ADLINK Jetson Thor Platforms
ADLINK unveiled DLAP‑701 edge platforms built around Nvidia Jetson Thor/IGX Thor modules, combining large unified memory, Arm Neoverse CPU cores and industrial I/O aimed at physical‑AI and robotics workloads. The platforms emphasize rugged packaging, high‑speed networking and multi‑model inference capacity rather than tiny single‑function accelerators. (itwire.com)
Most edge artificial intelligence boxes are built to do one narrow job, like spotting defects on a factory line or reading one camera feed. ADLINK’s new DLAP-701 goes the other way: it is built around Nvidia’s Jetson Thor module so one industrial computer can run several models at once on robots, medical scanners, or other sensor-heavy machines. (adlinktech.com) An edge computer is just a machine that does its thinking where the data is created instead of shipping everything to a cloud data center. That matters for a robot arm or a hospital imaging device because camera frames, sensor streams, and control loops arrive in milliseconds, not in the seconds you can waste waiting on the internet. (nvidia.com) The hard part is memory. If one system is watching video, reading lidar, and running a language model for instructions, it needs a big shared workspace instead of tiny separate buckets that force data to be copied back and forth. (developer.nvidia.com) Jetson Thor is Nvidia’s answer to that problem. Nvidia says the Jetson AGX Thor series delivers up to 128 gigabytes of memory and up to 2,070 floating-point-four trillion operations per second of artificial intelligence compute in a 40-watt to 130-watt power range. (developer.nvidia.com) Inside that module, the graphics processor handles the heavy matrix math and the Arm Neoverse central processor handles the operating-system and control work. Nvidia’s Jetson Thor datasheet lists a Blackwell graphics processor with a 14-core Arm Neoverse-V3AE central processor and low-power double-data-rate-five-X memory on the module. (developer.nvidia.com) ADLINK is not selling the module by itself. It is wrapping Jetson Thor in a finished industrial platform with slots, ports, and power design so customers can plug in cameras, storage, network cards, and specialized add-ons without designing a carrier board from scratch. (adlinktech.com) The DLAP-701 is aimed at jobs where one box has to juggle several kinds of data at once. ADLINK says the system targets “multi-model generative artificial intelligence,” multi-sensor processing, humanoid robotics, and medical imaging, which are all workloads that combine large models with real-time inputs. (adlinktech.com) The hardware choices tell you what kind of deployments ADLINK has in mind. The company says the DLAP-701 includes Peripheral Component Interconnect Express Generation 5 slots in x8 and x16 configurations plus Universal Serial Bus 3.2, which is the sort of expansion you need for fast capture cards, storage, or networking in a machine room or factory cabinet. (itwire.com) ADLINK launched the DLAP-701 alongside a tougher DLAP-IGX line built on Nvidia’s Industrial Graphics platform Thor family. Nvidia positions Industrial Graphics, or IGX, as the version for systems that need industrial reliability and functional safety, including robots and medical devices, while Jetson Thor is the smaller embedded route. (nvidia.com, adlinktech.com) That split is the real story here. For years, edge artificial intelligence hardware was sold as a cheap accelerator for one model near one camera, but ADLINK is betting the next wave looks more like a compact server at the edge: one rugged box, several sensors, several models, and enough memory and input-output bandwidth to keep them all running together. (adlinktech.com, nvidia.com)
