
AITOP ATOM Host *1 Power Adapter *1 (National Standard) Quick Guide *1
NVIDIA GB10 Grace Blackwell Superchip
20-core ARM CPU
Fifth-generation Tensor Cores
NVIDIA NVLinkâą C2C for high-speed interconnect
Up to 1000 AI TOPS (Tera Operations Per Second)
Supports models with up to 200 billion parameters
Achieves up to 10 quadrillion floating-point operations per second (FLOPS)

AMD Ryzen AI Max+ 395
CN„14,989.00

i914900K (24 cores, 3.25.6 GHz)
CN„22,799.00
128GB unified memory
Supports up to 4TB NVMe storage capacity
NVIDIA ConnectX-7 SmartNIC
High bandwidth, low latency communication
RJ45 10G Ethernet support
NVIDIA AI software stack technology
Compact and refined form factor suitable for desktop environments
Ideal for local AI development including prototyping, model fine-tuning, inference, data science, and edge AI applications
Enables local AI model optimization with data privacy and accelerated iteration without reliance on cloud services
Supports seamless expansion by connecting two AITOP ATOM systems for heavier AI workloads
Suitable for edge deployment supporting machine learning, computer vision, and multimodal applications
High-performance and reliable operation in a compact architecture
Supports complete data science workflows with strong data privacy protection