Cooperation with Nvidia, IBM and Google: Intel announced yesterday (Tuesday) the launch of its two new flagship products for the data center and artificial intelligence market: Xeon 6 processors that include the new extremely powerful processing cores (Performance Core), known by the code name "Granite Rapids; and the AI accelerator, Gaudi 3, developed in Israel by Habana Labs. These products bring significant performance improvements for artificial intelligence (AI) and high-performance computing (HPC) workloads, and are considered an important strategic step in restoring Intel's position as a technology leader in the data market chin.
Among the main partners that will use the new processors introduced by Intel, you can find Google, IBM, Dell and even the competitor Nvidia. Google Cloud will integrate Xeon 6 processors into its cloud services, while IBM will be the first to offer Gaudi 3-based cloud services within IBM Cloud. Dell will allow its customers to use both Xeon 6 processors and Gaudi 3 accelerators for advanced AI applications, and Nvidia will integrate Xeon 6 processors as a host platform (Host CPU) in its DGX systems, which are used for intensive computing and machine learning purposes.
Intel is especially proud to announce that four major computer manufacturers: Lenovo, Dell, and Supermicro will sell the Gaudi 3 accelerators to their customers in the market. At the same time, the number of customers of Bana Labs continues to grow, and among them leading companies such as: IBM, Bosch, Roboflow, Seekr, NAVER, Landing AI, NielsenIQ and Bharti Airtel use Gaudi 2 technology in their artificial intelligence systems.
Xeon 6 processors show an impressive 5.5x improvement in AI performance compared to market competitors, and in HPC high-performance computing up to 2.1x. The processors include key technologies developed in Israel, such as: Intel AMX, which enables the acceleration of AI operations directly in the processor and Intel TDX, which provides increased security for virtual machines in the cloud and in data centers.
The Israelis also led the development of the powerful performance core, the P-core, in Intel's development centers in Israel."
Another Israeli angle is found in the Gaudi 3 accelerators, developed by the Bana Labs teams and now officially launched. These accelerators provide twice the computing power for AI and twice the network bandwidth compared to the previous generation. In addition, they offer a 1.8 times better cost-benefit ratio compared to the NVIDIA H100 when working with models
Like LLaMA 2 and LLaMA 3. It should be noted that the performance of Gaudi 3 is expected to improve as new versions of the software, such as the upcoming Synapse AI 1.18, arrive and allow improved performance in training and model processing.
Another advantage of Gaudi 3 is the use of standard Ethernet communication, which allows maximum flexibility in combination with existing infrastructures and lower costs compared to dedicated communication solutions. Also, Gaudi 3 offers simple expansion capabilities and broad support for tools and software, which make management and maintenance easier.
Intel even presented for the first time the future data center processor, Clearwater Forest, which will be based on advanced manufacturing technology 18A. This is the first processor in the Xeon family to use this technology, and it guarantees full compatibility with existing infrastructures and further improvement in performance and energy efficiency.
Justin Hotard, Senior Vice President and CEO of the Data Center and AI Group at Intel, said: "With Xeon 6 and Gaudi 3, we provide our customers with the most advanced tools for developing and using AI solutions. Our partnership with leading companies such as: Google , IBM and Nvidia emphasizes our commitment to provide flexible, efficient and economical systems in energy consumption, adapted to the various needs of the industry."
Xeon 6 processors are available now to select customers, and wider availability is expected in the coming quarters. The Gaudi 3 accelerators will be available to the general public between the third quarter of 2024 and the first quarter of 2025.