (GPU) instead of its central processing unit (CPU).
Los Alamos National Labs (LANL) is based in (fittingly) Los Alamos, New Mexico. it is currently operated by the University of California, which has contracted for the ability to manage the lab. This may have caused the confusion.
Also, Lawrence Livermore National Labs (LLNL) is based in Northern California, so that may have caused the confusion as well.
Not a terribly serious concern, but their fact's should be straight. The lab is not in California, it is in New Mexico... Editors: shame on you!
The experiment isn't done to show us what supernovea look like, it's more like they've discovered that the gpu from their graphics cards are better suited to the types of operations required to define a supernova than a traditional cpu. The video is just a shiny bonus.
NVIDIA's GPUs are only one or two bits short of perfect 32 bit IEEE floats. (ATI's are still at 24 bit floats.)
See Karl Hillesland and Anselmo Lastra's cool work on measuring this error on current GPUs, GPU Floating-Point Paranoia for much more information.
-matt
I believe this is the Mike Batt version.
ZEN is a prime number in base-36
Err, what? All graphics cards that implement ARB_color_buffer_float has to implement IEEE 32-bit floats, as stipulated by ARB extension specification. (of course, this is assuming that the scientists are using the color buffer to encode information)R B/color_buffer_float.txt
http://oss.sgi.com/projects/ogl-sample/registry/A
Basically, any up-to-date ATi or NVidia gfx cards are capable of true IEEE 32-bit floating point numbers. What really worries me about the research is that they're not using 64-bit!
I tried... When I OC too high it just locks the computer up and scrambles the onscreen fonts... For a real show turn off the lights, voltmod a card, remove the hsf during HL2, and pour saltwater on it when it starts smoking. ;)
It's been found that GPUs, despite their impressive floating point capabilities, can't compare to heavily-optimized and cache coherent CPU implementations of large matrix operations, such as ATLAS. The exception is when the result is to be displayed anyway, as in scientific visualization and Scout. The real drawback of GPUs is the readback speeds. When the result is done, if it isn't to be displayed, it must be read back into the CPU memory. This is notoriously inefficient. PCIe is improving this, but it's still a serious problem.