It seems we're having a week of super power and super computing here at CCL, first had the new flag ship i7 from intel, the hot new FirePro from AMD, Intels new Xeon super computer chip and now we have the all new Tesla K20-series graphics processing unit.

The new GPU based accelerators are for specific high-performance computing applications and were announced at the SC12 supercomputer conference this week. NVIDIA's new power horse is based on the GK110 processors and is set to stun the HPC world.

Tesla K20X
2688 stream processors
6GB of GDDR5 memory
3.95TFLOPS single-precision
1.31TFLOPS double-precision

Tesla K20
2496 stream processors
5GB of GDDR5 memory
3.52TFLOPS single-precision
1.17TFLOPS double-precision

Not only are these more powerful than the last generation but they're also a massive three times more energy efficient, which is obviously impressive for so many reasons. Much like the new Intel hardware we saw this week, the K20 is perfect for things like CFD, CAE, chemistry and physics calculations, imaging and weather modelling and it's capable of doing a damn fine job of it all too.

Both the K20X and K20 are already in the hands of a few super computer clients around the world, enough to perform 30PFLOPS of processing in fact, which is only equivalent to all the computational performance of the 10 fastest super computers combined for the last year... wow!

Enjoyed this article? share it!