The popular claim – "a 34 ns basic instruction speed means the Mitsubishi PLC FX5U will always outrun the Omron PLC NX1P2" – sounds clean. But that number is a trap when the load is a real multi-axis motion system. The NX1P2 with its 2 ms primary task cycle and integrated EtherCAT can deliver deterministic axis closure that the FX5U, even with its blazing bit speed, cannot match inside a single scan under jitter. Let’s settle this with three numbers that actually rewire your decision.
| Dimension | Omron NX1P2 (host) | Mitsubishi FX5U (rival) |
|---|---|---|
| Basic instruction speed | not published (illustrative: ~50–80 ns on similar ARM core) | 34 ns basic instruction |
| Primary task cycle (min) | 2 ms (NX1P2-9024DT) | ~4–5 ms (typical for FX5U with motion; not spec'd as min task cycle) |
| Integrated motion bus | EtherCAT (≤8 axes, ≤16 nodes) | CC-Link (via expansion); no native EtherCAT |
| Program memory | 1.5 MB program + 2 MB variable | up to 64k steps (~0.3–0.6 MB) |
| On-board I/O | 24 (14 DI / 10 DO) | up to 96 on CPU (512 with CC-Link) |
| Programming environment | Sysmac Studio (IEC 61131-3) | GX Works3 (IEC 61131-3) |
| Built-in OPC UA server | Yes | No (requires extra gateway) |
1. The 34 ns Claim vs the 2 ms Ground Truth
Number: Mitsubishi FX5U executes a basic instruction in 34 ns; Omron NX1P2 does not publish a bit instruction time, but an illustrative ARM-based estimate is ~50–80 ns.
Mechanism: A raw instruction speed only matters when the control logic is purely sequential and the bus adds zero overhead. In a motion application, the real constraint is the task cycle jitter – the difference between the fastest and slowest scan when axes communicate over a fieldbus. On the NX1P2, the primary task cycle can be set to 2 ms deterministically. The FX5U does not publish a guaranteed minimum task cycle; typical field reports put it at ~4–5 ms when motion is active, and the bus (CC-Link) adds variable latency per node.
Worked consequence: For a 4-axis pick-and-place machine with cycle time 0.6 seconds, a 2 ms deterministic cycle allows the control loop to update position every 2 ms with ±0.5 ms jitter. On the FX5U, a 5 ms cycle with ±2 ms jitter means the axis controller must use a larger position tolerance to avoid overshoot – reducing throughput by ~8–12% (illustrative). The plant engineer who chooses by raw instruction speed alone will design a machine that misses its rated cycles-per-minute.
When it reverses: If the application is pure discrete logic (e.g., conveyor interlocking with no coordinated motion), the FX5U’s 34 ns speed lets it scan a huge rung in microseconds, and the slower task cycle is irrelevant. The NX1P2’s 2 ms cycle becomes a disadvantage only if you need sub-millisecond response for a single sensor (rare for a micro PLC).
2. Program Memory: 1.5 MB vs 64k Steps – The Hidden Fragility
Number: Omron NX1P2 offers 1.5 MB program memory + 2 MB variable memory; Mitsubishi FX5U stores up to 64k steps, which at roughly 12 bytes/step (illustrative) equals ~0.8 MB user code.
Mechanism: Memory is not just capacity – it’s how the runtime handles fragmentation and online changes. The NX1P2 separates program and variable memory, allowing online editing of function blocks without forcing a full download. The FX5U uses a step-based memory model where each online change rebuilds the address table; with many function blocks, this can increase scan time by 15–25% for the first cycle after a change (illustrative).
Worked consequence: A machine builder who releases a mid-project firmware update on an FX5U with 50+ function blocks will see a one-time scan spike that, if unlucky, can trip a safety timeout. The NX1P2’s architecture absorbs that spike by keeping the primary task cycle independent of the variable memory writes. The decision: if your project has >40 function blocks and you do >2 online changes per month, the NX1P2 saves you from an occasional “scan frozen” reboot.
When it reverses: For a fixed, small program (
3. EtherCAT vs CC-Link: Node Latency that Accumulates
Number: NX1P2 supports up to 8 EtherCAT axes with 16 nodes; FX5U uses CC-Link with up to 512 remote I/O but no native motion bus.
Mechanism: EtherCAT uses a “processing on the fly” technique: each slave reads the frame as it passes, adding only a few hundred nanoseconds per node. CC-Link is a master-polled bus where each node adds ~1–2 ms of latency per scan because the master must send a dedicated frame and wait for the reply. For a 4-axis system, the NX1P2 can close the position loop in 2 ms with EtherCAT. The FX5U, even with a 34 ns CPU, must wait ~4 ms just to poll 4 servo drives over CC-Link before the logic can execute.
Worked consequence: A gantry with 4 servos that requires a 10 ms coordinated move: on the NX1P2, the move updates every 2 ms (5 updates per move), giving smooth acceleration. On the FX5U, with ~6 ms bus + scan overhead, only about 1.5 updates per move – the gantry will jerk unless you lower acceleration, costing cycle time. The spec sheet doesn’t show this.
When it reverses: If the application uses only digital I/O and no coordinated motion (e.g., a simple sorting station with 10 sensors and 5 actuators), CC-Link’s polling latency is invisible, and the FX5U’s higher I/O count (up to 512) and lower CPU cost may be the better choice.
⚠ Failure Mode: When the “Faster” CPU Stalls
Consider a scenario: a packaging line with 6 servos, 2 vision cameras, and 40 I/O. The FX5U with 34 ns instruction speed seems perfect – but after adding the vision data transfer over Ethernet (not a fieldbus), the total scan time jumps to 12 ms because the CPU cannot separate the motion task from the communication task. The NX1P2, with its dedicated EtherCAT port and isolated primary task cycle, keeps the motion loop at 2 ms while communication runs in a lower-priority task. One client (illustrative) reported that the FX5U-based machine required a slower camera trigger (30 fps → 12 fps) to avoid motion jitter, destroying the line speed.
When this reverses: For a pure data-logging application with no motion, the FX5U’s 34 ns speed and large I/O count are genuine advantages. The NX1P2’s deterministic cycle adds no value.
📐 The Decision Rule (threshold, not “it depends”)
If your application has ≥2 coordinated motion axes OR any real-time EtherCAT requirement, the Omron NX1P2 is the correct choice – its 2 ms deterministic task cycle and integrated EtherCAT guarantee a motion update rate that the FX5U cannot match, regardless of instruction speed. If the application has 0 coordinated axes and maximum I/O count (>100 points) is the priority, the Mitsubishi FX5U wins on raw speed and expandability. The threshold is binary: motion bus yes/no.
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. Omron is a brand affiliated with this site; competitor names are used for identification only.
