DFI EC900-QCS 8-Core Qualcomm QCS6490 Edge AI Computer with 12 TOPS NPU and DIN Rail Mount

Autonomous mobile robots, carts and small machines often need local AI inference without the power budget of an x86 box PC. The EC900-QCS from DFI is an edge AI computer for that role: a small fanless unit that clips to a DIN rail and runs from a 5 V DC supply. DFI aims it at AMR and box PC applications, and as a preliminary product it is available to order now.

Compute comes from the Qualcomm QCS6490, with eight Kryo 670 cores at up to 2.7 GHz and 12 TOPS of INT8 AI performance, so detection and classification models can run on the device rather than upstream. It is the same processor DFI uses on its QCS051 series board. Memory is 4 GB or 8 GB of LPDDR5x on board, storage is 128 GB of UFS 3.1 by default with a microSD slot, and the EC900-QCS ships with Yocto Linux on kernel 6.6, with Ubuntu 20.04 available on request.

External I/O covers one GbE port on RJ45, one RS-232 port on DB9, two USB 3.1 Gen 1 Type-A ports, an HDMI 1.4 output with audio (1080p under Linux), an audio jack and 8-bit DIO on a terminal block, set to six lines by default. Inside are three USB 2.0 headers, two more RS-232 headers and a USB-C port used only for image download. An M.2 3052 B key slot with a Nano SIM takes an optional LTE module, an M.2 2230 E key slot takes optional Wi-Fi, and there are five antenna openings.

Power draw is 3 W typical and 7.25 W maximum through a two-pin 5 V DC jack. The EC900-QCS is rated from 0 °C to 60 °C, with vibration testing to IEC 60068-2-64, and DFI lists CE, FCC Class A, UKCA and RoHS compliance. We supply it configured and tested, with the Linux image, storage and any wireless modules fitted. For Arm compute on a 9 to 36 V DC input with two GbE ports, see the EC900-8MM-4GB, or the EC910-8MM-4GB for four serial ports.