The new Cooper Lake processors, and the other components of the Intel announcement, all share a focus on improving AI-related capabilities.
Today, the Intel Data Platforms Group is announcing that their 3rd Gen Intel Xeon Scalable processors, code-named Cooper Lake, are shipping. That’s the feature announcement, but it is accompanied by others around the company’s hardware and software AI portfolio.
“With network transformation around 5g, AI and the intelligent edge rising up, we are seeing an explosion of demand and use cases, and new modernized workloads delivered by new modernized services,” said Lisa Spelman, Intel corporate vice president and general manager, Xeon and Memory Group. “Our goal is to service every single data centre workload.”
The new Cooper Lake processors announced are 4- and 8-socket models.
“Their enhanced speed allows for maximizing performance of highest priority workloads,” Spelman said. “They provide up to 1.98x higher database performance, compared to a 5-year old platform.”
What’s of more interest here beyond the raw performance is the new processors’ AI capabilities. They are the first mainstream server processor with built-in bfloat16 support, which will make them more suitable for AI-related use cases on general CPUs, including deep learning, VM density, in-memory database, and analytics-intensive workloads. They will also support applications that include image classification, recommendation engines, speech recognition and language modeling.
“bfloat16 is a compact numeric format which uses fewer bits to achieve greater speed with only slightly lower accuracy than the FP32 format,” Spelman stated. “It also requires very minimal software changes for customers.”
bfloat16 support will accelerate both AI training and inference performance in the CPU. Intel-optimized distributions for leading deep learning frameworks [including TensorFlow and Pytorch] support bfloat16 and are available through the Intel AI Analytics toolkit. Intel also delivers bfloat16 optimizations into its OpenVINO toolkit and the ONNX Runtime environment to ease inference deployments.
As part of the 3rd Gen Intel Xeon Scalable platform, Spelman also introduced the Intel Optane persistent memory 200 series. It provides up to 4.5TB of memory per socket to manage data intensive workloads.
“In the past, tradeoffs between capacity speed cost persistence, and we have been working to close these gaps,” she said.
Finally, Spelman announced the availability of the next-generation high-capacity Intel 3D NAND SSDs, the Intel SSD D7-P5500 and P5600, which are targeted at systems that store data in all-flash arrays. These 3D NAND SSDs are built with Intel’s latest triple-level cell [TLC] 3D NAND technology and an all-new low-latency PCIe controller, as well as advanced features to improve IT efficiency and data security.
Intel also announced their first AI-optimized FPGA, the Intel Stratix 10 NX FPGAs.
“This is Intel’s first AI-optimized FPGA, and provides up to 15x more compute performance than todays Stratix 10 MX FPGAs,” said David Moore, Corporate Vice President & General Manager, Progammable Solutions Group at Intel.
These FPGAs provide customizable, reconfigurable and scalable AI acceleration for compute-demanding applications like natural language processing, fraud detection and Smart Cities. They also contain new AI-optimized arithmetic blocks called AI Tensor Blocks, which contain dense arrays of lower-precision multipliers typically used for AI model arithmetic.
“Our current DSP blocks are for more general purpose compute,” Moore indicated. “AI Tensor Blocks are optimized for the AI space. That’s what lets us pack 15x more compute into the same footprint.”
The Intel Stratix 10 NX FPGA is expected to be available in the second half of 2020. However, the 3rd Gen Intel Xeon Scalable processors, the Intel Optane persistent memory 200 series, and the Intel SSD D7-P5500 and P5600 3D NAND SSDs are all available today.