Cloud GPU Service offers exceptional parallel computing capabilities, making it ideal for scenarios such as AI training, AI inference, scientific computing, game streaming, cloud rendering, and video encoding/decoding.
AI Training
GPU compute-optimized GN10Xp and GT4 instances are equipped with V100 and A100 GPUs respectively. They have strong double-precision floating-point computing capabilities and are equipped with NVLink bandwidth ranging from 300 to 600 GB/s, which is suitable for large-scale AI training acceleration.
AI Inference
GPU compute-optimized PNV4 and GN7 instances are equipped with A10 and T4 GPUs respectively, providing high-performance single-precision floating-point operation capabilities with excellent power efficiency. They are widely used in both online and offline inference business scenarios.
NPU compute-optimized PTX1, equipped with the Zixiao C100 accelerator card, is well-suited for AI inference scenarios, delivering excellent cost-effectiveness in scenarios such as CV, OCR, and NLP.
Scientific Computing
GPU compute-optimized GN10Xp and GT4 instances are equipped with V100 and A100 GPUs respectively. They feature large video memory, high computing power, and low-latency interconnects between cards, providing large-scale parallel computing capabilities for scientific computing scenarios such as computational fluid dynamics, seismic analysis, and genomics.
Cloud Gaming (GS)
GPU rendering GNV4v and GN7vw instances are equipped with A10 and T4 GPUs, respectively, and feature GRID drivers tailored for cloud gaming rendering scenarios. They offer 1/2 and 1/4 GPU partitioning configurations, aiming to optimize the cost per game while ensuring high-quality visuals and low latency.
GPU rendering GA3 instances, powered by Tencent Cloud’s Star Lake GA01, are specifically designed for game streaming scenarios. They support 1/6 fine-grained partitioning, delivering a highly cost-effective game streaming solution.
Cloud Rendering
GPU rendering GNV4 and GN7vw instances are equipped with A10 and T4 GPUs, respectively, and feature GRID drivers optimized for rendering scenarios. They enable rapid image and video rendering and enhance the efficiency of both online and offline rendering in scenarios such as film and television animation and industrial design.
Video Encoding and Decoding
The video-enhanced GN7vi instances, powered by T4 GPUs, are equipped with Tencent’s self-developed Bright Eyes Video Fusion AI technology, featuring a Top Speed Codec encoding and decoding engine along with an image quality enhancement toolkit. These instances are suitable for both on-demand and live streaming scenarios.