» Current | 2018 | 2017 | 2016 | Subscribe

Linley Newsletter

Cascade Lake Speeds Deep Learning

September 4, 2018

Author: Bob Wheeler

Cascade Lake-SP is a post–Moore’s Law design, boosting performance through new features rather than classic core-count and process-technology scaling. This second-generation Xeon Scalable processor refreshes the Purley server platform Intel unveiled with Skylake-SP in 3Q17. It’s not only socket compatible but also sports the same number of cores and threads as well as the same L3 cache size. It includes hardware mitigation for several side-channel vulnerabilities such as Spectre and Meltdown. Intel is building Cascade Lake-SP in its 14++ process, which will deliver undisclosed but small clock-speed improvements, and it expects production shipments in 4Q18.

The processor’s major microarchitecture addition is the new Vector Neural Network Instruction (VNNI) extension. The first set of features introduced under Intel’s Deep Learning Boost brand, VNNI accelerates inference workloads. For convolution loops operating on 8-bit integers, the new vector unit fuses three instructions into one. In parallel with Cascade Lake-SP, Intel is optimizing its software, including the MKL-DNN (Math Kernel Library for Deep Neural Networks).

As expected, Cascade Lake-SP is Intel’s first processor that enables Optane DC persistent-memory modules, which are DIMM-like devices based on 3D XPoint memory. They implement a DDR4 electrical interface, and the processor’s memory controllers allow each channel to have both a standard DIMM (SDRAM) and an Optane DC module. A 2S server can handle 6TB of Optane memory plus a lesser amount of DDR4 SDRAM. The primary use for this new memory type is in large databases—specifically, those that won’t fit in 1.5TB of DRAM.

Lacking a new process node or new microarchitecture, Cascade Lake-SP delivers only a modest upgrade over Skylake-SP. It offers a big performance boost for certain workloads but underwhelming improvements for the broad market.

Subscribers can view the full article in the Microprocessor Report.

Subscribe to the Microprocessor Report and always get the full story!

Purchase the full article

Events

Linley Fall Processor Conference 2018
Covers processors and IP cores used in embedded, communications, automotive, IoT, and server designs.
October 31 - November 1, 2018
Hyatt Regency, Santa Clara, CA
Register Now!
More Events »

Newsletter

Linley Newsletter
Analysis of new developments in microprocessors and other semiconductor products
Subscribe to our Newsletter »