What Is a SoC FPGA? Combining CPU and Programmable Logic in One Chip (Zynq and Beyond)

A SoC FPGA puts processor cores and programmable logic on a single chip, two functions on the same device instead of two separate parts on a board. This has been the market direction for roughly a decade: most current parts are hybrids rather than plain FPGAs. As Adam Szychulec, Head of Hardware / Embedded at InTechHouse, puts it, an FPGA used to be an FPGA; today you get ARM cores you can run Linux on, the fabric, often an integrated GPU, and peripherals for fast interfaces, all on one die.

Key Takeaways

• A SoC FPGA combines a processor and programmable logic on one chip. It is not a niche design choice; it has been the mainstream direction in the FPGA market for about a decade.
• "SoC" alone is nearly meaningless today, since almost every modern device qualifies. "SoC FPGA" or "CPU + FPGA on one chip" is the precise term this article actually means.
• The processor runs the operating system and control flow; the fabric takes on work that needs to be parallelised. An FFT is the honest exception, since it still computes better in the processor.
• Resources are often left deliberately spare in the fabric, so hardware-level functions can be added later without a board change.
• InTechHouse has delivered several Zynq and Zynq UltraScale+ systems for telecom, aerospace and oil & gas customers, and works day to day in Vivado, Vitis, Quartus, Lattice Radiant and Lattice Diamond.

What Is a SoC FPGA?

A SoC FPGA puts processor cores and programmable logic on a single chip: two functions, one device, rather than a processor and an FPGA connected across a circuit board.

That combination has been the direction of the FPGA market for roughly a decade, and most current parts sold today are hybrids of this kind rather than plain, processor-free FPGAs. Today's devices typically add ARM cores capable of running Linux, the fabric itself, often an integrated GPU, and peripherals built for fast, standard interfaces, all on one chip.

This guide covers what the term actually means, how work splits between processor and fabric, where the combination earns its place, and where to find a team that has shipped systems built on it.

System on Chip: Why the Term Needs Narrowing Before It Is Useful

"System on chip" on its own has become close to meaningless, and that is worth saying before using the term again.

In general use, SoCs integrate complete systems, a processor, a memory controller, peripherals, sometimes graphics, that would once have needed multiple separate chips on a board. That is why the phone in your pocket, a general purpose application processor and a basic microcontroller are all, technically, systems on chip. The category is too wide to describe anything specific on its own.

The narrower thing this article is actually about is programmable logic on the same die as the processor. That is what "SoC FPGA" means here, and it is the term worth using instead of bare "SoC" from this point on.

SoC FPGA vs FPGA: The Key Differences Against Traditional FPGAs

The main difference against traditional FPGAs comes down to one thing: where the processor lives.

On a traditional FPGA, a processor had to sit beside the device on the board as a separate chip, or be built inside the fabric itself as a soft core, at the cost of logic resources that would otherwise go to the application. On a SoC FPGA, the processor is hard silicon on the same die as the fabric, connected by high bandwidth on-chip interconnect rather than by traces across a board.

What that actually contains matters to an architect. Zynq UltraScale+, for example, has fabric, graphics capability and two distinct processor types: Cortex-A cores for application software, and Cortex-R cores for real-time work. An application core running Linux next to a real-time core next to programmable fabric, all on one die, is a different design conversation from "a CPU plus an FPGA" wired together on a board. Vendor and third-party IP cores cover most of the standard interfaces on these devices, which is what makes design reuse practical.

Understanding ASIC, FPGA and SoC as Three Different Answers

An application specific integrated circuit is custom silicon commissioned for one product: it carries non recurring engineering costs and a mask set before a single working part exists, which is why the decision is framed against production volume, and why low volume products rarely reach it.

A SoC FPGA gives a design team an integrated processor-plus-fabric system without that upfront commitment, at the cost of a higher per-unit price and, unlike an ASIC, never quite as optimized for the one function it runs. The full decision between the three is covered in a dedicated guide.

SoC Integration: Performance and Power Efficiency on One Chip

Integration is a mechanical argument, not a marketing adjective, so it is worth explaining what it actually buys a design.

Moving the processor onto the same die as the fabric removes the board traces that would otherwise sit between them, which cuts latency and the power spent driving signals across external interfaces. One device instead of two also means fewer components, simpler power management, and one less connection that can fail, which is where the reliability argument actually comes from, rather than from the silicon being inherently more dependable.

Worth being honest about the other side: a SoC FPGA is not low-power in absolute terms, and it is a poor fit for a battery powered device compared with a simple microcontroller doing the same job. The power efficiency case here is relative, against a two-chip solution doing the same work, not an absolute claim about low power operation on its own.

Check FPGA vs microcontroller guide.

How the Work Gets Split: Partitioning Between Processor and Fabric

This is the part that distinguishes a real answer from a generic explainer: a method with a worked judgement behind it, not a rule of thumb.

The operating system and the control flow run on the processor. Work that must be parallelised, large data volumes, image processing, signal processing, goes to the fabric instead.

"You put the operating system on the processor, and then the tasks that need parallelising, processing large volumes of data, image, signals, or where you simply need a lot of flexibility in the system for future development, or unusual interface configurations. Then these are very good solutions, combining the advantages of FPGAs and processors. Of course there are certain disadvantages that come with it too."

‍

Adam Szychulec, Head of Hardware / Embedded, InTechHouse

Not every heavy computation belongs in the fabric side of that split. An FFT, for instance, computes better in the processor, worth stating plainly because it proves this partitioning rule was actually applied to real designs rather than recited from a textbook.

The flexibility argument is stronger on a SoC than on a plain FPGA specifically because of this split: resources are often left spare in the fabric so functions can be added later without a PCB change, hardware-level functions rather than a firmware update, while the processor side is still updated in software on its own. Space and aerospace customers do this on purpose.

Check our FPGA hardware acceleration guide.

Common Applications Where the Hybrid Earns Its Place

Where this combination earns its place: environments working with different kinds of data at once, video, telemetry, sensor data; parallel processing of large volumes within a short time; unusual interface configurations a standard processor cannot offer; and flexibility for future development without a board respin.

More broadly, these parts also appear in motor control and industrial drives, instrumentation, consumer electronics and IoT devices, wherever a design needs a general-purpose processor and dedicated parallel logic in the same product.

Adaptive SoC and What the Vendors Call These Devices

"Adaptive SoC" is AMD's current marketing name for this class of device, worth knowing if you are searching vendor material directly, though it is a branding term rather than the framework this article uses.

The wider landscape: AMD/Xilinx offers Zynq and Zynq UltraScale+; comparable families exist at Intel/Altera; Microchip has its own SoC FPGA line; Lattice builds comparable parts on RISC-V. Choosing between them is its own decision. Check Choosing an FPGA vendor and family guide.

Zynq Experience: What InTechHouse Has Actually Built

InTechHouse has delivered several Zynq and Zynq UltraScale+ systems for telecom, aerospace and oil & gas customers, working day to day in Vivado, Vitis, Quartus, Lattice Radiant and Lattice Diamond across FPGA and SoC design generally.

Worth stating plainly: these solutions combine the advantages of both worlds, and there are certain disadvantages too. The one worth naming is the tooling. Many of these tools work the way they work, with plenty of bugs and in constant development, and the workflow splits between the logic itself and how software gets loaded onto the device, and with which tools. That tooling experience is a second competence layer on top of the hardware design itself, and it is a real part of what a customer is buying from a team that has already climbed that curve.

Not sure whether your project needs a SoC FPGA or a plain FPGA? Request an FPGA architecture assessment.

‍