Doubling down: Intel’s 8-core Xeon processor raises the performance bar for rugged systemsStory
September 13, 2021
In August 2021, Intel announced the new Intel Xeon W-11000E Series processor (formerly known as “Tiger Lake-H”), designed for the embedded market. This new processor follows the announcement of similar 11th-generation processors introduced for the commercial market several months earlier.
For designers of rugged commercial off-the-shelf (COTS) systems, this new processor – which supports critical features for embedded systems, such as DRAM error correction (ECC) and extended operating temperatures – is a great addition to Intel’s range of embedded processors that also offer extended life cycle availability, which is especially important for typically long-life military programs. At the high end of the Xeon W-11000E series is an 8-core device that doubles the core count of previous-generation quad-core processors, making it all the more attractive for embedded virtualized applications.
The Intel Xeon W-11000E provides enterprise-class virtualization with its large core count, the ability to support as much as 128 GB of memory, and its use of Intel Virtualization Technology (VT-x, VT-d). The processor’s high core count enables system designers to reduce their platform’s SWaP profile by consolidating multiple independent processing applications onto a single module.
What does this mean for embedded system designers? Processing tasks that formerly required multiple modules or multiple systems can now be consolidated into a single slot. Additionally, the Intel Xeon W-11000E boosts single-thread performance by as much as 32%, increases multithread performance by as much as 65%, and improves graphics performance by as much as 70%.
For deployed general-purpose applications, the Intel Xeon W-11000 is aimed at applications that require extreme single-threaded performance. The high-end 8-core device runs at 2.6 GHz nominally (4.7 GHz with turbo), enabling multiple time-sensitive tasks to be run in parallel, minimizing the latency in reaching critical decisions. For artificial intelligence (AI) and machine learning (ML) applications, the processor family uses built-in support for VNNI instructions and the Intel OpenVINO framework. It also accelerates vector processing algorithms using AVX-512 engines, which support accelerated floating-point math algorithms at more than double the speed of previous generations.
For applications needing intensive math-processing capabilities, embedded designers have traditionally relied on external graphics processing units (GPUs), such as those supplied by NVIDIA. Now, for many of these DSP applications, the Intel Xeon W-11000E provides TeraFLOP-level performance sufficient to meet or even exceed the previous generation’s capabilities. Eliminating an external GPU delivers significant size, weight, power, and cost (SWaP-C) savings to systems built around Intel’s new processor.
In the area of embedded graphics, this 11th-generation processor includes Intel’s Gen 12 graphics engine, which enables as many as three 4K display interfaces. It delivers 2D and 3D performance that rivals, and often exceeds, the performance of many of the discrete graphics chips from other vendors that have been used in embedded systems in the past.
An example of a new single-board computer (SBC) that’s able to maximize the features of Intel’s new processor is Curtiss-Wright’s VPX6-1961. (Figure 1.) This rugged 6U OpenVPX board has eight high-speed hyperthreading cores, a dual-channel DDR4 memory subsystem connected directly to the processor, and support for as much as 64 GB of SDRAM with ECC support. Additional features include up to 1 TB of high-speed NVMe on-board SSD memory. Dual XMC mezzanine sites support a wide variety of expansion mezzanine daughtercards, such as high-performance FPGA [field-programmable gate arrays], general-purpose GPUs, and storage modules. Four 10G Ethernet ports are included for control plane connectivity, supporting 10GBASE-T and 10GBASE-KR interfaces. The board also supports 1000BASE-T and 1000BASE-KX/BX for backwards compatibility to 1G-based systems.
[Figure 1 | The VPX6-1961 SBC enables data to flow from the VPX backplane directly to the XMC sites to support demanding high-bandwidth applications such as sensor and data processing or video/graphics mezzanine modules. Curtiss-Wright photo.]
The Intel Xeon W-11000E processor is a helpful addition to the range of processors available for designers of deployed military systems. Its combination of high performance and maximum connectivity aim it squarely at use in high-performance embedded computing systems, general computing, and mission-processing solutions, as well as use in multi-SBC systems used in advanced processing and ISR [intelligence, surveillance, and reconnaissance] applications.
Aaron Frank is senior product manager at Curtiss-Wright Defense Solutions.
Curtiss-Wright Defense Solutions https://www.curtisswrightds.com