Home · Industries · Edge AI

Industry Route · Edge AI

Inference where the cameras are, not where the cloud is

Edge AI moves the model to the site: NVR rooms, camera cabinets, gateways at the network edge. The board needs compute class, an accelerator slot and a way to report back — the three models below quote those exact fields on their datasheets.

RYZEN AI 9 365 · CORE ULTRA · M.2 PCIE 5.0 · NVMe · 4G/5G BACKHAUL

Edge AI cabinet with cameras and inference units

At The Edge

Four things an inference node asks for

  • AI-class compute: the HK800CA’s CPU field reads AMD Ryzen AI 9 365 — the Ryzen AI line for on-device inference at the node.
  • Accelerator and storage bandwidth: the RP7003AW quotes an M.2 2280 slot on PCIe 5.0 x4; the TOP800B adds a PCIe 4.0 x4 slot beside its M.2 2280 PCIe 4.0 x4 NVMe.
  • Big memory envelope: DDR5 96GB on the TOP800B and RP7003AW holds multiple models and large video buffers at once.
  • Backhaul: B-Key with Micro-SIM on the TOP800B and a B-Key slot supporting 4G/5G or SSD on the RP7003AW keep sites talking over cellular.

Deployment Flow

From model to running node

Four steps we walk with system integrators on edge AI roll-outs.

Model & workload

Inference type, batch size and latency budget — what runs on the node defines the CPU class.

Node sizing

CPU, DDR5 capacity and the accelerator slot — PCIe 4.0 or 5.0 M.2 — matched to the model.

Site power & link

DC input range for the cabinet; cellular backhaul slot for sites without a wired uplink.

Fleet rollout

Burn-in tested nodes, locked configurations and a remote service plan for each site.

Questions

Edge AI, answered

The HK800CA carries the AMD Ryzen AI 9 365 on its CPU field — the AI-class Ryzen platform — while the TOP800B runs Core Ultra 5/7 at 28W for desktop-class inference headroom.

The RP7003AW quotes an M.2 2280 slot on PCIe 5.0 x4; the TOP800B adds a PCIe 4.0 x4 slot beside its M.2 2280 PCIe 4.0 x4 NVMe.

The TOP800B carries a B-Key with Micro-SIM for 4G/5G, and the RP7003AW quotes a B-Key slot supporting 4G/5G or SSD — camera sites stay connected without a wired uplink.

Get A Quote

The model runs where the cameras are — that is the edge.

Send the workload, model size and site power — we’ll size the node within 24–48 hours.

Request a Quote Contact Us