Amazon EC2 G5g instances

Best price performance in Amazon EC2 for Android game streaming

Amazon EC2 G5g instances are powered by AWS Graviton2 processors and feature NVIDIA T4G Tensor Core GPUs to provide the best price performance in Amazon EC2 for graphics workloads such as Android game streaming. They are the first Arm-based instances in a major cloud to feature GPU acceleration.

With G5g instances, game streaming customers can run Android games natively on Arm-based instances, encode the rendered graphics, and stream the game over network to a mobile device. This helps reduce development effort by eliminating the need for cross-compilation or emulation, and lowers the cost-per-stream per hour by up to 30%.

The G5g instances are also a cost-effective platform for machine learning inference. They are ideal for deploying deep learning applications that need access to NVIDIA GPUs and their associated AI libraries.

Introducing Amazon EC2 G5g Instances Powered by AWS Graviton2 | Amazon Web Services (1:06)


Best price performance for Android game streaming

With G5g instances, game streaming customers can run Android games natively on Arm-based instances, encode the rendered graphics, and stream the game over the network to a mobile device, without the need for cross-compilation or emulation. This helps to simplify development effort and reduce time to market, while lowering the cost-per-stream per hour by up to 30%.

Cost-effective machine learning inference

G5g instances are powered by AWS Graviton2 processors and feature NVIDIA T4G Tensor Core GPUs to provide cost-effective machine learning inference. This helps reduce the infrastructure cost of running deep learning models in production, and allows developers and businesses to deploy deep learning capabilities more pervasively in their applications.

Enhanced security and maximized resource efficiency

G5g instances are powered by AWS Graviton2 processors and are built on the AWS Nitro System. AWS Graviton2 processors feature always-on 256-bit DRAM encryption and 50% faster per-core encryption performance compared to first-generation AWS Graviton. The AWS Nitro System is a combination of dedicated hardware and lightweight hypervisor which delivers practically all of the compute and memory resources of the host hardware to your instances for better overall performance and security. G5g instances also support encrypted EBS storage volumes by default.

Extensive software support

AWS Graviton processors, based on the 64-bit Arm architecture, are supported by popular Linux operating systems, including Red Hat Enterprise Linux, SUSE, and Ubuntu. Many popular applications and services for security, monitoring and management, containers, and CI/CD from AWS and independent software vendors (ISVs) also support AWS Graviton2-based instances.


Powered by AWS Graviton2 processors

AWS Graviton2 Processors are based on 64-bit Arm Neoverse cores and custom silicon designed by AWS for optimized performance and cost. AWS Graviton2 Processors deliver 7x more performance, 4x more compute cores, 5x faster memory, and 2x larger caches versus first-generation AWS Graviton Processors.

Featuring NVIDIA T4G Tensor Core GPUs

NVIDIA T4G GPUs deliver enhanced graphics and gaming performance when combined with AWS Graviton2 processors. With support for NVIDIA RTX technology through RT cores, they allow real-time ray-traced rendering, delivering photorealistic objects and environments with physically accurate shadows, reflections, and refractions.

Built on the AWS Nitro System

The AWS Nitro System is a rich collection of building blocks that offloads many of the traditional virtualization functions to dedicated hardware and software to deliver high performance, high availability, and high security while also reducing virtualization overhead.


“At, we want to offer mobile gamers the best experiences regardless of their smartphone hardware specs” said Rosen Sharma, President and CEO “There are two challenges in offering a cloud-based mobile gaming service. The first challenge is technical—for mobile gaming you need powerful GPUs and Arm-based CPUs working together. The second challenge is to be able to offer mobile gaming at a price point which makes the business viable. Amazon EC2 G5g instances are a major leap forward in both these dimensions, and we have been working closely with AWS to run NowCloudOS on top of G5g instances to bring these benefits to mobile gaming."

Rosen Sharma, President and CEO -

“Voice chat in video-game communities is essential for socialization and engagement, but suffers from rampant toxicity that harms players and increases churn. Modulate uses machine learning to proactively detect harassment and other harms in live voice conversations, creating a more safe and inclusive place for gamers to play. Amazon EC2 G5g instances are an essential backbone in Modulate's capability to offer realtime cost-effective voice moderation technology, and have reduced our machine learning inference costs by up to 70% ."

Carter Huffman, CTO and Co-founder - Modulate

“Embedded software development for automotive software-defined functions such as ADAS, autonomous driving, and infotainment faces challenges. The differences between the development environment and the edge slow innovation and time to market due to the need to cross-compile and revalidate the software stack between the hardware platforms. By employing the SOAFEE framework running on Amazon EC2 G5g instances, powered by Arm-based Graviton2 CPUs and Nvidia T4G Tensor Core GPUs, automotive developers can be confident in having close alignment, or environmental parity, between car and cloud.”

