GPU Cloud Computing (GPU) is an elastic computing service that provides GPU computing power. It has super computing power and serves multiple application scenarios such as deep learning, scientific computing, graphic visualization, and video processing.As the first cloud service provider in Asia Pacific, Alibaba Cloud is ready to provide you with computing power at your fingertips, effectively relieving computing pressure, improving your business efficiency, and helping you improve enterprise competitiveness.
Up to 4 * NVIDIA A10-24G card;Training application, AI reasoning, scientific computing and other business scenarios applicable to artificial intelligence algorithms
Up to 8 * NVIDIA V100-16G card;Business scenarios such as training/reasoning applications and scientific computing applicable to artificial intelligence algorithms
Up to 4 * NVIDIA T4-16G card;Applicable to image/voice recognition, cloud real-time rendering and other business scenarios
example
4-core 15G
Purchase duration
1 month
number
one
Enquiry in progress
GPU software to improve your computing efficiency
Alibaba Cloud provides you with proprietary auxiliary tools
AIACC-Training
AIACC-Inference
FastGPU
cGPU
EAIS
DPCA AI acceleration training engine
AIACC Training is a DPCA AI accelerated training engine launched by Alibaba Cloud. It has been deeply optimized for Alibaba Cloud environment, which can significantly improve the efficiency of distributed training and significantly improve the utilization of network bandwidth.At present, AIACC Training has set two world records: Stanford Dawnbench Imagenet has the fastest training speed in the world Stanford Dawnbench Imagenet has the lowest training cost in the world
Capable of providing
Support four mainstream frameworks
Tensorflow, Pytorch, MXNet and Caffe four distributed training frameworks
50% to 300% better performance
Network model suitable for bandwidth density
High performance communication between single machine multi card/multi machine multi card
Support FP16 gradient compression and mixed precision compression
API Extension of MXNet
Support data+model parallelism of insightface type
AIACC Inference is a DPCA AI accelerated reasoning engine launched by Alibaba Cloud. It has been deeply optimized for Alibaba Cloud environment, which can significantly improve GPU utilization and reasoning business performance.At present, AIACC Conference has set two world records: Stanford Dawnbench Imagenet has the lowest reasoning delay, ranking first in the world Stanford Dawnbench Imagenet has the lowest reasoning cost in the world
Capable of providing
Support multiple frameworks
Tensorflow, Pytorch, MXNet and other deep learning frameworks that can derive ONNX models for GPU inference optimization
FastGPU is a set of fast deployment tools for AliCloud GPU instance clusters. It helps users deploy GPU computing resources on AliCloud with one click, making it easy to adapt, deploy and run anywhere. It provides users with a time-saving, economical and convenient solution for building AliCloud GPU instance clusters instantly.
Capable of providing
Rapid deployment
Provide convenient APIs to quickly deploy offline training/reasoning scripts in AliCloud GPU instance clusters
Convenient management
Provides convenient command line tools for managing the running status and life cycle of AliCloud GPU instance clusters
Efficient and time-saving
Users do not need to perform tedious deployment operations such as computing, storage, and networking related to Alibaba Cloud's IAAS layer. When they obtain cluster resources, they automatically obtain the corresponding environment
A software that creates and runs multiple GPU containers on the GPU, isolates GPU resources, and enables multiple containers to share a GPU.CGPU can run multiple containers on a single graphics card, isolate GPU applications between multiple containers, and improve the utilization of GPU hardware resources.
Capable of providing
GPU sharding
Divide GPUs to improve GPU utilization
Shared GPU
Multiple AI applications share GPU to save costs
Flexible matching
Flexible segmentation of computing power and video memory to meet application requirements
Alibaba Elastic Accelerated Computing Instances (EAIS) is an elastic accelerated computing instance that can flexibly add GPU acceleration resources to the Alibaba ECS instance.You can select the most suitable ECS instance in Alibaba Cloud according to the overall computing and memory requirements of your application, and then configure the required level of GPU driver reasoning acceleration to effectively use resources and save costs as much as 50%.
