DEEPX packs 64-channel video decode into a 30W PCIe card
DEEPX announced the DX-H1 V-NPU, an edge AI semiconductor that integrates video decoding, encoding, and transcoding with AI inference capabilities. This PCIe card is designed for real-time, large-scale video monitoring and processing, offering up to 50 TOPS (INT8) AI performance and supporting 64-channel H.264/265 1080P/30fps video decoding and 32-channel encoding at 30W power consumption.
Key Takeaways
- DX-H1 V-NPU combines AI inference with video decoding, encoding, and transcoding on a single PCIe card.
- The card is rated at 50 TOPS (INT8) and 30W power consumption.
- Codec support includes 64-channel H.264/H.265 1080p/30fps decoding and 32-channel encoding at 30W.
- DEEPX lists 24GB LPDDR5 memory, PCIe Gen3 x16 host interface, and low-profile card dimensions of 68.1 x 164.3 x 18.01 mm.
- The product targets security, smart buildings, smart factories, smart cities, logistics, transportation, and smart retail use cases.
Why It Matters
DX-H1 V-NPU packages AI inference and video codec functions into one 30W PCIe card, which matters for deployments that need local processing without adding separate decoder hardware. DEEPX is explicitly pitching it for real-time, large-scale monitoring, with claims around lower hardware costs, lower electricity costs, and up to four cards per 2U server. The broader signal is that edge video stacks are getting denser and more integrated, with local analytics replacing more of the backhaul to VMS systems. Watch for which server form factors and channel counts DEEPX actually certifies in customer deployments.
Read full article at deepx.ai