The Xeon 6 processors supply a 4x efficiency enchancment and practically 3x higher rack density in contrast with second-generation Intel Xeon processors, Intel acknowledged.
Taking goal at Nvidia and focusing on massive AI processing wants, Intel introduced the Gaudi 3 AI accelerator chip, which it says is 40% on common extra energy environment friendly than comparable Nvidia H100 chips. (Pictured above is Intel CEO Pat Gelsinger unveiling the Gaudi 3 accelerator on the Intel Imaginative and prescient occasion in Phoenix, Ariz.)
“The Intel Gaudi 3 AI accelerator will energy AI methods with as much as tens of hundreds of accelerators linked by way of the widespread normal of Ethernet,” Intel acknowledged. For instance, 24 200-gigabit Ethernet ports are built-in into each Intel Gaudi 3 accelerator, offering versatile and open-standard networking.
Intel Gaudi 3 guarantees 4x extra AI compute and a 1.5x improve in reminiscence bandwidth over its predecessor, the Gaudi 2, to permit environment friendly scaling to help massive compute clusters and eradicate vendor lock-in from proprietary networking materials, Intel acknowledged.
The concept is that the accelerator can ship a leap in efficiency for AI coaching and inference fashions, giving enterprises a alternative in what methods they deploy when taking generative AI to scale, Katti stated.
The Intel Gaudi 3 accelerator will probably be obtainable to authentic tools producers within the second quarter of 2024 in industry-standard configurations of Common Baseboard and open accelerator module (OAM). Dell Applied sciences, Hewlett Packard Enterprise, Lenovo and Supermicro are among the many distributors that may implement Gaudi 3 in servers and different {hardware}. Common availability of Intel Gaudi 3 accelerators is ready for the third quarter of 2024.
