Drilling Down Into Nvidia’s “Pascal” GPU

April 19, 2016 Timothy Prickett Morgan Nvidia made a lot of big bets to bring its “Pascal” GP100 GPU to market and its first implementation of the GPU is aimed at its Tesla P100 accelerator for radically improving the performance of massively parallel workloads like scientific simulations and machine learning algorithms. The Pascal GPU is a beast, in all of the good senses of that word, and warrants a deep dive that was not possible on announcement day back at the GPU Technology Conference. We did an overview of the chip back on announcement day, as well as talking about Nvidia’s own DGX-1 hybrid server platform, which pairs two current Intel Xeon processors with right of the Pascal-based Tesla cards to create a system that delivers 170 teraflops of half…


Link to Full Article: Drilling Down Into Nvidia’s “Pascal” GPU