AI • IOT • NETWORK EDGE

While recent trends in high-end rugged computing have emphasized heterogeneous architectures, the Intel^® Xeon^® D family makes a strong case for general-purpose architectures. As the first Xeon processor qualified for extended temperature ranges, the Xeon D offers more flexibility and higher performance than is typical for embedded products.

Specifically, the Xeon D line supports a temperature range stretching from -40C to 85C. The Xeon D family isn’t the first set of Intel^® chips to support this range, but the company’s previous extended-temperature products were derived from its Intel Atom^® processor line. Intel Atom CPUs emphasize low power consumption rather than high performance.

As such, the Xeon D processors, offer considerably greater performance than older extended-temperature processors. This means the new processors can take on tasks that previously required a heterogeneous architecture, e.g., a CPU plus an FPGA. By running everything on a Xeon D instead, you get a simpler, more scalable system architecture.

The Xeon D family also offers an impressive degree of flexibility. With a wide range of core counts (4-16), clock speeds (1.3GHz – 2.7GHz Turbo), and TDPs (20-65W), the family can tackle a wide range of requirements.

A white paper from Curtiss-Wright explains how these new capabilities allow the Xeon D family to take on many different application roles and system tasks. The highlighted systems are:

• A 3U system designed for non-throttling performance up to an 85^°C sidewall temperature in a conduction-cooled environment.
• A two-socket 6U system with support for up to 16 cores per socket and up to 32GB of DRAM per CPU.
• A 6U system with just one CPU socket but far more memory (up to 128GB).

This variety of configurations illustrates how Xeon D can scale to meet many different needs—including extreme requirements in terms of temperature, compute, and memory capacity.

On that last point, Curtiss-Wright points out that sensor fusion – the process of combining data from multiple sources – is a good example of a memory-hungry application. Multiple high-resolution and/or high-sampling-frequency sensors can generate enormous amounts of data very quickly.

To analyze the data in real time, a system needs both massive compute power and an extraordinarily large RAM. The single-socket, 128GB system illustrates how architectures based on Xeon processors can rise to the challenge – and how taking the CPU-only approach results in a simpler design than an equivalent heterogeneous system.

In short, the high-performance, flexible configuration and broad temperature range of the Xeon D make it an excellent choice for a variety of use cases. The ability to scale out to 32 cores or up to 128GB of RAM is particularly noteworthy, while the ability to scale down to 4 cores means that low-power requirements can be addressed with the same architecture.