Specifications
CPU
Intel Core Ultra 7 358H (12Xe, 16 cores, 1.9-4.8GHz) · Intel Core Ultra 5 325 (4Xe, 8 cores, 2.1-4.5GHz)
AI Capability
Up to 180 TOPS (Ultra 7 358H) · up to 95 TOPS (Ultra 5 325)
Graphics
Intel Xe 3 LPG integrated graphics
Memory
DDR5-6400/7200 CSODIMM up to 128GB · 2x 262-pin SO-DIMM (no ECC)
Storage
PCIe Gen5 NVMe via M.2 Key M
Expansion
M.2 E Key, B Key and M Key; Advantech iDoor and PoE modules
LAN
Up to 4x Intel GbE / 2.5GbE
Serial / Fieldbus
6x serial ports, 2x CANBus, 8-bit DIO
USB
4x USB 3.2 + 2x USB 2.0
Display
3 displays — 2x 4K HDMI + 1x 4K USB Type-C
Power
12-24V wide-range DC input
Operating Temp
-20°C to 60°C extended (-40°C to 85°C storage); fanless
Security
On-board TPM
Software
Advantech Edge AI SDK and WEDA — MLOps, digital twin interface, container/AI model management, remote deployment, silicon-optimised container library
Dimensions
260 x 145 x 60 mm
OS
Windows 11 LTSC / Ubuntu 24.04
Overview
The Advantech ARK-2252 is a high-performance fanless embedded box PC built on Intel Core Ultra Series 3 processors (codenamed Panther Lake). The top configuration pairs a Core Ultra 7 358H with Intel Xe 3 LPG graphics to reach up to 180 TOPS of AI capability, while the Core Ultra 5 325 variant delivers up to 95 TOPS — all without a fan or a discrete GPU. Memory goes to 128GB of DDR5-6400/7200 CSODIMM across two sockets, and storage uses PCIe Gen5 NVMe. I/O is unusually dense for the 260 x 145 x 60mm chassis: up to four 2.5GbE LAN ports, six serial ports, dual CANBus, 8-bit programmable DIO, six USB ports and triple 4K display output. Modular M.2 and Advantech iDoor expansion add wireless, vision and PoE as needed.
Key Benefits
GPU-class inference, fanless: up to 180 TOPS from the CPU, Xe 3 GPU and NPU together means vision AI, VLM inference and robotics perception run in a sealed, silent chassis suitable for dusty or acoustic-sensitive environments. Built for industrial integration: four LAN ports, six COM ports, dual CANBus and 8-bit DIO let the ARK-2252 act as the single controller for a machine cell — cameras, PLCs, drives and sensors — rather than needing a separate gateway. Wide thermal and voltage envelope: -20°C to 60°C operation and 12-24V input suit cabinets, vehicles and outdoor enclosures. Managed at scale: the Advantech Edge AI SDK and WEDA platform provide MLOps, a digital twin interface, container and AI model management, remote deployment and a silicon-optimised ready-to-dev container library, which is the difference between piloting one unit and operating a fleet.Applications
Vision AI and machine inspection, intelligent robotics and AMR controllers, factory automation and machine-cell control, video analytics with PoE cameras, AGV and vehicle computing, self-service and kiosk systems, and edge deployment of containerised AI models across distributed sites.
Request a Quote — ADVANTECH ARK-2252 — FANLESS BOX PC, INTEL CORE ULTRA SERIES 3, 180 TOPS
QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.
Get Your Quote →