Nervana accelerators are aimed at large data centers and cloud environments
At the Hot Chips 2019 show this week, Intel introduced new processors aimed at accelerating both deep learning training and inferencing workloads in large enterprise data centers and cloud environments.
The Lowdown: The two high-performance artificial intelligence (AI) accelerators will be part of Intel’s larger Nervana Neural Network Processor (NNP) portfolio and are aimed at helping corporations and other large organizations more quickly process, analyze, and gain insights from huge amounts of data and then make business decisions based on those insights.
The Details: The chips will enable organizations to address both the deep learning training and inference aspects of networks using AI models. Both are designed to accelerate large and complex workloads running on Xeon-based systems. The Nervana NNP-T (code-named “Spring Crest”) is aimed at running deep learning training workloads – where neural networks are taught such things as object identification – using minimal amounts of power while striking a balance between computing, communication, and memory.
The Nervana NNP-1 (code-named “Spring Hill”) was built for inference – where what is learned through training is put into use – with high levels of programmability, performance, and power efficiency. The chips are based on Intel’s 10-nanometer “Ice Lake” Core processors and are the fruits of investments the chipmaker has made in Israeli AI start-ups like Habana Labs and NeuroBlade. Facebook reportedly already is using the chips.
The Impact: In modern data centers and cloud environments, it’s the data – and the ability to draw actionable information from that data as quickly as possible – that matters most, and using AI, machine learning, and deep learning techniques has become the best way to do this.
Background: Intel is not alone in developing chips specifically for machine learning workloads. Google has its Tensor Processing Units (TPUs), Nvida has NVDLA (Nvidia Deep Learning Accelerator), and Amazon Web Services (AWS) has its Inferentia inference chip. In addition, start-up Cerebras Systems this week unveiled a huge chip with 1.2 trillion transistors aimed at AI applications.
The Buzz: “To get to a future state of ‘AI everywhere,’ we’ll need to address the crush of data being generated and ensure enterprises are empowered to make efficient use of their data, processing it where it’s collected when it makes sense and making smarter use of their upstream resources,” said Naveen Rao, vice president and general manager of Intel’s Artificial Intelligence Products Group. “Data centers and the cloud need to have access to performant and scalable general-purpose computing and specialized acceleration for complex AI applications. In this future vision of AI everywhere, a holistic approach is needed – from hardware to software to applications.”