More diversity for new challenges
Boost for embedded computing
Several hardware and software technologies are currently converging. This will bring about a lasting change in the embedded computing sector.
At least with Intel-based assemblies, everything seems to be as usual: The assemblies are gradually being equipped with processors of the “Coffee Lake” generation and the next generation change is being prepared. The industry is thus following the Intel roadmap, as usual. Less familiar is the fact that Intel has extended the availability of selected CPU models from 7 to 15 years. This means that it is no longer necessary to automatically schedule each new CPU generation in order to have the longest possible availability. Depending on the time horizon and performance requirements, users – and also suppliers – can omit one or the other CPU generation.
If you are looking for more performance in the graphics area, the AMD Ryzen family, which is increasingly penetrating the embedded computing sector, is the right choice. The high quality of the integrated AMD GPUs is also indirectly confirmed by Intel, because Intel now also incorporates AMD GPUs in some of its new high-end NUC computers – in the embedded sector, however, no comparable combination is yet available.
The generation change in ARM assemblies is already more turbulent, such as the change from i.MX6 to i.MX8: The considerable increase in performance can be bought with increased development costs, both on the hardware and software side. The latter, in particular, is in great demand because the topic security is also increasingly demanded at the Internet of Things (IoT) edge. This is where Microsoft wants to score points with “Azure Sphere” and securely network devices with microcontrollers – from the chip to the cloud.
Azure Sphere includes three components with which Microsoft wants to combine experience from the cloud, software and device hardware to form a secure platform: In addition to the Azure Sphere operating systems and cloud security services, there are Azure Sphere microcontrollers (MCUs). Even if the individual components of the concept cannot convince everyone at the moment, it must be acknowledged that Microsoft is working with partners for the first time on a consistent solution for microcontrollers – in other words, it is active well beyond the usual PC environment.
With “Windows for ARM”, the software giant has not distanced itself quite so far from its roots. In contrast to earlier efforts in products for the ARM architecture, “Windows for ARM” can provide an often useful compatibility to 32-bit x86 programs. Especially for embedded projects that need to use older Windows x86 application software, a hardware alternative can now open up. However, extensive testing must show whether it works properly, because the hardware requirements are high, and the closer to the x86 hardware has been programmed, the more critical it becomes.
This step from Microsoft to ARM is not only due to the stagnating notebook and PC business, but also to a shift in innovation towards smartphones and tablets – two device groups where Microsoft’s successes has so far been quite moderate. ARM, on the other hand, not only dominates this business, but also wants to expand its success into the x86 device domains. For example, ARM-based 64-bit processors are penetrating into the server market, and therefore driving forward the development of high-end ARM devices. For embedded computing, however, developments for the notebook segment and below are currently of greater importance.
Alternatives for processor architectures are always welcome in the embedded sector; however, a shift does not occur all too often so far, because backward compatibility is usually required for the software. Instead of new developments, upgrades for the functional expansion of the target application are often worked on. If the new software overloads the existing hardware, modular systems can score points by using faster CPU cards and modules. But here too, a change is emerging with the next generations of PCI Express (4.0 and 5.0). This is because the high data rates overburden many connectors that are used for modular embedded assemblies. New connectors must therefore be used and thus the connector compatibility for most standards is lost.
- Boost for embedded computing
- »Tensor-Flow« as a standard
