Amazon Redshift RG Instances: Graviton-Powered Performance for Data Warehousing and Lake Queries

Amazon Redshift continues to innovate with the introduction of RG instances, a new family powered by AWS Graviton processors. These instances offer significant performance improvements and cost savings for data warehouse workloads, while also integrating a query engine that seamlessly runs SQL analytics across both data warehouse tables and data lakes. This Q&A explores key features and benefits of the new RG instances.

What are Amazon Redshift RG instances?

Amazon Redshift RG instances represent the latest generation of compute-optimized instances for the cloud data warehouse service. Powered by AWS Graviton processors, they deliver up to 2.2x faster performance than the previous RA3 instances while reducing cost per vCPU by 30%. The RG family includes the rg.xlarge and rg.4xlarge sizes, designed for workloads ranging from small departmental analytics to production use. These instances also come with an integrated data lake query engine that allows you to run SQL queries across both Redshift tables and Amazon S3 data lakes (Apache Iceberg and Parquet formats) from a single engine, achieving up to 2.4x faster performance for Iceberg and 1.5x for Parquet compared to RA3. This blend of speed, cost efficiency, and unified querying makes RG instances ideal for modern analytics and agentic AI workloads.

Amazon Redshift RG Instances: Graviton-Powered Performance for Data Warehousing and Lake Queries — Source: aws.amazon.com

How do RG instances compare to RA3 instances?

The new RG instances are designed as a direct upgrade to the current RA3 family. For example, the ra3.xlplus is replaced by rg.xlarge (4 vCPU, 32 GB RAM), while the ra3.4xlarge becomes rg.4xlarge with an improved vCPU count (16 vs 12) and memory (128 GB vs 96 GB). This gives a 1.33:1 ratio in both compute and memory, offering more resources at a lower price per vCPU. Benchmark data shows that RG instances run data warehouse workloads up to 2.2x faster. Additionally, the integrated data lake query engine boosts performance on Apache Iceberg queries by up to 2.4x and on Parquet by up to 1.5x. For customers running mixed warehouse and lake workloads, this translates to significant cost savings and simplified operations. Use the AWS Pricing Calculator to estimate specific savings based on your workload patterns.

What is the integrated data lake query engine?

The integrated data lake query engine is a built-in feature of RG instances that enables you to run SQL analytics across both your data warehouse tables and data lakes stored in Amazon S3—all from a single engine. This eliminates the need for separate query tools or data movement between environments. The engine supports open table formats like Apache Iceberg and Apache Parquet, and it is enabled by default on new RG instances. Performance improvements are substantial: up to 2.4x faster for Iceberg and 1.5x for Parquet compared to RA3 instances. This unified approach reduces operational overhead and total cost for customers who need to query structured warehouse data alongside diverse, cost-effective data lake datasets.

How do RG instances benefit AI and analytics workloads?

AI agents and modern analytics demand high query volumes and low latency. RG instances are specifically designed to handle these requirements. With up to 2.2x faster performance and 30% lower cost per vCPU, they can efficiently scale to support both human-driven BI dashboards and autonomous AI agents. The integrated data lake query engine further accelerates queries on Iceberg and Parquet datasets, which are common in machine learning pipelines. For example, Redshift already improved query speed for dashboards and ETL by up to 7x in March 2026, and RG instances build on that foundation. This makes them well-suited for near-real-time analytics, goal-seeking AI agents, and other latency-sensitive applications.

How can I get started with RG instances?

Getting started is straightforward. You can launch new Amazon Redshift clusters using RG instances or migrate existing RA3 clusters through the AWS Management Console, AWS CLI, or AWS API. The integrated data lake query engine is enabled by default, so no additional configuration is needed. To estimate cost savings for your specific workloads, use the AWS Pricing Calculator. For migration guidance, refer to the Amazon Redshift documentation.

What are the key use cases for RG instances?

RG instances are ideal for a range of analytics scenarios. For smaller deployments, the rg.xlarge (4 vCPU, 32 GB) is perfect for departmental analytics and small clusters. The rg.4xlarge (16 vCPU, 128 GB) supports standard production workloads with medium data volumes. Both sizes excel in situations requiring high query concurrency, such as BI dashboards, ETL pipelines, and near-real-time analytics. Additionally, the integrated data lake query engine makes them a strong choice for organizations that want to combine structured warehouse data with data lake lakes for cost-effective storage and diverse analytics—especially those using Apache Iceberg or Parquet formats. Agentic AI workloads, which generate high query volumes, also benefit from the improved price-performance.

What cost savings can I expect with RG instances?

On a per-vCPU basis, RG instances are priced 30% lower than RA3 instances. For example, if you are currently using ra3.4xlarge instances, migrating to rg.4xlarge gives you more vCPUs and memory at a lower price. The exact savings depend on your workload patterns, including query volume, data lake usage, and instance count. Because RG instances also run queries faster (up to 2.2x for warehouse and up to 2.4x for Iceberg), you may need fewer compute resources to achieve the same performance, further reducing costs. Use the AWS Pricing Calculator to input your specific details and get an accurate estimate. Overall, combined savings from lower unit cost and higher efficiency can be substantial.

Tags: