Embedded computing is shaped not only by processor performance, but by how effectively a complete system moves, stores and processes data. A 2019 review of the VITA technology industry points to several developments with relevance for designers of high-performance embedded systems: open standards, modular architectures, new memory approaches and faster interconnects. The report also highlights how demanding computing markets, including supercomputing and defense, were influencing the technologies likely to reach embedded platforms.
For industrial automation engineers and system integrators, the practical question is not whether every emerging technology will appear in a control system. It is how changes in computing architecture may affect platform selection, system integration and the ability to upgrade complex equipment over time.
Open standards and modular embedded computing
VITA is an industry association focused on ANSI-accredited open system architectures for critical embedded applications. Standards such as VPX and OpenVPX provide a common framework for modular board-and-system designs. The value of an open architecture is that it can give equipment designers and integrators a defined basis for combining components, rather than relying entirely on a single proprietary implementation.
The report connects this standards-based approach with emerging military concepts such as Modular Open Systems Approach (MOSA) and other platform initiatives. Its broader point is that complex platforms may be designed as collections of interoperable subsystems. Instead of treating a vehicle, aircraft or other system as one fixed unit, designers can consider how functional modules may be integrated and updated within a larger architecture.
That direction is relevant to industrial systems where computing, communications, sensing and control functions must work together. Modular design can make interfaces and component boundaries important considerations from the outset. Open standards do not remove the need for integration work, but they provide a shared technical foundation for it.
Memory bandwidth and I/O become design constraints
One of the report’s central technical themes is the gap between a processor’s ability to perform calculations and the rate at which data can reach it. CPUs and GPUs can be constrained when memory cannot supply data quickly enough. Increasing memory bandwidth can help, but wide memory buses also bring power demands into the design trade-off.
Several memory technologies were under development or in use in the period covered by the report, including high-bandwidth memory, 3D XPoint, MRAM, phase-change memory, ReRAM and stacked SRAM and DRAM. These approaches involve different balances of density, performance and power consumption. The report also discusses in-memory computing, in which memory devices perform some data manipulation rather than requiring all data to be moved into processor registers first.
I/O is closely connected to the same bottleneck: data transfers commonly move information into memory before processing. The report describes work on faster electrical links, including 112 Gb/s connections for chip-to-chip and die-to-die communication, as well as PCIe 6.0 and Compute Express Link (CXL). CXL is described as a cache-coherent link based on PCIe. These technologies illustrate why board, backplane and interconnect design can become increasingly demanding as data rates rise.
For embedded designers, the implication is to assess the whole data path. Processor specifications alone do not establish system performance. Memory capacity and bandwidth, I/O links, board layout and interconnect choices all influence how a platform handles data-intensive workloads.
Computing architectures are diversifying
The report’s discussion of supercomputers and AI processors reflects a wider shift toward heterogeneous computing. GPUs and other specialized processors are used alongside CPUs to handle particular workloads. The report also describes large-scale investment in exascale systems, where multiple processing elements and high-speed interconnects work together.
It does not predict that general-purpose CPUs will immediately disappear. Instead, it frames the transition as uneven: specialized processors may take on more work, while CPUs continue to serve general-purpose tasks, housekeeping and error handling. The exact balance depends on application requirements and the maturity of the technologies involved.
This distinction matters in embedded applications, where workload, power, cooling and system constraints can differ from those in a data center. A design choice that suits a supercomputer does not automatically suit an embedded platform. However, developments in high-performance computing can influence the available processors, memory technologies and interconnect options for embedded systems.
Designing for change through modularity
The report links modular military-platform concepts with open standards and faster technology updates. It describes a direction in which hardware components and software capabilities can be changed incrementally, rather than requiring a complete platform redesign for every improvement. It also introduces the idea of a “system of subsystems” to describe interconnected, modular capabilities.
For system architects, this highlights the importance of defining boundaries between subsystems and understanding the interfaces that connect them. Standards activity may evolve as new requirements emerge, including requirements for subsystem interconnects. Modularity can support change, but only when system interfaces, integration responsibilities and technical dependencies are addressed as part of the architecture.
What industrial teams can take from the trends
- Evaluate the full computing chain. Consider processors, memory and I/O together, especially for data-intensive workloads.
- Examine platform interfaces. Open architectures and established standards can provide a basis for modular integration, but system-level compatibility still needs to be verified.
- Match technology to the application. Specialized processors and advanced memory approaches should be assessed against workload and platform constraints, rather than adopted solely because they appear in high-performance systems.
- Plan for technology evolution. Modular architectures can make incremental changes more manageable when subsystem boundaries and interconnects are clearly defined.
Conclusion
The VITA industry review presents a 2019 snapshot of an embedded-computing landscape being influenced by open standards, modular system concepts and growing demands on memory and I/O. Its enduring engineering lesson is to treat computing performance as a system property: processors, memory, interconnects and interfaces must work together. For industrial automation teams selecting or integrating high-performance platforms, understanding those relationships can support more deliberate architecture decisions and a clearer path for future system changes.