Amazon Redshift has long been a leader in cloud data warehousing, and the latest innovation—RG instances powered by AWS Graviton—takes performance and cost efficiency to new heights. This article answers common questions about these new instances, including how they compare to previous generations, their integrated data lake query engine, and how they support both human and AI-driven workloads.
What are Amazon Redshift RG instances?
Amazon Redshift RG instances are a new family of compute instances built on AWS Graviton processors. They are designed to deliver significantly better performance for data warehouse workloads—up to 2.2 times faster than the earlier RA3 instances—while reducing cost per vCPU by 30%. The RG lineup includes sizes like rg.xlarge (4 vCPU, 32 GB RAM) for small departmental analytics and rg.4xlarge (16 vCPU, 128 GB RAM) for standard production workloads. These instances come with an integrated data lake query engine enabled by default, allowing you to query both Amazon Redshift tables and Amazon S3 data lakes using a single SQL engine. This blend of speed, lower cost, and unified querying makes RG instances ideal for handling high-volume analytics and agentic AI workloads that demand low latency.
How do RG instances compare to RA3 instances?
RG instances represent a major leap over RA3 instances. For example, the ra3.4xlarge has 12 vCPU and 96 GB RAM, while the recommended replacement rg.4xlarge offers 16 vCPU and 128 GB RAM—a 33% increase in both resources. This hardware upgrade, combined with the efficiency of AWS Graviton, yields up to 2.2x faster performance for data warehouse workloads. For data lake queries involving Apache Iceberg, RG instances can be up to 2.4 times faster than RA3, and for Apache Parquet, up to 1.5 times faster. Additionally, the cost per vCPU is 30% lower with RG instances, making them a more economical choice for organizations that need to scale query volumes. The integrated data lake query engine is also a key differentiator, eliminating the need for separate query tools and simplifying operations. AWS recommends using the Pricing Calculator to estimate savings based on your specific workload patterns.
What is the integrated data lake query engine?
The integrated data lake query engine is a built-in capability of Amazon Redshift RG instances that lets you run SQL analytics across both your data warehouse tables and your Amazon S3 data lake from a single engine. This means you no longer need to set up separate query services or move data between systems. The engine is enabled by default on new RG instances and supports popular open table formats like Apache Iceberg and Apache Parquet. It delivers performance improvements of up to 2.4x for Iceberg and up to 1.5x for Parquet compared to RA3 instances. This integration reduces total analytics costs by eliminating data duplication and simplifying operational overhead. Whether your data is structured warehouse tables or diverse lake datasets, you can query it all with consistent SQL syntax and performance optimizations. For more details, see What are RG instances?.
How do RG instances benefit AI and agentic workloads?
AI agents can generate enormous query volumes—far beyond typical human usage—which can quickly drive up operational costs. Amazon Redshift RG instances are specifically designed to meet this challenge. The combination of Graviton-based performance (up to 2.2x faster than RA3) and a lower price per vCPU (30% less) means that even at high query rates, costs remain manageable. The integrated data lake query engine also reduces latency because agents can access both warehouse and lake data without separate hops. In March 2026, Redshift improved new query performance by up to 7x for BI dashboards and ETL workloads, which directly benefits autonomous, goal-seeking AI agents that require near-real-time responses. RG instances thus provide the speed, scalability, and cost efficiency needed to support today's agentic AI analytics without budget surprises.
What performance improvements can I expect with RG instances?
Performance gains with RG instances are substantial across multiple fronts. For traditional data warehouse workloads, you can see up to 2.2 times faster execution compared to RA3 instances. When querying data lakes, the integrated engine accelerates Apache Iceberg queries by up to 2.4x and Apache Parquet queries by up to 1.5x. Additionally, a previous update in March 2026 improved new query speeds for BI dashboards and ETL pipelines by up to 7x, further enhancing low-latency analytics. These improvements stem from the AWS Graviton processor's efficiency and the optimized architecture of RG instances. For real-world numbers, benchmark tests have shown significant reductions in query runtimes for both simple aggregations and complex joins. However, actual performance will depend on your data size, schema, and query patterns. AWS recommends testing with your own workloads using the console or CLI to measure the impact.
How do I get started with RG instances?
Getting started with Amazon Redshift RG instances is straightforward. You can launch a new cluster directly from the AWS Management Console, or use the AWS CLI or API for automation. When creating a new cluster, simply select the RG instance type that fits your workload—such as rg.xlarge for small deployments or rg.4xlarge for production. The integrated data lake query engine is enabled by default, so you can immediately run SQL queries across your S3 data lake. You can also migrate an existing RA3 cluster to RG by taking a snapshot and restoring it on an RG instance. No code changes are required for your existing queries or applications. For cost estimation, use the AWS Pricing Calculator with your specific workload patterns. Once your cluster is running, you can connect using standard SQL clients and start benefiting from improved performance and lower costs right away.
What are the pricing considerations for RG instances?
RG instances offer a 30% lower price per vCPU compared to RA3 instances, making them a cost-effective choice for organizations with heavy query volumes. However, the overall cost also depends on factors like storage (managed storage for RA3 vs. the integrated data lake engine with its own S3 costs), data transfer, and the specific instance size you choose. For example, an rg.4xlarge instance has 16 vCPU and costs less per vCPU than an ra3.4xlarge, while also providing more memory (128 GB vs 96 GB). To accurately estimate your savings, AWS recommends using the Pricing Calculator with your workload patterns—including query mix, data volumes, and AI agent query rates. Additionally, because the integrated data lake query engine reduces the need for separate query services and data movement, you may see further operational savings. Always compare reserved instance options if you have predictable workloads, as they can offer additional discounts over on-demand pricing.