Tech companies working with large volumes of data needing faster and more powerful data processors are under pressure to modernize their ageing data centre systems. In a bid to capitalise on that, and offer their clients more powerful, and cost-effective solutions, Intel has unveiled its latest Xeon 6 processors featuring Efficient-cores (E-cores), which are aimed at delivering superior performance and power efficiency for high-density, scale-out workloads within data centres.
Intel CEO Pat Gelsinger, at the launch in Taipei, said, “Intel is one of the only companies in the world innovating across the full spectrum of the AI market opportunity – from semiconductor manufacturing to PC, network, edge and data centre systems. Our latest Xeon, Gaudi and Core Ultra platforms, combined with the power of our hardware and software ecosystem, is delivering the flexible, secure, sustainable and cost-effective solutions our customers need to maximize the immense opportunities ahead.”
Xeon 6 Platform
Intel’s introduction of the Xeon 6 platform and its family of processors, including both E-core (Efficient-core) and P-core (Performance-core) SKUs, addresses these challenges by catering to a wide range of use cases and workloads, from AI and high-performance computing to scalable cloud-native applications.
These processors offer a unified hardware platform and software stack across the 6700 series and 6900 series platform offerings. They bring significant advancements in double data rate (DDR5), peripheral component interconnect express 5.0 (PCIe 5), ultra path interconnect (UPI), and compute express link (CXL), setting new standards in performance and efficiency.
The Xeon 6 6700 series delivers up to 1.4x higher memory bandwidth and up to 1.1x increase in input/output (I/O) bandwidth compared to the 5th Gen Intel® Xeon® processors, enabling faster and more efficient data handling.
The Xeon 6 6900 series offers up to 1.8x increased inter-socket bandwidth compared to its predecessor, enhancing communication within the system and significantly boosting performance.
Fabricated on the latest strides in silicon architecture, the Xeon 6 processors with E-cores are fine-tuned for dense, scale-out workloads, encompassing cloud-native applications, content delivery networks, network microservices, and consumer digital services. These processors deliver exceptional performance per watt, culminating in efficient computation that curtails energy expenditures and bolsters sustainability objectives.
Support for CXL® 2.0 simplifies connectivity between computers and additional components such as accelerators and memory expanders.
Moreover, Xeon 6 processors with E-cores offer significant rack density advantages, enabling a 3-to-1 rack-level consolidation and freeing up space for innovative new AI projects. Additionally, they are optimized for network and edge workloads, providing up to 2.7x higher 5G-user plane function performance per watt and up to 3.5x higher next-generation firewall performance per watt compared to 2nd Gen Xeon processors. This increased compute efficiency supports confidential computing capabilities for 5G networks and edge-to-cloud infrastructures.
The support for CXL 2.0 streamlines connectivity between computers and supplementary components like accelerators and memory expanders.
Furthermore, Xeon 6 processors with E-cores present notable rack density benefits, enabling a 3-to-1 rack-level consolidation and liberating space for pioneering AI initiatives. Additionally, they are optimized for network and edge workloads, delivering up to 2.7x superior 5G-user plane function performance per watt and up to 3.5x enhanced next-generation firewall performance per watt in contrast to 2nd Gen Xeon processors. This heightened computational efficiency underpins confidential computing capabilities for 5G networks and edge-to-cloud infrastructures.
Gaudi 3 AI Accelarators
Intel has also announced significant advancements in their line of AI accelerators working on generative AI, and LLMs, and launched the Gaudi-3 AI accelerator, aiming to make AI workloads faster and more cost-effective. Leveraging the dominant infrastructure choice, x86, which operates at scale across nearly all data centre environments, Intel is paving the way for seamlessly integrating the power of AI while ensuring cost-effective interoperability and reaping the benefits of an open ecosystem of developers and customers.
At the heart of this endeavour are Intel Xeon processors, serving as the ideal CPU head node for AI workloads. When combined with Intel Gaudi AI accelerators, purposefully designed for AI tasks, they form a powerful solution that seamlessly integrates into existing infrastructure.
The Gaudi architecture stands as the only MLPerf-benchmarked alternative to Nvidia H100 for training and inference of large language models (LLMs), offering customers the sought-after GenAI performance with a price-performance advantage. This advantage provides choice and facilitates fast deployment at lower total operating costs.
Intel offers standard AI kits comprising eight Intel Gaudi 2 accelerators with a universal baseboard (UBB) at $65,000, estimated to be one-third the cost of comparable competitive platforms. Additionally, kits including eight Intel Gaudi 3 accelerators with a UBB are listed at $125,000, approximately two-thirds the cost of comparable competitive platforms.
The Intel Gaudi 3 accelerators promise significant performance improvements for training and inference tasks on leading GenAI models, enabling enterprises to unlock the value in their proprietary data. In an 8,192-accelerator cluster, Gaudi 3 is projected to offer up to 40 per cent faster time-to-train compared to the equivalent size Nvidia H100 GPU cluster and up to 15 per cent faster training throughput for a 64-accelerator cluster versus Nvidia H100 on the Llama2-70B model. Moreover, Intel Gaudi 3 is expected to provide an average of up to 2x faster inferencing compared to Nvidia H100 when running popular LLMs such as Llama-70B and Mistral-7B.
To ensure the widespread availability of these AI systems, Intel is collaborating with at least 10 top global system providers, including six new providers introducing Intel Gaudi 3 to the market. Among these collaborators are Asus, Foxconn, Gigabyte, Inventec, Quanta, and Wistron, expanding the production offerings alongside leading system providers Dell, Hewlett-Packard Enterprise, Lenovo, and Supermicro. This collaboration aims to make advanced AI capabilities accessible to a broader audience, driving innovation and progress in the field of artificial intelligence.
