AI Training Server and AI Inference & Fine-tuning Server

The CT100 Server is a high-performance AI training server that supports up to 8 GPU accelerator cards, assembled and provisioned by PerfXLab. It offers an ideal solution for both intelligent and general-purpose computing. With an architecture that boasts independent intellectual property rights, it delivers powerful high-precision and multi-precision hybrid computing capabilities. Its integrated computing and memory design makes it a versatile computing engine for the intelligent era. This server is well-suited for large models, intelligent computing, data centers, general-purpose computing, education, and research applications.

• CUDA Compatibility: CUDA-compatible, the CT100 8-card MetaX C-Series GPGPU Server supports most industrialstandard AI libraries and functions.
• PerfXAPI and PerfXLM: Heterogeneous computing framework and the special framework drastically lowered the users’ cost of acquiring hardware, from 5~50%. Constantly updated massive operation instance library, work together with customers to enter the era of computing power.

CR100 Server is a dedicated AI reference and fine-tuning server with up to 16 Iluvatar MR100 GPU. Suitable for large language models reference and fine-tuning, computer vision, etc. Assembled and provisioned by PerfXLab. It supports mixed-precision inference with FP32, FP16, and INT8, achieving instruction set enhancements, increased computational density, and rebalanced compute and storage.

• CUDA Compatibility: CUDA-compatible, the CR100 16-card MR100 server supports most industrial-standard AI libraries and functions.
• PerfXAPI and PerfXLM: Heterogeneous computing framework and the special framework drastically lowered the users’ cost of acquiring hardware, from 5~50%. Constantly updated massive operation instance library, work together with customers to enter the era of computing power.