Due to the high-efficiency computing needs of enterprises driven by the generative AI craze, China Telecom officially announced the launch of the "hicloud AI Computing Power Cloud" GPU Cloud Rental Service, providing NVIDIA GPU Cloud compu...
Due to the high-efficiency computing needs of enterprises driven by the generative AI craze, China Telecom officially announced the launch of the "hicloud AI Computing Power Cloud" GPU Cloud Rental Service, providing NVIDIA GPU Cloud computing power resources, with diversified specifications and flexible lease periods as the main features, providing sliced GPU specifications, breaking through the market's general restrictions on renting entire machines, and matching with China Telecom's independently developed AI computing power platform, it helps enterprises quickly deploy AI training and promotion environments, and accelerate the transformation and implementation of smart applications.
China Communications Technology pointed out that there are high-end spots that require renting a whole server (8 GPUs) in one go in the market. "hicloud AI computing power cloud" uses NVIDIA MIG (Multi-Instance GPU, multiple execution individual GPUs) technology to divide the single GPU into independent slice combinations, each slice has a dedicated computing core, memory and cache.
Flexibility specifications: Users can choose to rent 1/4 or 1/2 slice GPUs according to the model training specifications and budgets, or adopt 1, 2, 4, and 8 slice GPUs. Benefit optimization: significantly reduces the company's early hardware investment costs, avoid resource waste caused by indeterminate model introduction, and effectively combines efficiency and cost-effectiveness."hicloud AI computing power cloud" provides services with PaaS model. Developers do not need to set up and install a complex bottom layer system. They only need to complete model training, recommendations, project management and teamwork through a direct web interface. This model can effectively reduce the technical gate and communication costs of the development team and accelerate the development process of the AI model.
Considering the rapid iteration of AI models and project requirements, this service adopts a time-based flexible lease design, and can start renting within a minimum of 1 hour, allowing enterprises to more flexibly master the R&D rhythm and budget configuration.
