Reading Assignment II: NVIDIA Volta and Turing

Computer architectures and in particular accelerator architectures are evolving to support more efficiently AI-type and ray-tracing workloads.

In particular, deep-learning applications require the calculation of several matrix-matrix (also called tensor) multiplications and additions during the neural network training and inference. 

To give you a little bit of material on these topics, we are uploading some materials about the new NVIDIA Volta GPUs and tensor corse we suggest you to read the paper:

TensorCorePaper.pdf Download TensorCorePaper.pdf

You can find some information about tensor cores and how to program them. The supercomputer in Umeå HPC center can provide access to Volta GPUs!

Very recently, the new Nvidia Turing architecture have been introduced. Together with tensor cores, Turing GPUs feature RT (Ray Tracing) cores. You can find more information here Links to an external site..