At the "AI Everywhere" event, Intel launched the 5th-gen Xeon processor, code-named Emerald Rapids, and Intel Core Ultra mobile processors to compete in an AI era. Intel also showcased the Gaudi3 AI accelerator.
As outlined in Intel's press release, the 5th-gen Xeon showcases an impressive 21% average surge in general computing performance and facilitates a remarkable 36% rise in average performance per watt. For clientele adhering to a typical five-year refresh cycle and upgrading from earlier generations, the potential to slash their total cost of ownership by up to 77% emerges as a compelling proposition.
The 5th-gen Xeon boasts support for up to 64 cores per CPU and nearly triple the maximum last-level cache compared to the prior generation. Additionally, it provides eight channels of DDR5 per CPU, backing DDR5 at speeds of up to 5,600 megatransfers per second (MT/s) while elevating inter-socket bandwidth through Intel UPI 2.0, delivering speeds of up to 20 gigatransfers per second (GT/s).
Meanwhile, made on Intel 4 process technology and composed of CPU, GPU, and NPU, which was branded as Intel AI Boost, the Intel Core Ultra presents an integrated Intel Arc GPU3, equipped with a maximum of eight Xe-cores, AI-driven Xe Super Sampling (XeSS), support for DX12 Ultimate, and the potential to deliver twice the graphics performance compared to its predecessor. Intel Core Ultra is the most AI-capable and power-efficient client processor in Intel's history, said the chipmaker.
Gelsinger also shared an update on Intel Gaudi3, which is set to debut in 2024. Intel said its Gaudi pipeline, an AI accelerator for deep learning and large-scale generative AI models, is expanding rapidly and is expected to compete with rival products from Nvidia and AMD.
In collaboration with OEMs, including Cisco, Dell, HPE, IEIT Systems, Lenovo, and Super Micro Computer, Intel will begin to ship the 5th gen Xeon in the first quarter of 2024. On the other hand, Intel said its Intel Core Ultra will power 230 of the world's first AI PCs offered by Acer, Asus, Dell, Chromebook, etc.
Intel CEO Pat Gelsinger said at the event that the industry is still in the very early stage of an AI era. He said Intel is trying to drive AI Everywhere into every aspect of applications and every device in the data center, the cloud, the edge, and the PC as well.
Gelsinger said the world is moving from high-end training to inference everywhere, with a few people creating weather models and many people using them. He highlighted three reasons behind integrating inference capabilities onto the device, including the law of economics favoring device-based processing over rented cloud services, the law of physics where the practicality of immediate local execution versus round trips to the cloud, and the law of the land in which Intel will operate models locally on the edge and device within every geographic area where Intel holds a presence.
