The Kria module is designed to ship compute and AI workload processing for bodily AI. | Credit score: AMD
Superior Micro Gadgets Inc. is making a full-stack play for robotics with its new Ryzen AI Embedded X100 sequence and Kria AI Robotics platform. AMD in the present day promised deterministic real-time management, unified CPU–GPU–NPU reminiscence, and an open, non–vendor-locked software program stack that it claimed can outperform NVIDIA Orin and Thor on system-level robotics workloads.
Constructed on a unified reminiscence structure that reduces latency throughout CPU, GPU, and NPU, the Ryzen AI Embedded X100 line targets “agency” and “laborious” real-time management by way of BIOS and Linux optimizations, AMD QoS (high quality of service) options, and Zen-based virtualization.
On prime of the silicon, AMD is rolling out the Kria AI system-on-module and Robotics Growth Package, the open-source AMD Robotics Sophie Suite software program stack, and a curated Robotics Associate Community. The Santa Clara, Calif.-based firm stated they place it to help every little thing from industrial arms and autonomous cell robots (AMRs) to humanoids with a single, scalable platform.
AMD launches Ryzen AI embedded X100 processor
AMD earlier this month introduced the brand new Ryzen AI Embedded X100 household of system-on-a-chip (SoC) parts.
The brand new chips ship a unified CPU–GPU–NPU structure with a give attention to embedded AI use circumstances, stated the corporate. It designed the X100 for high-end robotic and edge AI functions primarily based on its earlier sibling, the P100.
AMD delivered a “shot throughout the bow” of NVIDIA‘s chipset designed for bodily AI functions. The X100 is now AMD’s flagship SoC for robotics functions.
The Ryzen AI Embedded X100 Sequence processor is optimized for bodily AI functions. Credit score: AMD
What X100 provides for robotic builders
The X100 provides unified reminiscence and heterogeneous compute for robotics workloads, and it explicitly reduces knowledge copies. That is key for duties similar to notion, sensor fusion, and planning duties, and unifies all of it on a single platform.
Builders can deploy the X100 into quite a lot of agency or laborious real-time functions. “Agency actual time” is achieved by way of Linux + BIOS optimizations (interrupt latency goal < 7 μs at six-nines). As well as, through the use of AMD QoS, the system can ship L3 cache reservation, reminiscence bandwidth allocation, and thread isolation.
For laborious real-time use circumstances by way of virtualization, AMD really useful the Zen hypervisor, FreeRTOS digital machine, cache coloring, and VM isolation. That is in distinction to conventional dual-device CPU with a discrete GPU structure in a robotic.
Save the date for RoboBusiness 2026Evaluating X100 system-level efficiency with NVIDIA
NVIDIA is at the moment the dominant platform for bodily AI. The discharge of the AMD X100 offers builders an alternative choice to NVIDIA structure, whereas additionally placing value stress on NVIDIA.
“In that market [aerospace and defense market, with a focus on signal-processing workloads], we’re really delivering 3 times the FP32 compute efficiency relative to NVIDIA Thor,” stated Rob Bauer, senior supervisor of product administration and advertising and marketing for the x86 Embedded APU portfolio at AMD.
“They’re fairly extremely optimized round inferencing in that structure, proper?” he stated. “We’re nonetheless delivering floating-point sign processing efficiency right here with the X100 sequence, so that enables us to do some fairly superb issues.”
AMD claimed that its Kria AI SoM delivers 3.4x higher real-time efficiency over NVIDIA Thor T5000, has 1.6x spare compute capability, and provides 2.3x extra agentic AI capability.
AMD X100 efficiency versus NVIDIA for robotics workloads. | Credit score: AMD
AMD introduces the Kria AI SoM
In the present day, AMD introduced the Kria AI SoM (system on module) constructed on the X100 SoC. The structure encompasses a 1.5 ÎĽs control-loop closure and adoption of COM-HPC as a substitute of a proprietary module.
In contrast with its shopper Ryzen merchandise, AMD stated the Kria/X100 line provides prolonged reliability, burn-in, check protection, and match charges essential for an industrial software.
A brand new Kria AI robotics improvement package can be being launched to jump-start builders who wish to check the brand new compute structure.
The dev package features a robotics service card: GMSL, high-speed Ethernet, IMU, Wi-Fi/Bluetooth, and so on. There’s an built-in FPGA on the baseboard for real-time I/O, sensor fusion, and security options.
There are additionally open-source baseboard schematics obtainable to hurry the event of manufacturing designs.
“What makes this [Kria] SoM differentiated is tying it again to the suggestions [from customers],” stated KV Thanjuvar Bhaaskar, robotics lead and senior supervisor at AMD. “They need open, non-vendor-lock-in. So what we did is undertake an open customary. Right here, the usual we adopted is COM HPC. AMD can construct it, and every other firm on the planet can undertake it and construct it. So we don’t wish to have a proprietary kind issue.”
The Kria system helps ROS 2 / Nav2 acceleration on X100, with a give attention to deterministic management and compatibility with present robotics workflows.
AMD ROCm is supported as a typical AI stack from cloud to robotic and delivers robot-optimized AI pipelines.
AMD is positioning Kria as the muse for robotics improvement. | Credit score: AMD
AMD builds robotics companion community for Kria
AMD stated it’s dedicated to constructing a community of robotics companions, primarily sensor and software program firms, to help the X100 and Kria Robotics SoM of their respective merchandise. A few of these companions embrace Open Robotics, Open Navigation, and OpenCV, in addition to sensor and authentic design producer (ODM) companions similar to analog/IMU distributors and system integrators.
The aim of this system is to give attention to curated collaborations whereas supporting full robotics methods with the required sensing, compute, software program, and integrations.
Early prospects embrace Castec International, with its semiconductor fab AMR, and humanoid robotic developer Basis Robotics, which is migrating from Intel and NVIDIA to AMD X100 and future FPGA-based hand management.
The put up AMD unveils Kria module for real-time management, unified reminiscence for robots appeared first on The Robotic Report.
