Turbo Oblako, a subsidiary of RTK-TsOD, has expanded its GPU infrastructure by adding Nvidia RTX Pro 6000 Blackwell Server Edition accelerators. The new hardware is designed to support high-performance computing and AI-related tasks, including large language model inference and fine-tuning.
The RTX Pro 6000 features 96 GB of video memory, which allows for hosting AI models on a single GPU. The hardware supports FP4, FP8, FP16, and FP32 computing formats, enabling flexibility for various workloads such as 3D visualization, rendering, and video processing.
Users can deploy configurations ranging from one to eight GPUs per virtual machine with dedicated access. The service provides pre-configured environments with CUDA, PyTorch, TensorFlow, vLLM, and Triton Inference Server to streamline the deployment process for developers.
This new addition complements the existing L40S and H200 accelerators already available in the provider's cloud. According to product director Tatyana Mironova, the RTX Pro 6000 fills a specific niche between general-purpose inference GPUs and high-end hardware for heavy AI training.
