DEEPX unveils AI chip solution

DEEPX, a South Korean AI semiconductor firm, is set to enter the global AI chip race with its comprehensive All-in-4 AI Total Solution being unveiled at CES 2024. The technology consists of four chips for on-device AI such as physical security systems, machine vision, smart mobility, robot platforms and AI servers.

From its CES booth, DEEPX simultaneously showcased how the four AI chips can be applied to various applications along with its DXNN development environment, which can run all four chips using a single software framework — a capability other global AI chip companies do not offer. Since its inception, this has been DEEPX’s strategy for addressing the fragmented on-device AI market, giving the brand an edge in providing customized solutions optimized for customers’ products and applications.

DEEPX Unveils ‘All-in-4 AI Total Solution’ of Four AI Chips to Capture the On-Device AI Market at CES 2024. Source: DEEPX

The All-in-4 AI Total Solution includes DX-V1, DX-V2, DX-M1, and DX-H1. Specializing in vision systems, DX-V1 and DX-V2 are standalone chips. DX-V1 is capable of processing the latest AI algorithms such as Yolov7 in a single camera, and DX-V2 is optimized for autonomous driving, robot vision, and other applications that require processing of 3D sensors outside of the camera. On a single chip, DX-M1 supports real-time AI computation processing of more than 30 frames per second (FPS) for 16 or more channels of multichannel video. For AI servers, DX-H1 is an AI inference-type solution that maximizes performance, power, and cost efficiency compared to general-purpose GPUs dedicated to AI inference.

DEEPX is running an Early Engagement Customer Program (EECP) to give customers early access to the DX-V1, a one-chip solution for smaller camera modules; the DX-M1, an AI accelerator solution for M.2 modules; and DXNN, DEEPX’s unique developer environment. As it strives to democratize on-device AI, the brand is currently offering these solutions in the mass production pre-qualification stage — with more than 40 global customers trialing implementation into their products — and will enter mass production for availability later this year — fully embedded into customers’ products by 2025.