If there are any kids within earshot, send them off to play or cover their ears before reading any further, we wouldn’t want their little ears being exposed to excited obscenities that may follow. As in, “Holy sh*t, how much RAM!?” Try 24GB of GDDR5 memory, which is how much Nvidia decided to use on its Tesla K80 dual-GPU accelerator, the new flagship offering in the Tesla family.
For the past two years, Nvidia has launched high-end supercomputing hardware at the International Conference for High Performance Computing, Networking, Storage and Analysis (sensibly abbreviated as SC). Today, at SC14, Nvidia is following that trend, except this time it’s bringing a new dual-GPU product to the table. The new K80 is based on a revised version of the GK110 chip, dubbed GK210, with 13 of Kepler’s 15 SMX’s enabled and a 300W TDP. This may allow the card to hit higher Boost frequencies than the desktop-oriented Titan Z, which struggled to deliver equivalent scaling to two Titan Black’s in an SLI configuration.
