Every embedded product with a processor at its centre starts from the same engineering trade-off: design the system-on-chip (SoC) directly onto a custom board, or buy it pre-integrated on a system-on-module (SoM). That choice locks in development time, board risk, unit economics, and lifecycle exposure before the prototype powers on, and it becomes expensive to reverse once the architecture is frozen.

Where the compute platform lives is a volume and lifecycle decision before it is a performance one.

SoM vs SoC: Key Differences at a Glance

An SoC is the silicon itself: processor cores, graphics, and peripheral controllers on a single die. A system-on-module packages that silicon with memory, power management, and storage on a compact board, ready to mount on a carrier.

• Form: A single chip that needs a complete board designed around it, against a pre-integrated board that drops onto a carrier.
• Design Effort: A full high-speed board for a chip-down design, against a carrier board alone for a module.
• Time to Market: Longer for chip-down, where every high-speed revision adds weeks; shorter for a module, where the hardest layout is already proven.
• Unit Economics: Lower bill of materials (BOM) at volume for chip-down, versus lower non-recurring engineering (NRE) for a module.
• Flexibility: Full control of layout and form factor, against the constraints of the module footprint and connector.
• Lifecycle: The original equipment manufacturer (OEM) manages component obsolescence on a chip-down board; the module vendor manages it on an SoM.

Side-by-Side: Differences between SoC and SoM

Headline specifications are the least informative point of comparison between a system-on-module and a chip-down board, because the same silicon can sit inside either approach. NXP's i.MX 95 applications processor, for example, ships as a bare chip for chip-down designs and as a system-on-module from several vendors, often catalogued as IMX95. What separates the two is who engineers the board around the processor. At what stage, and that difference plays out across six dimensions that decide the programme outcome.

Integration

A chip-down design integrates everything on one printed circuit board (PCB). Removing the board-to-board connector gives the designer full control over placement, layer stack-up, and form factor, and eliminates a mechanical interface that would otherwise need qualifying. An SoM takes the opposite route: the processor subsystem arrives pre-integrated, and the carrier board handles only product-specific input/output (I/O) and power entry.

Development Complexity

The chip-down burden concentrates in the processor subsystem. Low-power double data rate (LPDDR) memory needs length-matched routing within tight timing budgets, fine-pitch packages often force high-density interconnect (HDI) stack-ups, power rails must sequence in the order the silicon expects, and every high-speed interface needs signal integrity validation. System-on-module design removes most of that risk, often leaving a carrier board of four to six layers. 

Neither approach escapes the fundamentals, though: PCB design still decides whether the carrier or the custom board performs as intended.

Performance

Identical silicon performs identically in principle. In practice, a chip-down board gives the designer full control over thermal path, power delivery, and memory configuration, so all three can be sized around the real workload. An SoM fixes memory size and type at the module level, and its compact footprint and connector can limit how much heat leaves the processor. 

The constraint rarely shows on the bench. It surfaces under sustained inference load, when a processor with nowhere to shed heat throttles below its rating.

Reliability

A connector-based SoM must be qualified against shock, vibration, and thermal cycling, because the board-to-board interface is a failure point that a one-board design does not carry. That is why ruggedised and automotive designs often favour solder-down modules or chip-down boards. 

The counterweight is maturity: a proven module arrives with the hardest part of the board already validated in volume production, which lowers first-revision risk.

Product Lifecycle

An SoM vendor absorbs memory and component obsolescence behind a stable interface, revising the module internally while the carrier stays unchanged. Industrial modules commonly carry availability commitments of ten years or more. A chip-down design places every end-of-life notice on the OEM, but frees the product from dependence on a single module vendor's roadmap.

Cost

An SoM lowers NRE and can shorten development by months, in exchange for a higher per-unit BOM, since every unit carries the module vendor's engineering. A chip-down board reverses that equation. The crossover point depends on lifetime volume and on the commercial value of reaching market sooner, which places the decision in the business case as much as the schematic.

When to Choose an SoC

A full-custom chip-down board makes sense when lifetime volume is high enough to recover the added engineering investment across the production run. Many programmes reach it in stages: they launch on an SoM, prove demand, and migrate to a chip-down board once volumes justify it, using the module's reference design as the starting schematic. 

The conditions that favour an SoC are consistent:

• High Production Volume: Enough units to amortise NRE against a lower BOM.
• Tight Form Factor: Size or weight limits that cannot accommodate a module and carrier stack.
• Harsh Environments: Vibration and temperature profiles where a board-to-board connector is a liability.
• High-Speed Design Capability: In-house or partner expertise in memory routing and signal integrity.
• Workload-Specific Optimisation: Thermal and power design tuned around a known sustained load.

When to Choose an SoM

An SoM suits low-to-mid volume, high-mix programmes where time to market and engineering risk outweigh per-unit cost. It also occupies a different position from a single-board computer (SBC). An SBC is a complete, ready-to-run board suited to prototypes and low volumes, while a module on a custom carrier gives a production product its own form factor and I/O. Choosing between an SoM and an SBC is therefore a question of how close a product sits to production; the role of SBCs in embedded applications offers wider context.

Before committing, evaluate system-on-module manufacturers on longevity commitments that match the product's service life, board support package (BSP) quality and maintenance, and whether the form factor follows an open standard such as Smart Mobility Architecture (SMARC) or a proprietary pinout that ties the carrier to one vendor.

• Low to Mid Volume: Production runs too small to recover chip-down NRE.
• Many Product Variants: One module shared across several carrier designs.
• Compressed Schedule: A launch date that cannot absorb multiple high-speed board revisions.
• Upgrade Path: Pin-compatible modules that allow a processor refresh without redesigning the carrier.

FAQ

What Is the Difference between SoC and SoM?

An SoC is a single chip that integrates the processor, graphics, and I/O controllers. An SoM is a small board that combines an SoC with memory, storage, and power management, ready to mount on a carrier board that supplies the product's connectors and interfaces.

Engineering the Board Around Your Compute Platform

Whichever path a programme takes, the board around the processor decides whether the silicon delivers its datasheet performance. PCI Ltd is an electronics manufacturing services (EMS) partner with over 50 years of experience, engaging at the platform decision, before the architecture is frozen. ISO 9001, IATF 16949, and ISO 13485 certifications cover industrial, automotive, and medical programmes. 

Capabilities relevant to SoC and SoM programmes include:

• Electronic hardware design: carrier board and chip-down design, covering power delivery, memory routing, and thermal strategy from the first revision.
• Single-board computers: SBC design and manufacture for programmes that need a ready-to-run platform.
• PCB assembly: precision surface-mount technology (SMT) placement for fine-pitch ball grid array (BGA) SoCs and module mounting.
• Product design verification: validation under sustained load, shock, vibration, and thermal cycling.
• High-mix low-volume manufacturing: production built for variant-heavy, module-based programmes.

Choosing Where Your Compute Platform Lives

The SoC or SoM decision rests on four variables: volume, schedule, environment, and lifecycle. Each is known, or knowable, before the first board revision, which is when the decision is cheapest to settle. If a new embedded product is on your roadmap, contact us about which platform fits the build, from carrier board design through to volume production.