Desktop and Application Streaming

Amazon WorkSpaces launches Graphics G6, Gr6, and G6f bundles

We are excited to announce the availability of new Amazon WorkSpaces G6, Gr6, and G6f Graphics bundles for WorkSpaces Personal and WorkSpaces Core. These bundles are built on Amazon EC2 G6, Gr6, and G6f instances respectively and powered by NVIDIA L4 Tensor Core GPUs. When running representative graphics and ML inference workloads, the G6 family delivers up to 2x better performance for deep learning and graphics workloads compared to G4dn instances.

The G6 family also introduces fractional GPU options that make GPU-accelerated virtual desktops accessible at a fraction of the previous cost. For more information on pricing, please see our pricing pages for WorkSpaces Personal and WorkSpaces Core.

Customers such as Halliburton have validated the performance of G6 WorkSpaces, and particularly the high-memory Gr6 variants, as an upgrade for their workflows.

“Halliburton‑Landmark participated in the AWS Gr6 WorkSpaces preview as part of our iEnergy Subsurface offering. The G6 family demonstrated improved performance for subsurface workloads, enabling a smoother experience for working with large seismic datasets.”

What’s New

Amazon WorkSpaces now offers three new GPU-accelerated bundle families with 12 different configurations. As of this publication, these are available in 13 AWS Regions: US East (N. Virginia), US West (Oregon), Canada (Central), Europe (Paris, Frankfurt, London), Asia Pacific (Tokyo, Mumbai, Sydney, Seoul), South America (São Paulo), and AWS GovCloud (US-West and US-East).

  • Graphics.g6f – Fractional GPU bundles (1/8 to 1/2 GPU) that deliver GPU acceleration at significantly lower price points, perfect for users who need graphics capabilities but don’t require a full GPU. Available in four sizes from 2 vCPU/8GiB RAM to 16 vCPU/64 GiB RAM.
  • Graphics.g6 – Full GPU bundles optimized for graphics-intensive applications, offering balanced compute and memory with one complete NVIDIA L4 GPU per instance. Available in five sizes from 4 vCPU/16 GiB RAM to 64 vCPU/128 GiB RAM.
  • Graphics.gr6 and Graphics.gr6f – High-memory bundles with a 1:8 vCPU to RAM ratio, designed for memory-intensive applications like large-scale CAD models and simulations. These bundles provide up to 256 GB of system memory.

All bundles are powered by NVIDIA L4 Tensor Core GPUs, which feature fourth-generation Tensor Cores and third-generation RT Cores with support for NVIDIA RTX technology. Each full GPU instance type (g6 and gr6) provides 24 GB of VRAM. This is 50% more than the prior generation g4dn WorkSpaces, which offered a maximum of 16GB VRAM. This increase in VRAM helps enable smooth performance for demanding applications like 3D modeling, video editing, and real-time rendering.

Lower Entry Point for GPU Workloads

One of the most significant improvements is the introduction of fractional GPU bundles, which reduce the entry cost for GPU-accelerated WorkSpaces by ~64%. Here is a comparison, showing the price for the entry level GPU model of two generations. The comparison shows a Windows WorkSpace (license included), at monthly billing, in us-east-1.

  • Previous minimum: Graphics.g4dn at $537/month (4 vCPU, 16 GB RAM, full GPU).
  • New minimum: Graphics.g6f.large at $193/month (2 vCPU, 8 GB RAM, 1/8 GPU).

This lower-cost entry point makes GPU acceleration accessible to a much broader range of use cases and budgets. For customers who need light GPU capabilities – such as basic CAD work, video conferencing with virtual backgrounds, or entry-level content creation – fractional GPU bundles provide the right amount of GPU resources without paying for capacity they won’t use.

The Graphics.g6f.xlarge bundle offers an even more compelling comparison: at $277/month with 4 vCPU and 16 GB RAM, it costs ~48% less than the previous entry-level Graphics.g4dn while providing the same CPU and memory resources with 1/8 GPU allocation.

Performance and Value

The new G6-based bundles deliver significant improvements across all tiers:

Compared to Graphics.g4dn (4 vCPU, 16 GB RAM, 1 GPU, 16 GB VRAM):

  • Graphics.g6.xlarge provides up to 2x better graphics performance
  • 50% more GPU memory (24 GB vs. 16 GB)
  • Up to 65% better price-performance ratio
  • Monthly: $650 vs. $537 (21% higher price for up to 2x GPU performance)

Compared to GraphicsPro.g4dn (16 vCPU, 64 GB RAM, 1 GPU, 16 GB VRAM):

  • Graphics.g6.4xlarge provides up to 2x better graphics performance
  • 50% more GPU memory (24 GB vs. 16 GB)
  • Up to 42% better price-performance ratio
  • Monthly: $1,353 vs. $959 (41% higher price for up to 2x GPU performance)

These improvements mean you work with higher-resolution assets and run more demanding applications while achieving better overall value despite the higher absolute prices.

