Radxa, an embedded computing specialist, has launched AICore SG2300x, a system-on-module (SOM) designed for Edge AI projects that can power substantial GenAI models.

The AICore SG2300x is built around the SOPHON SG2300x system-on-chip and includes eight Arm Cortex-A53 processor cores. Key features of this product include a dedicated tensor processing unit claiming 32 tera-operations per second, 16 GB of LPDDR4X RAM, and a 64 GB eMMC storage module. Additionally, it has a PCI Express 3.0 four-lane root complex and provides hardware decoding for video channels.

Radxa claims that with AICore SG2300x, users can run popular GenAI models directly on the device. The company also indicates that the module can support various machine-learning frameworks like TensorFlow and PyTorch. Furthermore, Radxa plans to release an open-source carrier board design and a design guide for those wishing to construct their own.