Table of Contents 94e1i
A Intel revealed, this Tuesday (9), the Intel Gaudi 3, your youngest AI accelerator aimed at large-scale AI computing. Manufactured using a 5 nanometer process, the launch offers better energy efficiency and more processing power, compared to the previous generation (Gaudi 2). 6q1a7
Intel Gaudi 3 2c2jq
According to company, who presented the Intel Gaudi 3 during the event Intel Vision 2024, the new lineup delivers up to 4x better computational AI performance than the previous generation, plus a 1,5x increase in memory bandwidth and a 2x increase in network bandwidth. With the increase, the company's solution emerges as one of the few alternatives on the market for large-scale Generative AI infrastructure.
Intel Gaudi 3 stands out as the GenAI alternative that presents a compelling combination of price performance, system scalability, and time-to-value advantage.
Justin Hotard, Intel Executive Vice President and General Manager for Data Center and AI Group
the chips Gaudí 3 features 64 Tensor Processing Cores (TPCs) customized for AI and are fully programmable, with 8 Matrix Multiplication Engines (MME), which results in excellent performance in heterogeneous computational processing. Each MME coprocessor is capable of executing 64 thousand parallel operations, which guarantees the ability to deal with complex matrix operations, something essential for Deep Learning algorithms.
Intel also highlights that the accelerator Intel Gaudi 3 promises to serve companies from the most diverse areas, such as finance, manufacturing and healthcare, that are seeking to quickly expand accessibility to AI and the transition of generative AI projects, from experimental phases to full-scale implementation.
The company ensures that the solution will meet these requirements and will also offer versatility through community-based open software and industry-standard open Ethernet, helping companies flexibly scale their AI systems and applications.
Among the main highlights of the Intel Gaudi 3 still are:
- AI-dedicated compute engine: The solution allows for a high degree of computational efficiency, making them adept at handling complex matrix operations, a type of computation that is fundamental to deep learning algorithms. This unique design accelerates the speed and efficiency of parallel AI operations and s multiple data types, including FP8 and BF16;
- Memory Boost for LLM Capacity Requirements: 128 gigabytes (GB) of HBMe2 memory capacity, 3,7 terabytes (TB) of memory bandwidth, and 96 megabytes (MB) of integrated static random access memory (SRAM) provide ample memory for processing large GenAI datasets in fewer Intel Gaudi 3s, particularly useful for serving large language and multimodal models, resulting in increased workload performance and data center cost efficiency;
- Efficient system scaling for Enterprise GenAI: Twenty-four 200 gigabit (Gb) Ethernet ports are integrated into each Intel Gaudi 3 accelerator, providing flexible, open-standard networking;
- Open Industry Software for Developer Productivity: Intel Gaudi software integrates the PyTorch framework and offers optimized models based on the hugging face community, the most common AI framework for GenAI developers today. This allows GenAI developers to operate at a high level of abstraction for ease of use and productivity and ease of portability of models across different types of hardware.
Fight with AMD and NVIDIA for the AI market 1zb72
The launch of Intel intensifies the dispute with other manufacturers, such as AMD e NVIDIA, the latter which even recently announced the H200, Ryzen Embedded 8000 U for embedded systems for industrial applications.
Currently, many AI tools are already used in industry, mainly in the field of computer vision and Big Data processing. With this, together, manufacturers are already preparing for a new computing generation, where AI promises to dominate even more spaces in the most diverse areas.
Availability 5683v
The accelerator Intel Gaudi 3 will be available to original equipment manufacturers (OEMs) in the second quarter of 2024 in standard configurations. Among the notable partnerships are Dell Techonolgies, HPE, Lenovo and Supermicro, with solutions using the new accelerators arriving on the market in the third quarter.
See also:
Qualcomm RB3 Gen 2 is launched with 10x more AI performance. Check out all the launch details.
reviewed by Glaucon Vital in 9 / 4 / 24.
