Wise Cube

LUISUAN's newly launched Wise Cube is an efficient, convenient solution with integrated hardware and software. Its modular design allows it to be customised on demand and expanded flexibly. It integrates many core modules while being widely applicable in scientific research, disasters rescue, mine petroleum, geological prospecting, mobile police affairs and other areas.

It's designed together with its own protective container, essentially like a portable suitcase that is outstandingly anti-pressure and anti-shock. Both structural integrity and security of its internal data can be perfectly assured with airborne, land transport and any other forms of transportation. 

Corresponding servers, systems, softwares etc. for each core module of this solution are available for users to integrate as optional settings:

For GPU computing power server, users could choose from a pool of brands including NVIDIA, HUAWEI Ascend, Cambricon, Moore Threads, SOPHGO, MetaX, TsingMicro, KunLunXin, Emflame etc., to respond to complex computational tasks and achieve AI training and inference. 

For interchangeable CPU server, X86, Hygon, HUAWEI Kunpeng, Phytium, Sunway, Zhaoxin, Loongson etc. are available for optional settings, so as to attain operations of diverse application software and services for clusters management. 

For scattered file system and network switches, users could choose from unified storage services such as support files, objects, Block Storage, HDFS etc., hence efficient data storage and management, satisfying the needs of diverse scenarios. 

For high efficiency full flash EBOF, LUISUAN GP5000/6000 full flash EBOF is available for settings, which provides flexible NVMe-oF storage expansion services, meeting the needs of high-speed access and storage for massive amount of data. 

For artificial intelligent platforms, PyTorch, TensorFlow, PaddlePaddle, Transformer etc. are available as basic softwares. Triton, vLLM, DeepSpeed, Megatron etc. are available as inference training frameworks. Deep learning platform, AIOps, RLHF etc. are available as pre-install applications. DeepSeek-V3-671B, DeepSeek-R1-Distrill-Llama-70B, LLaMA-7B etc. are available as pre-install models for optional settings. 

Three application scenarios are launched:

LUISUAN-DeepSeek Training & Inference AIO (DeepSeek's API cache technology + "storage for computing power" concept + LUISUAN's GP5000/6000 full flash EBOF = higher efficiency, lower costs, out-of-the-box AI solution)

Data Safe Case (Due to issues like long construction period of computing centre, scarcity of NVIDIA GPU card etc., original AI model, sample data, checkpoint process data gathered, finely trimmed AI LLM are all stored inside Wise Cube, ensuring "eternal" data security, allowing companies engage in LLM training and optimisation, hence competent in the fierce competition)

Cross-origin data transfer (2023 domestic corporation data transfer scale↑230%, e.g. 1000M direct line, 1PB ⟼ 99.4 days; Wise Cube, 1PB ⟼ 9.6 hrs)