IN Brief:
- IQ-2390 combines application processing, machine vision, AI acceleration, and a dedicated real-time MCU on one industrial platform.
- Dual Gigabit Ethernet with TSN, ECC protection, and operation from -30°C to +115°C support industrial deployment.
- Evaluation kits are expected in early 2027, placing the processor at the design-in rather than broad-production stage.
Qualcomm Technologies has introduced the Dragonwing IQ-2390, a compact industrial processor that combines application computing, machine vision, on-device AI, deterministic networking, and real-time control resources on one platform for HMIs, gateways, PLCs, vision systems, and energy-management equipment.
The IQ-2390 is the first device in Qualcomm’s new Dragonwing IQ2 Series and extends the company’s industrial portfolio towards equipment where integration, lifecycle support, and predictable I/O matter more than headline accelerator performance. Qualcomm is positioning the device below its higher-performance IQ6, IQ8, IQ9, IQX, and IQ10 families, giving developers another point between conventional microcontroller-based control and larger edge-AI computers.
The processor combines one Arm Cortex-A78 core with three Cortex-A55 cores, an Adreno 704 GPU, Hexagon V661 processing resources, and a dedicated RISC-V microcontroller for real-time tasks. Its Hexagon NPU is rated at 1.1 TOPS. That is modest beside premium AI accelerators, but the target applications include operator interfaces, compact machine-vision systems, industrial gateways, scanners, building controllers, and other nodes where a limited amount of inference sits alongside conventional control and communications workloads.
Networking is built around two Gigabit Ethernet interfaces with Time-Sensitive Networking support. TSN gives industrial designers mechanisms for scheduling and controlling time-critical Ethernet traffic rather than relying on best-effort packet delivery alone. The processor also integrates PCIe, USB, storage, serial interfaces, and camera support, reducing the amount of companion logic required around the main SoC.
Qualcomm specifies operation from -30°C to +115°C and includes hardware ECC for memory integrity, a 0.8mm ball-pitch package, hardware-backed security, and a product-longevity programme exceeding ten years. Those details are central to the industrial positioning. Controllers and gateways can remain in service for years after equivalent consumer silicon has been replaced, and changing a processor inside a qualified machine can trigger software, EMC, thermal, and functional revalidation.
The software environment spans Qualcomm Linux based on Yocto, Ubuntu, Android, and Zephyr, allowing developers to separate application and real-time workloads while retaining a common platform. Board-support packages, development tools, documentation, and evaluation hardware are intended to move designs from initial evaluation towards production without forcing OEMs to construct the software stack from scratch.
Third-party hardware is already being prepared around the device. SECO is developing Compact Vision 5 and a single-board computer based on IQ-2390, while Engicam is preparing a system-on-module. Those products matter because an industrial processor normally reaches the market through modules and boards that solve power, memory, connector, thermal, and mechanical integration before an equipment manufacturer writes its own application.
The device sits alongside a broader Dragonwing push into industrial edge computing. The Dragonwing IQ8-based Arduino VENTUNO Q addresses heavier local AI workloads, illustrating how far the portfolio now stretches. Many controllers do not need that level of acceleration; they need enough local inference for vision or anomaly detection while preserving deterministic networking, conventional I/O, and real-time behaviour.
That distinction is important as industrial edge products accumulate AI branding. A machine controller may use a neural network for inspection or condition monitoring, but it still has to boot predictably, service control loops, exchange time-sensitive data, and operate through temperature and vibration conditions that have little resemblance to a desktop development kit. Integration only becomes useful if those conventional requirements survive consolidation onto the same processor.
Qualcomm has opened an early-access programme for IQ-2390 and its Q-2390 commercial counterpart. Evaluation kits are expected in early 2027, so the processor is currently a design-in platform rather than a broadly available production component. OEMs will still need to measure sustained thermal performance, real-time latency, memory bandwidth, peripheral behaviour, and software maturity under their own workloads.
The IQ-2390 therefore broadens Qualcomm’s industrial range downwards rather than extending peak performance upwards. Its engineering case rests on how effectively it can replace several processing and control functions without adding qualification or software complexity. For compact edge systems, that may prove a more useful measure than another increase in TOPS.