Capable of providing
50% reduction in reasoning costs
Satisfy the user to select the most suitable ECS overall calculation instance type, and separately formulate the required GPU reasoning acceleration level, which reduces the cost of GPU reasoning instances by 50%
Flexible CPU and GPU ratio
Flexible allocation of CPU and GPU resources according to user needs, and accurate acquisition of user needs
Elastic expansion
Easily expand and reduce the amount of reasoning acceleration, helping users pay only for the resources they need
AliCloud GPU ECS is equipped with the industry's super powerful GPU computing card. Combined with the high-performance CPU platform, a single instance can provide mixed precision computing performance of up to 5PFLOPS.
Excellent network performance
The VPC network of the AliCloud GPU ECS instance supports a maximum of 24 million PPS and 160Gbit/s intranet bandwidth.
Flexible purchase mode
It supports flexible resource payment modes, including monthly package, pay as you go, preemptive instance, reserved instance coupon, and storage capacity unit package.You can purchase as needed to avoid resource waste.
Application scenarios
Deep learning
Graphic visualization
video processing
Scientific computing
Strong training ability, excellent reasoning ability
Deep learning has made significant breakthroughs and been widely used in industry. In order to enable computers to "understand" human language, natural language processing has been widely used in text classification, recommendation systems and other directions with the significant progress of deep learning;Speech synthesis and speech recognition are also widely used in intelligent question answering and chat robots.As the most mature field of deep learning applications, the image field, with the help of Alibaba Cloud's powerful GPU computing power, can identify images more accurately, improve the accuracy, and also improve the operating efficiency.
Able to solve
Powerful training ability
The latest GPU has achieved excellent acceleration in AI and data analysis on various scales to meet extremely severe computing challenges.At the same time, Alibaba Cloud provides a variety of GPU instance types, providing flexibility for different computing power and scenario requirements.
Excellent reasoning ability
It provides industry-leading reasoning ability. The latest GPU has realized powerful diversified uses through full range of precision (FP32, FP16, INT8 to INT4) acceleration.
Industry leading solutions for engineering simulation and analysis can provide high performance, scalability and enterprise class reliability.With the help of GPU's huge video memory capacity and super performance, use the required computing power to perform complex simulations and solve challenging problems.
Able to solve
Optimized solution
Responsible for CFD modeling, greatly shortening the solution time
Electronic design of accelerated computational electromagnetics
When designing high-performance electronic products and components, simulate electromagnetic performance and accurately predict electromagnetic radiation, interference and signal transmission
Engineering simulation
In the cloud, improve work efficiency and enable IT departments to save budget by virtualizing applications
In the field of video processing, there are also problems of large amount of computation and long processing time. GPU can be used for optimization because of its high parallelism of computing tasks.At present, GPU is mainly used for large-scale high-definition video transcoding, 4K/8K high-definition live broadcast, multi person video conference, source repair and other fields
Able to solve
High performance
High degree of optimization to improve computing performance
Strong calculation force
Fast processing of multi frame data, providing computing power for processing a large number of computing tasks
GPU has played a powerful role in scientific computing fields requiring large-scale parallel computing, such as weather prediction, oil and gas exploration, molecular dynamics, etc.By providing the computing power of large-scale floating point operations, and perfectly combining with elastic computing, efficient computing performance can be provided online or offline
Able to solve
Elastic expansion
Elastic capacity expansion combined with ESS and SLB
Super strong calculation force
Provide the latest model GPU and the most convenient deployment method to meet the powerful computing needs of scientific computing
Shenshi TechnologyThe Alibaba Cloud based elastic supply solution, combined with preemptive GPU instances, can quickly obtain the required computing resources at a low cost.
Speak fluentlyThe cloud GPU can be purchased on demand, and the hardware can be upgraded at a relatively low cost.The rich cloud ecosystem also reduces the cost of system construction and operation and maintenance.
Maverick TranslationThe flexible use scheme of GPU provides flexible and sufficient GPU computing resources, which can be applied at any time.
