Hybrid Computing Updates
by Michael Hay on Jun 10, 2010
In case you missed it the second fastest super computer in the world is based on hybrid compute technology. It incorporates Intel and NVidia processors to achieve extreme scale and performance. If you look at the number of cores, 120640, you will see that it has a little more than half of the number as the fastest super computer, the Cray Kraken at 224162.
In other news AMD recently announced the Fusion line of APUs, adding the GPU block directly onto the die to create a true many core processor. Since AMD, like NVidia, is being used for GPGPU, I expect that an iteration of these new processors will begin to infiltrate both business and academic/research high performance computing. Also this makes me ponder what will happen to Cray’s next generation supercomputer architecture if they adopt a future version of the AMD fusion processor?
It clear to me that the trend of the many core compute platform continues its influence on IT!
Comments (2 )
Make’s you wonder what sort of IO/s you could get with a vanilla Linux server using a GPGPU to do the RAID calcs in a software RAID setup. Does Linux LVM support GPGPU yet?
Mike, I’m not sure that there is yet support in the storage stack for GPGPU. There is a very intriguing whitepaper from NVida which talks about the potential for applying GPUs for extremely high performance virus scanning.