Until the introduction of multicore technology, performance increases were primarily achieved by increasing the clock frequency. Today, performance increases at a given clock rate primarily through the number of computing cores per processor, so that more tasks can be executed in parallel. With the eighth generation of embedded Intel Xeon and Intel Core processors, both of these factors have now been tweaked once again: for the first time, a 6-core processor is available for embedded computers and, for the first time, it can process twelve threads at up to 4.4 GHz.

The first 'COM Express Type 6' module with up to six processor cores: The conga-TS370 with Generation 8 Intel Embedded Xeon and Intel Core processors.

© congatec

With the increase in cores and clock frequency, the overall computing power increases considerably, which benefits vision control applications and control computers for complex CNC machines, for example, as well as control room computers and monitoring systems for networked factories with support for up to three independent displays in 4k. But what exactly is the performance increase of the processors?

Initial tests with Congatec's 'COM Express Type 6' modules show that these new 6-core processors offer between 45 and 50 % more multi-thread and 15 to 25 % more single-thread performance compared to variants with the 7th generation of Intel core processors.

Powerful six-packs

For a given thermal power dissipation (TDP), system developers can achieve a higher bandwidth with lower overall power consumption. As a result, they can now install high-end computing power and features even deeper in the field and significantly increase the functionalities of their applications. The microarchitecture of the Generation 8 processors has not been changed compared to Generation 7, only the 14 nm manufacturing technology has been further optimized, resulting in lower leakage currents and allowing higher clock rates. As the processor core is otherwise almost identical, no adjustments are necessary on the software side; this enables a seamless migration to the latest processor generation.

Fast upgrades possible

The cooling concept of the COM Express standard is standardized to such an extent that system designs can be equipped with new modules of comparable TDP classes without great effort.

© congatec

If OEMs have designed their systems with Computer-on-Modules, they can very quickly equip them with this processor technology, as systems with a given TDP can be raised to the latest performance level simply by replacing the modules. The COM Express Type 6 specification ensures that modules offer identical dimensions, functions and cooling solutions as ready-to-purchase supercomponents, which already include a complete board support package (BSP) with all the necessary standard drivers. Specifications for the structure of a carrier board also make it easier for developers of industrial-grade computers to develop a system solution tailored to their own requirements. Compared to full-custom designs, it is assumed that developers save around 50 to 90 % of the effort required to develop a suitable solution at board level. If standard motherboards do not meet the requirements of the specifications or are too large, computer-on-modules should always be used - mass production excluded.

Congatec, for example, offers suitable modules. The acquisition of Real-Time Systems (RTS) - manufacturer of a real-time hypervisor for OEMs in the field of industrial control systems and medical technology - means that real-time control systems can be supplemented with additional virtual machines for IoT or Industry 4.0 connectivity. The two additional cores compared to the previously available quad-core solutions are virtually predestined for these two additional functions.

More tasks? More cores!

In automation, however, this increase in performance also benefits systems that have to manage and coordinate many different tasks in parallel. These include, for example, control computers for robotics, production cells as well as complex packaging and machine tools. With multicore technology, for example, individual functional elements of a machine such as HMI and (real-time) control can be operated on an embedded hardware platform, which reduces the number of computer systems in a machine, thereby reducing costs and increasing reliability. In the field of motion control, for example, motion axes can be assigned to individual processor cores. The data exchange required to coordinate the motion sequences takes place in the system itself in such installations and is therefore much more efficient than if separate axis drive controllers were networked - especially as networked systems also have latencies that can reduce the precision of a machine.

The collaborative robotics market also represents a new but booming segment for powerful multicore systems. According to Markets und Markets, it is expected to grow at an annual rate of no less than 56.94% between 2017 and 2023 to reach a total global volume of 4.28 trillion US dollars. Collaborative robots require even more computing capacity, as they require additional subsystems for situational awareness and adaptive control in addition to the actual control system. These can be systems such as Lidas and (stereoscopic) cameras for obstacle detection as well as subsystems for adaptive control and pathfinding in mobile solutions. These systems require additional computing instances that can be implemented very efficiently in the control systems via more computing cores.

IoT networking and condition monitoring

Automation OEMs also want more computing cores for IoT and Industry 4.0 integration and condition monitoring of their machines and systems. At this point at the latest, however, virtualization makes sense in order to separate the individual tasks from one another. If, depending on the software, the real-time controls of individual components can still run as multi-threaded software on an operating system instance, the gateway functionality should be separated from the real-time controls. Intelligent condition monitoring can also be efficiently integrated in this way, with local rule engines evaluating and monitoring the condition of the mechanical components - for example via a vibration analysis. This also sometimes requires significant computing power. High-end embedded systems for machine control systems can therefore never have enough computing cores and computing power per core.

The key technical data

Congatec already offers the first modules with the new generation of processors: The new Conga TS370 modules with Intel Xeon and Intel Core i7 processors or quad-core Intel Core i5 processors have a TDP of 35 to 45 W and support up to 32 GByte DDR4 2666 RAM, so that even with comprehensive virtualization with several virtual machines, each instance has more than enough RAM available. Error correction code (ECC) is optionally supported for safety-critical applications, such as situational awareness for collaborative robots. The modules are also characterized by their high-bandwidth I/Os. These include 4x USB 3.1 Gen 2 (10 Gbit/s), 8x USB 2.0 as well as 1x PEG and 8x PCIe Gen 3.0 lanes for powerful system expansions. The high long-term availability of at least ten years, Intel Optane Memory Support and the extended security features with Intel Software Guard Extensions, Trusted Execution Engine and Intel Platform Trust Technology are also impressive. The modules support all common Linux operating systems as well as the 64-bit versions of Microsoft Windows 10 and Windows 10 IoT.

Author: Zeljko Loncaric is a marketing engineer at Congatec.