AMD Ryzen AI Halo
AMD Ryzen AI Halo
AMD Ryzen AI Halo mini PC
A miniature workstation with the power of AMD Ryzen AI Max+ for running large language models locally
Local AI development without cloud costs or performance compromises. The AMD Ryzen AI Halo mini PC is built on the AMD Ryzen AI Max+ 395 processor with Zen 5 architecture, integrated AMD Radeon 8060S graphics with 40 RDNA 3.5 compute units, and 128GB of LPDDR5x coherent memory with 256GB/s throughput. This combination allows you to run models with up to 200 billion parameters locally, right on your desk, without relying on cloud infrastructure or recurring GPU instance costs. In tests against the Apple M4 Pro, the AMD Ryzen AI Halo achieves up to 7.3× higher performance in generative AI tasks, while the AMD XDNA 2 NPU with up to 50 TOPS ensures efficient acceleration of AI inference even outside the main GPU. All this in a body measuring just 150 × 150 × 45.4mm and weighing under 1.2kg.
AMD Ryzen AI Max+ 395 5.1GHz
AMD Radeon 8060S
Windows 11 Pro
Key features of the AMD Ryzen AI Halo PC
- The AMD Ryzen AI Max+ 395 processor with Zen 5 architecture, up to 5.1GHz frequency, 16 cores, and 32 threads provides the performance for AI development, fine-tuning, and regular developer workflows.
- The integrated AMD Radeon 8060S graphics with RDNA 3.5 architecture and 40 compute units achieve 60 TFLOPS of FP16 performance for both AI inference and generative graphics tasks.
- 128GB of LPDDR5x coherent memory with 256GB/s throughput lets you load and run AI models with up to 200 billion parameters locally, without needing cloud infrastructure.
- The AMD XDNA 2 NPU with up to 50 TOPS of performance provides hardware acceleration for AI inference alongside the main GPU, increasing the system's overall efficiency and power savings.
- The complete AMD ROCm software platform with support for PyTorch, vLLM, Ollama, LM Studio, and LoRA/QLoRA frameworks ensures Day 0 support for modern AI development tools.
- The 2TB M.2 SSD storage with SED encryption offers fast data access and secure storage for sensitive projects and models.
- Connectivity includes 3× USB-C (one for power), 10Gbps Ethernet, Wi-Fi 7, Bluetooth 5.4, and HDMI 2.1b for connecting to peripherals and external monitors.
- Support for both Windows 11 and Linux gives developers the freedom to choose their platform, while the 120W TDP ensures energy-efficient operation even under a full AI workload in a compact Mini ITX design.
AMD Ryzen AI Halo lets you end your reliance on the cloud
Training and inferencing large AI models usually means either expensive cloud instances or performance compromises. The AMD Ryzen AI Halo eliminates this choice. 128GB of coherent LPDDR5x memory with 256GB/s throughput lets you load and run models with up to 200 billion parameters locally, right on your desk. No per-token charges, no waiting in queues for shared infrastructure. This allows developers, researchers, and data scientists to iterate faster, and sensitive data remains under their control at all times.
Faster than Apple M4 Pro, more efficient than NVIDIA DGX Spark
In generative AI tests, the AMD Ryzen AI Halo achieves up to 7.3× higher performance than the Apple M4 Pro on models like Ace Step or Hunyuan 3D. Compared to the NVIDIA DGX Spark, it offers higher tokens-per-second throughput on models like GLM 4.7 Flash or Qwen 3.6, all while supporting both Linux and Windows 11. The DGX Spark is limited to Linux only; the AMD Ryzen AI Halo gives developers freedom of platform choice and a better price-to-performance ratio.
Software ready from the first boot
Hardware without a functional software ecosystem is just potential. The AMD Ryzen AI Halo comes with full support for AMD ROCm software and Day 0 compatibility with tools like PyTorch, vLLM, Ollama, LM Studio, and LoRA/QLoRA frameworks for fine-tuning. The pre-installed AMD Ryzen AI Halo Developer Center application then serves as a central hub for accessing playbooks, tools, and updates. From first power-on to the first token takes minutes, not hours spent configuring environments.
Supercomputer performance in a book-sized body
With dimensions of 150 × 150 × 45.4mm and a weight of under 1.2kg, competing hardware would barely be enough for office work. The AMD Ryzen AI Halo fits a 16-core Zen 5 processor, integrated RDNA 3.5 graphics with 60 TFLOPS of FP16 performance, and a dedicated NPU with up to 50 TOPS of performance into this compact body, all at a TDP of just 120W. For developers who need an AI supercomputer on the go, in a small office, or as a supplement to their main workstation, this is performance that simply hasn't fit into such a small space before.
Specifications
Hard Drive
Memory
Use
Processor
GPU series/model
Graphics card
Equipment
Colour and design
Outputs
Case
Operating system
Package contents
AI
AI specifications
| Code: | AMD1pc1 |
| Warranty: | 24 months |
| Links: | Producer's Website: |
| Manufacturer information |
AMD Logic Drive, Bianconi Avenue, Citywest, Dublin, D24 T683, Ireland [email protected] 
|