High-Memory Bundles

For customers with memory-intensive workloads – particularly in architecture, engineering, and construction – the new high-memory bundles solve a long-standing challenge. Many CAD and BIM applications require substantial system RAM to handle large assemblies and complex models. Customers have asked us for WorkSpaces GPU hardware with a greater amount of system RAM available. A prior generation, the GraphicsPro instances (based on G3 EC2 instances) delivered this, but the g4dn bundles had lower amounts of RAM available.

Customers have asked us for a modern replacement to the original GraphicsPro (G3) bundles. We are happy to announce Gr6 WorkSpaces GPU bundles to meet this need. The Gr6 GPU bundles deliver increased RAM for higher-end configurations which require more memory.

Compared to GraphicsPro (g3.4xlarge) (16 vCPU, 122 GB RAM, 8 GB VRAM):

  • Graphics.gr6.4xlarge provides a modern GPU chipset.
    • g6 instances use the NVIDIA L4 GPU; g3 instances used the NVIDIA Tesla M60 GPU.
  • 5% more system RAM (128 GB vs. 122 GB)
  • 3x more GPU memory (24 GB vs. 8 GB)

For customers who need the high memory capacity, the Gr6 bundles provide a compelling upgrade path with significantly better performance and more GPU memory.

The Graphics.gr6f.4xlarge bundle (16 vCPU, 128 GB RAM, 1/2 GPU) offers another option for workloads that need substantial system memory but moderate GPU capabilities.

Key Features and Benefits

High-Performance Graphics Acceleration

The NVIDIA L4 GPUs in G6-based bundles feature third-generation RT Cores that accelerate ray tracing workloads, making them ideal for rendering photorealistic scenes in real-time. Whether you’re working with CAD applications, 3D animation software, or video editing tools, these bundles provide the performance needed for professional creative workflows. The performance improvement over g4dn instances means faster render times, smoother viewport performance, and the ability to work with more complex scenes.

Enhanced AI and ML Capabilities

Fourth-generation Tensor Cores deliver up to 2x better performance for AI-powered applications compared to g4dn instances on representative ML inference workloads. This makes G6-based bundles well-suited for creative professionals using AI-enhanced tools for image upscaling, video enhancement, generative design workflows, and real-time AI-powered effects in creative applications.

Flexible Memory Options

The new high-memory bundles offer up to 256 GB of system memory with a 1:8 vCPU to RAM ratio. This makes them ideal for memory-intensive applications like large-scale 3D modeling, complex CAD assemblies, simulation workloads, and data visualization that require substantial RAM alongside GPU acceleration.

Built on AWS Nitro System

All G6-based WorkSpaces bundles are built on the AWS Nitro System, which provisions GPUs in pass-through mode to deliver performance comparable to bare-metal instances. This architecture ensures you get maximum performance from the underlying hardware while maintaining the security and isolation benefits of virtualized environments.

Use Cases

The new G6-based WorkSpaces bundles are designed for a wide range of graphics-intensive use cases:

Architecture, Engineering, and Construction (AEC)

Run demanding CAD and BIM applications like Autodesk Revit, AutoCAD, and Bentley MicroStation with smooth performance for complex 3D models and assemblies. The high-memory gr6 bundles are particularly well-suited for large architectural projects and infrastructure designs that require substantial RAM to handle detailed models with thousands of components.

Media and Entertainment

Edit 4K and 8K video content, create 3D animations, and render visual effects with applications like Adobe Premiere Pro, DaVinci Resolve, Autodesk Maya, and Blender. The full GPU bundles provide the performance needed for real-time playback and effects processing, while fractional GPU bundles work well for basic editing and review workflows.

AI-Powered Creative Workflows

Accelerate AI-enhanced creative applications for image generation, video upscaling, and intelligent content creation using the enhanced Tensor Core performance. The improvement in ML inference performance enables real-time AI effects and faster processing of AI-powered tools.

Getting Started

To get started with the new G6 WorkSpaces, please see our Create a WorkSpace documentation.

Learn More

To learn more about the new G6-based WorkSpaces bundles and how they can accelerate your graphics-intensive workloads, visit:

We’re excited to see what you build with these new high-performance WorkSpaces bundles. As always, we welcome your feedback.

Dan is a Senior AWS Customer Success Specialist, focusing on helping customers scale and optimize end-user computing solutions. Dan also focuses on EC2, Microsoft, and Linux based workloads. Dan has been at AWS since March 2016, working with End User Compute solutions throughout. Previously, Dan has been an AWS Premium Support escalation engineer, Specialist Technical Account Manager, and Specialist Solutions Architect.
Ariel Fu is a senior product manager at AWS End User Computing team. Ariel has 15 years of experience in the IT industry, including 13 years as product manager in both startups and large corporations. Over the years, she strives to lead cross-functional and cross-border teams to build mobile/desktop apps and SaaS solutions that customers love.