Intel unveils new AI-focused Nervana processors in collaboration with Facebook

The Intel Nervana Neural Processor family are specially designed for matrix multiplication and convolutions, commonly required for deep learning processing.

Published Date
18 - Oct - 2017
| Last Updated
18 - Oct - 2017
 
Intel unveils new AI-focused Nervana processors in collaboration...

Intel unveiled a new family of chips yesterday, meant for AI processing. Known as the Intel Nervana Neural Processor family (NNP), the chips are expected to ship by the end of 2017. The company also worked with Facebook, taking its insights for chip design. The social media company has been one of the most prolific users of AI over the past year or so. Intel’s target market also includes the automotive industry, healthcare and more. “We are thrilled to have Facebook in close collaboration sharing their technical insights as we bring this new generation of AI hardware to market,” said Brian Krzanich, CEO of Intel.

It’s unlikely that Intel will sell this chips for PC building. Instead, they should be headed to data centers, taking on GPUs from Nvidia and Google’s Tensor Processing Units (TPUs). The new NNP family brings an application-specific integrated circuit (ASIC), which is specially designed for artificial intelligence processing.

While these are Intel’s first AI chips, the company has been amongst the most vocal firms in the AI space. It has teased these NNPs earlier, while making multiple purchases aimed at strengthening its AI division. While the NNP chips are obviously a result of its acquisition of chipmaker Nervana earlier, Intel had also acquired Movidius, the firm that made vision processing units (VPUs) running inside the DJI Phantom drones. The NNP chips were earlier codenamed “Lake Crest”.

The new Nervana chips are designed for matrix multiplication and convolutions, which are common for deep learning programs. Interestingly, while Intel’s Nervana chips are specially designed for AI, it’s unclear how they stack up against competing platforms like Google’s TPUs. Intel had earlier said that it intends to improve deep learning training speeds by 100 times, by 2020.

Further, Intel also announced investments in “frontier technologies that will be needed for other large-scale computing applications of the future”. The company is investing in neuromorphic processing, which refers to chips that are inspired by the human brand. These are self-learning chips that will allow computers to recognise patterns and make decisions based on the same. Intel says the potential benefits of such chips are “limitless”.

Alongside Neuromorphic chips, Intel is also working on quantum computing, with QuTech, it’s partner for the project.

Prasid BanerjeePrasid Banerjee

Trying to explain technology to my parents. Failing miserably.