Intel unveils AI chips for training and inference
Intel revealed new details of upcoming high-performance artificial intelligence (AI) accelerators: Intel Nervana neural network processors, with the NNP-T for training and the NNP-I for inference. Training is how deep learning applications are programmed by feeding them more input and tuning them. Inference is how they run, to perform analysis or make decisions.
“Data centers and the cloud need to have access to performant and scalable general purpose computing and specialized acceleration for complex AI applications,” said Naveen Rao, vice president and general manager for Intel’s Artificial Intelligence Products Group, in a statement.
Turning data into information and then into knowledge requires hardware architectures and complementary packaging, memory, storage and interconnect technologies that can evolve and support emerging and increasingly complex use cases and AI techniques. Dedicated accelerators like the Intel Nervana NNPs are built from the ground up, with a focus on AI.
The AI training chip is built from the ground up to train deep learning models at scale, and to prioritize two real-world considerations: training a network as fast as possible and doing it within a given power budget. The AI inference chip is for high-performing deep learning inference for major data center workloads, which Intel said is easy to program, has short latencies and fast code porting and includes support for all major deep learning frameworks.
Cerebras Systems unveils “Trillion Transistor Chip”
Cerebras Systems, a startup dedicated to accelerating Artificial intelligence (AI) compute, unveiled the largest chip ever built. Optimized for AI work, the Cerebras Wafer Scale Engine (WSE) is a single chip that contains more than 1.2 trillion transistors and is 46,225 square millimeters. The WSE contains 3,000 times more high speed, on-chip memory, and has 10,000 times more memory bandwidth.
In AI, chip size is important. Big chips process information quicker, producing answers in less time. Reducing the time-to-insight, or “training time,” allows researchers to test more ideas, use more data, and solve new problems. Google, Facebook, OpenAI, Tencent, Baidu, and many others argue that the fundamental limitation to today’s AI is that it takes too long to train models. Reducing training time removes a major bottleneck to industry-wide progress.
Huawei launches AI processor and computing framework
Huawei officially launched the Ascend 910, which the company describes as “the world’s most powerful AI processor” as well as “an all-scenario AI computing framework”, MindSpore.
The Ascend 910 is a new AI processor used for model training. After a year of ongoing development, test results now show that the Ascend 910 processor performs with much lower power consumption than originally planned. In a typical training session, the combination of Ascend 910 and MindSpore is about two times faster at training AI models than other mainstream training cards using TensorFlow.
Moving forward, Huawei will continue investing in AI processors to deliver more abundant, affordable, and adaptable computing power that meets the needs of a broad range of scenarios. In 2018, Huawei announced the three development goals for its AI framework: dramatically reduce training time and costs; use the least amount of resources with the highest possible OPS/W; and scenario adaptability including device, edge, and cloud applications.
The company also noted that MindSpore helps ensure user privacy because it only deals with gradient and model information that has already been processed. It doesn’t process the data itself, so private user data can be effectively protected even in cross-scenario environments. In addition, MindSpore has built-in model protection technology to ensure that models are secure and trustworthy.