So Nvidia made a lot of major announcements today during their first keynote of their GPU Technology Conference in San Jose, California. As always, their CEO, Jen-Hsun Huang, led the entire keynote and introduced a multitude of new technologies across the whole breadth of Nvidia’s product portfolios. The most notable announcements came in the Pascal GPU technology, GeForce GTX Titan Z and Erista, the successor of Maxwell.
He kicked off the presentation by talking about how GPU performance has improved over the years compared to their partners, the CPU.
He also talked about the expansion of GTC as a result of the growth of GPU, with over 3,500 attendees and nearly 729 different sessions, with a focus on Big Data Analytics, Machine Learning and Computer Vision.
So, the first announcement that Jen-Hsun announced was Nvidia’s own interconnect technology for GPUs called NVLINK which helps improve the performance and latency between GPUs.
This leads to the obvious progression of a new GPU that will utilize this technology. Then he started talking about Blaise Pascal…
This leads to the announcement of the Pascal GPU, which utilizes NVLink, 3D memory and is a module 1/3 the size of a full-size PCIe graphics card.
Which adds this to Nvidia’s new GPU roadmap, do note that Pascal is slated for 2016, and it pushes Volta out of the roadmap, which is curious.
After that, Jen-Hsun went over some machine learning slides and demonstrations that explained the relevance in GPU computing in helping simulate human neural networks. Following that, Jen-Hsun started to talk about Titan and a potential successor.
A Dual Titan GPU called the GeForce GTX Titan Z, with 5,760 CUDA cores, 12GB of RAM, 8 TeraFLOPS all for the low low price of $2,999.
Then, a self-driving Audi drove itself on stage with Jen-Hsun and he started to talk about how Nvidia’s Tegra K1 has been responsible for powering the self-driving car.
Last, but not least, he also decided to pull an Oprah and announced that everyone in the audience would be getting an Nvidia SHIELD.