John Heinlein, Vice President, Automotive Partnerships - North America, Arm

Anbox Cloud
“Canonical launched the Anbox Cloud Appliance in AWS Marketplace to give developers a simple way to prototype, develop and run Android in the cloud. Anbox Cloud now supports Amazon EC2 G5g instances and provides a complete solution that works seamlessly with the Android software stack to virtualize mobile apps, including games, and stream them securely at scale to mobile devices. With Anbox Cloud and G5g instances, customers can stream up to 55 individual sessions at 720p and 30 fps with up to 20% lower cost over comparable x86-based GPU instances.”

Simon Fels, Engineering Manager for Anbox - Canonical

Corellium logo
“The challenge with developing software for mobile devices is that physical devices don’t scale, but emulators don’t provide the same performance or accuracy. Corellium creates virtual models of Arm-based devices and runs them on Arm-based servers in the cloud. With Amazon EC2 G5g instances, mobile developers can leverage Corellium software to create Arm-native virtual Android devices that combine the fidelity and performance of a real device with the convenience and scale of the cloud.”

Chris Wade, CTO - Corellium

“Genymobile provides Genymotion, an Android as a service solution at an affordable price. With the release of Amazon EC2 G5g instances, we are able to go a step further by providing developers an Arm-native environment with hardware acceleration, enabling GPU-accelerated rendering and video encoding. Genymotion on G5g is our fastest and most accurate Android image so far. On top of that, our low-resource footprint implementation enables us to run over 60 Virtual Device instances on a single G5g server!”

Jean-Charles Leneveu, Product Manager - Genymobile

Tier IV
“Tier IV is the leading developer of Autoware, the world's first open-source software for autonomous driving. As part of our CI/CD loop, we were previously limited to evaluating the performance of our Yolo4 machine learning model on our in-car ROScube hardware. This has restricted the number of simulations that could be run concurrently. Since G5g instances feature the Arm-based Graviton2 processor and GPUs, we are able to leverage these instances to scale our CI/CD pipeline and run significantly more simulations iterations simultaneously, with unprecedented parity between cloud and embedded systems”

Eiji Sekiya, Director of Engineering, Tier IV

Ready Player One Tech
“At WePlay, we aim to build a cloud-native, decentralized, and character-identifiable virtual worlds. We are a cloud gaming platform, but not merely a cloud gaming platform. We utilize cloud gaming technology to bring all mobile gamers an unparalleled social experience. The biggest challenge we encountered was the compatibility of Arm-based apk files with today's available x86-based GPU-enabled instances. To translate Arm-based files into x86 and run them on x86 instances is not only more costly but ineffective. With the g5g instances, we can run Android games natively without a translation layer. We reduced our cost by about 30% compared to the g4dn.” 

Adrian Wang, CEO - Ready Player One Tech

Product Details

Instance Name NVIDIA T4G Tensor Core GPU GPU Memory (GB) vCPUs Memory (GB) EBS Bandwidth (Gbps) Network Bandwidth (Gbps) On Demand Price/hr* 1-yr ISP Effective Hourly (Linux) 3-yr ISP Effective Hourly (Linux
g5g.xlarge 1 16 4 8 Up to 3.5 Up to 10 0.42 0.252 0.168
g5g.2xlarge 1 16 8 16 Up to 3.5 Up to 10 0.556 0.334 0.222
g5g.4xlarge 1 16 16 32 Up to 3.5 Up to 10 0.828 0.497 0.331
g5g.8xlarge 1 16 32 64 9 12 1.372 0.823 0.549
g5g.16xlarge 2 32 64 128 19 25 2.744 1.646 1.098
g5g.metal 2 32 64 128 19 25 2.744 1.646 1.098




NVIDIA T4G tensor core GPU: Accelerated graphics and AI for the ARM-based AWS cloud


AWS On Air ft. Amazon EC2 G5g instances | AWS Events (15:15)

Getting Started with G5g Instances

Using Pre-Built AMIs from AWS
Using Amazon Deep Learning AMIs, you can quickly launch Amazon EC2 instances pre-installed with popular deep learning frameworks and interfaces such as TensorFlow and PyTorch to run sophisticated, custom AI models, experiment with new algorithms, or to learn new skills and techniques. To learn more, visit the Amazon Deep Learning AMIs product page.

Leveraging NVIDIA AI from the NVIDIA NGC Catalog
The NVIDIA Deep Learning AMI on AWS Marketplace comes preconfigured with all the necessary NVIDIA drivers, libraries, and dependencies to run Arm-enabled software from the NGC catalog, NVIDIA’s hub of GPU-optimized software. Learn more about using NVIDIA’s NGC Catalog on AWS.

Building your own custom AMI with NVIDIA drivers
The NVIDIA Data Center drivers provide the best performance for highly intensive computational workloads often found in deep learning, data science, and high performance computing (HPC). To learn how to install these drivers on your instance, click here.

Get started with AWS

Sign up for an AWS account

Sign up for an AWS account

Instantly get access to the AWS Free Tier.

Learn with simple tutorials

Learn with 10-minute tutorials

Explore and learn with simple tutorials.

Start building with EC2 in the console

Start building in the console

Begin building with step-by-step guides to help you launch your AWS project.