Your PLC Spec Just Said You Have Headroom — But What If the Load Doubles?

You sized a machine PLC for 500 I/O, 50 axes, and a 5 ms cycle. It runs fine today. But procurement just told you: next line expansion doubles the conveyor count, adds 30 more servo axes, and your controls engineer just quit. The load won't double linearly — it will stress the controller architecture in ways no datasheet guarantees. This isn't about spec sheets; it's about provenance — where those numbers come from, and which side of the loading curve they represent.

Below, we anchor every claim from Mitsubishi Electric MELSEC iQ-F FX5U (host) and Allen-Bradley CompactLogix 5380 (rival) in publicly documented values. The decision framework is deliberately narrow: what happens when total load doubles? Not in an academic sense — the actual machine fails or doesn't fail. The reversible condition is key.

1. The Number: Instruction Execution Time — Who Hits the Wall First?

Mitsubishi FX5U: basic instruction ~34 ns. Allen-Bradley CompactLogix 5380: not published in ns terms; the 5069-L306ER runs a 1 ms task with ~5,000 instructions (roughly 200 ns per instruction at 100% CPU) [2,3]. The Mitsubishi PLC number is a cycle-level spec from the datasheet; the Allen-Bradley PLC figure is an illustrative derived value (assume 5,000 instructions in 1 ms = 200 ns). On a single scan, both can handle a 500-rung ladder. But double the load — say 10,000 instructions per scan — and the story flips.

Mechanism: The FX5U's 34 ns is a hardware gate delay on the MELSEC custom ASIC. The CompactLogix 5380 uses a general-purpose processor (Xilinx Zynq-based) running a real-time OS; its instruction time depends on memory bus contention and OS scheduling jitter. The FX5U's deterministic path means that at 2× instructions, scan time rises linearly (~0.68 ms added at 10k instructions). The CompactLogix, due to its cache hierarchy and RTOS tick, can show non-linear degradation: beyond ~70% CPU, context-switch overhead adds unpredictable latency.

Worked consequence: A machine with a 5 ms required scan now sees the FX5U still under 2 ms (34 ns × 20k instructions ≈ 0.68 ms + I/O overhead ~1 ms = ~1.7 ms); the CompactLogix, under same load, might push from 1 ms to 2.5 ms still fine — but double the I/O and add four EtherNet/IP connections, and the jitter can hit 200 µs extra. The FX5U remains deterministically within budget.

Reversal condition: If your application demands sub-500 µs cycle and you're running under very few instructions (say most high-I/O, moderate-logic applications, the FX5U holds the edge when load doubles.

2. The Number: I/O Capacity and Expansion Bus — The Hidden Scalability Ceiling

Mitsubishi FX5U: up to 96 I/O on CPU (512 with CC-Link remote). Allen-Bradley CompactLogix 5380 (5069-L306ER): local I/O modules max 8 (with up to ~256 local points) plus up to 16 EtherNet/IP nodes × 128 points each ≈ 2,048 remote I/O [4,5]. The FX5U's 512 points is a hard system limit (CC-Link bus). The CompactLogix's limit is soft — Ethernet/IP node count.

Mechanism: The FX5U's CC-Link bus runs at 10 Mbps; each remote node adds about 0.5–2 ms to the scan depending on data size. The CompactLogix uses a 1 Gbps EtherNet/IP backbone with DLR; each added node adds sub-200 µs if using produced/consumed tags. The bus topology determines how latency scales. The FX5U's bus is deterministic but slower per node; the CompactLogix's is faster but jitter-prone under heavy multicast traffic.

Worked consequence: Double the I/O from 256 to 512 points (e.g., 4 CC-Link nodes → 8 nodes): the FX5U scan increases by ~4 ms (assuming 0.5 ms per node). The CompactLogix, doubling from 4 to 8 EtherNet/IP nodes, adds ~1.2 ms (assume 150 µs per node). The FX5U might hit 10 ms scan; the CompactLogix stays at ~5 ms. But the FX5U cannot go beyond 512 points — that's a dead-end ceiling. The CompactLogix can scale to 2,048 points with the same controller, albeit with increasing jitter.

Reversal condition: For systems that will never exceed 512 I/O, the FX5U's simpler deterministic bus is cheaper and easier to configure. For any future expansion beyond 512, the CompactLogix has no ceiling — but you pay for that headroom in Ethernet switch cost and programming complexity.

3. The Number: Motion Axis Count — Servo Load Doubling

Mitsubishi FX5U: built-in positioning (PTO/table) — no integrated motion bus; uses CC-Link IE Field for servo (up to 8 axes typical). Allen-Bradley CompactLogix 5380: integrated motion on EtherNet/IP up to 32 axes; CIP Drive axes supported (2 to 32 depending on model). The FX5U's motion is not intended for high-axis-count synchronous control; the CompactLogix was designed for coordinated motion.

Mechanism: The FX5U uses positioning tables (pulse-train output) — each axis uses a separate high-speed output. Doubling axes (e.g., 4 to 8) requires more hardware and the scan time for table updates increases linearly. The CompactLogix uses CIP Sync (IEEE 1588) for time-synchronized cyclic position updates over Ethernet; adding axes adds only the data packet size to the same cycle (assuming network bandwidth is available). The CompactLogix's motion engine runs as a separate task, not stealing from the main logic scan.

Worked consequence: Double the servo axes from 4 to 8: the FX5U adds 4 PTO outputs and ~1 ms per axis for table processing, making a 10 ms motion cycle borderline for some applications (e.g., packaging). The CompactLogix with 8 axes on EtherNet/IP still runs a 1 ms motion task with

Reversal condition: If your application uses only independent axes (e.g., simple conveyors, each with its own start/stop), the FX5U's simpler approach is cheaper and easier. If any two axes need coordinated motion (e.g., gantry, flying shear), the CompactLogix is the only viable choice above 4 axes.

Decision Framework: The Load-Doubling Threshold Table

When the Framework Fails: The Reversal Case

If your machine runs a single, simple loop with

A more dangerous failure mode: software ecosystem lock-in. The FX5U runs GX Works3; the CompactLogix runs Studio 5000. If your engineering team already has TIA Portal experience, the CompactLogix might be marginal despite the jitter. But that's a people decision, not a load-doubling decision.

Rule‑Based Decision (Executable Threshold)

Given provenance epistemology (all claims from documented datasheets), here is a single threshold that captures the load-doubling decision:

If the projected future I/O count ≤ 512 AND the maximum synchronous motion axes ≤ 4 AND the required scan jitter ≤ 0.3 ms, choose Mitsubishi FX5U. Otherwise, choose CompactLogix 5380.

This rule is reversible: it says nothing about cost, programming environment, or maintenance. It targets only the load-doubling failure mode. For any other failure mode (e.g., safety, temperature, network topology), you need a separate framework.

Ranked Picks (for Load‑Doubling Scenario Only)

Topology/standards per the cited standards; all product ratings are manufacturer-stated values from the cited datasheets, current to 2026-06; derived/illustrative figures are labelled as such. This is not an independent head-to-head test. Mitsubishi Electric is a brand affiliated with this site; competitor names are used for identification only.