Shulou(Shulou.com)11/24 Report--

On July 11, Intel AI product strategy and new Gaudi2 product conference was held, at which Intel officially launched the second generation Gaudi deep learning accelerator-Habana Gaudi2 for the Chinese market. As an important part of Intel's cloud-to-end product portfolio, Gaudi2 is committed to accelerating AI training and reasoning with leading performance-to-price advantages. The launch of this product has also brought new choices to the hot large model market recently.

Gaudi2 Deep Learning Accelerator and Gaudi2 mezzanine card HL-225B is based on the first-generation Gaudi high-performance architecture to improve computing performance and energy efficiency in all aspects. According to reports, Gaudi2 has 24 programmable Tensor processor cores, 21 100Gbps (RoCEv2) Ethernet interfaces, 96GB HBM2E memory capacity, 2.4TB / s total memory bandwidth, 48MB on-chip SRAM and integrated multimedia processing engine.

In terms of performance, the Gaudi2 Accelerator was certified in the MLCommons MLPerf benchmark released in June. According to MLPerf's test results, Gaudi2 has excellent results in several tests such as the GPT-3 model, the computer vision model ResNet-50 (using 8 accelerators), Unet3D (using 8 accelerators), and the natural language processing model BERT (using 8 and 64 accelerators). At present, it is claimed that the computing requirements of AI and LLM need to be expanded on a large scale. The test of MLPerf reflects the advantages of Gaudi2 in scalability, energy efficiency and ease of use.

In addition, Gaudi2 also supports Intel's SynapseAI software suite for deep learning, training and reasoning optimization on the Gaudi platform, which integrates PyTorch, TensorFlow, and DeepSpeed frameworks; supports Kubernetes choreography; customizes compilers, and has the support of Hugging Face, PyTorch Lightning, RedHat and other software partners.

In recent Hugging Face assessments, Gaudi2's performance in large-scale reasoning includes staying ahead of the industry when running Stable Diffusion, 7 billion, and 176 billion parameter BLOOMz models.

For the Chinese market, Intel and tide information ensemble to create and sell the wave information AI server NF5698G7 based on Gaudi2 deep learning accelerator, which integrates 8 Gaudi2 acceleration cards HL-225B and includes dual fourth-generation Intel Xeon scalable processors.

Facing China's vast AI market, Sandra Rivera, Intel Executive Vice President and General Manager of data Center and artificial Intelligence, said, "Intel is committed to accelerating the development of artificial intelligence technology by providing customers with a wide range of hardware choices and supporting an open software environment. With a portfolio of Xeon scalable processors and Gaudi2 Deep Learning Accelerator, Intel is reducing barriers to entry for artificial intelligence and strengthening customers' ability to deploy this critical business technology in the cloud through the network and intelligent edge, thereby helping to build the future of artificial intelligence in China. "

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.