The third generation of Intel Xeon Scalable processors has just been announced by Intel. These processors are especially designed for image classification tasks, recommendation engines, speech recognition and language modeling. Xeon Scalable 3rd Generation are characterized by offering great capacity for Artificial Intelligence systems. A market, on the other hand, where the presence of Intel is minority.
These Xeon Scalable processors appear to be the answer to the (partial) NVIDIA DGX-A100 systems that use AMD EPYC processors. NVIDIA's switch to AMD processors after so many years of working with Intel has surely been a major blow to the company. Furthermore, the company needs to improve its position in the AI market.
[amazon box="B0883NTLXM"]Table of Contents
Intel Announces 3rd Generation Xeon Scalable
The new Xeon Scalable 3rd Generation range starts from 18 cores and 36 threads and reaches 28 cores and 56 threads. The maximum exponent is the Xeon Platinum 8380HL which has 28 cores and 56 threads. This processor works at a base frequency of 2.9GHz and arrives in Boost mode at 4.3GHz with a total of 38.5MB of cache. It has Hexa-Channel DDR4 @ 3200MHz support and supports up to 4.5TB of DDR4 RAM. This new processor also offers support for PCI-Express 4.0 and has a TDP of 250W.
Intel press release
New 3rd Generation Intel Xeon Scalable Processors
Intel is further expanding its investment in accelerating the AI embedded in these new 3rd Generation Intel Xeon Scalable processors, by integrating bfloat16 support into the exclusive Intel DL Boost processor technology. bfloat16 is a compact numeric format that uses half the bits as the current FP32 format, but achieves similar model accuracy with minimal (if any) software changes.
The addition of bfloat16 support speeds up both AI training and inference performance on the CPU. Distributions optimized by Intel for the major deep learning frameworks (including TensorFlow and Pytorch) are compatible with bfloat16 and are available through the Intel IA Analytics toolkit. Intel also provides optimizations for bfloat16 in its OpenVINO toolkit and in the ONNX runtime to facilitate inference implementations.
The 3rd generation Intel Xeon Scalable processors (codenamed “Cooper Lake”) represent an evolution of Intel's 4-socket and 8-socket processor offering. The processor is designed for deep learning, virtual machine (VM) concentration, in-memory database, mission-critical applications, and deep analytics workloads. In this way, customers upgrading legacy infrastructures can expect an estimated average profitability of 1,9 times on common workloads3 and up to 2,2 times more VM4 compared to equivalent 4-socket platforms 5 years ago
New Intel Optane persistent memory
As part of the 3rd generation Intel Xeon Scalable platform, the company has also announced the Intel Optane 200 series of persistent memory, which provides customers with up to 4,5 TB of memory per socket to handle large data workloads such as in-memory databases, dense virtualization, analytics, and high-powered computing.
New Intel 3D NAND SSDs
For systems that store data on flash arrays, Intel has announced the availability of its next generation of high-capacity 3D NAND SSDs, the Intel SSD D7-P5500 and P5600. The new 3D NAND SSDs are built with Intel's latest Tri-Level Cell (TLC) 3D NAND technology and an all-new low-latency PCIe controller to be able to meet the intense IO requirements of AI and analytics workloads. In addition, they have advanced features to improve IT efficiency and data security.
First AI-optimized Intel FPGA
Intel has revealed its upcoming Intel Stratix 10 NX FPGAs, Intel's first AI-optimized FPGAs aimed at accelerating AI with high bandwidth and low latency. These FPGAs will offer customers more customizable, reconfigurable and scalable AI acceleration for demanding computing applications such as natural language processing and fraud detection. Intel Stratix 10 NX FPGAs include high-bandwidth memory (HMB), high-performance networking capabilities, and new AI-optimized arithmetic blocks called AI Tensor Blocks, which contain dense sets of multipliers. lower precision, commonly used for AI arithmetic models.
OneAPI Cross Architecture Development for Continuous AI Innovation
As Intel expands its portfolio of advanced AI products to meet the diverse needs of its customers, it is also paving the way for developers by simplifying heterogeneous programming with its portfolio of OneAPI cross-architecture tools to accelerate performance. and increase productivity. With these advanced tools, they can accelerate AI workloads on Intel CPUs, GPUs, and FPGAs, future-proofing your code for current and future generations of Intel processors and accelerators.
Enhanced portfolio of Intel Select solutions that addresses key IT requirements
Intel has enhanced its portfolio of Intel Select solutions to accelerate the implementation of the most urgent IT requirements, highlighting the value of delivering pre-verified solutions in today's changing business context. Today 3 new solutions and 5 enhanced solutions from Intel Select, focused on analytics, AI and infrastructure were announced. Intel Select's enhanced solution for genomics analysis is being used worldwide to find a vaccine for COVID-19 and the new Intel Select solution for VMware Horizon VDI on vSAN is being used to enhance distance learning.
