Saturday, August 30, 2008

CPU vs GPU -- totally misleading, but totally AWESOME


Mythbusters’ GPU vs. CPU demo: 2100 paintballs make the Mona Lisa in 275ms

Well, this really is awfully misleading, but it sure is awesome. The idea is that a normal CPU processes things one at a time, like the single-paintball robot, and a GPU has parallel processors that can do (in this case) 2100 times the work in a single stroke. Well, considering CPUs are moving to serious multi-core architecture and GPUs are being used to do CPU work, the whole demonstration is a little dubious. Still, how filthy was that? [via TGDaily]


Mythbusters duo launches new GeForce, codenamed Mona Lisa PDF Print E-mail
By Theo Valich
Thursday, August 28, 2008 16:03
vote nowBuzz up!
San Jose (CA) – Even though Nvidia’s Nvision tradeshow did not achieve its goal of 10,000 visitors, more than a thousand gathered at the Center for Performing Arts to witness the ending of the event. The duo behind the popular Mythbusters showed the results of six months of work, demonstrating the difference between a CPU and a GPU, following the conventional wisdom of parallel computing. Dubbed Smiley and Mona Lisa, Adam Savage and Jamie Hyneman presented two robots that represented the difference between a CPU and a GPU. Smiley was given a task to draw a smiley using conventional CPU techniques, doing one thing at a time. Smiley was a relatively simple robot, while Mona Lisa consisted out of "1100 massively parallel barrel processors", dwarfing the 240 shaders offered by a GeForce GTX 280 chip. These 1100 "shaders" painted a picture of Mona Lisa in a matter of a few hundreds of a second, resulting in lots of stunned faces, laughter and a big round of applause for the Mythbusters crew. We won't dwell into the specs of these two, but Mona Lisa's fill-rate is quite interesting: 1100 (paintball) balls travelled eight feet onto wood-enforced canvas in just in just 275 ms, creating a picture of Mona Lisa within three quarters of a second. We guess you could say Nvision ended with a bang (sorry, we couldn’t resist.) There was no information if and when these barrel processors will be available for sale.