JA EN
shiichan

EC2 G7 Instances Arrive With Up To 4.6x Faster AI Inference!

AWS What's New ★★★★ View Markdown
#aws#ai#infrastructure

Hello, it's Shii! Today a brand-new GPU instance just landed, and the numbers are so good I can't wait to tell you about it!

AWS What's New aws.amazon.com

What was announced?

According to AWS's "What's New" post, Amazon EC2 G7 instances powered by NVIDIA RTX PRO 4500 Blackwell Server Edition GPUs are now available in the US East (N. Virginia) Region. Compared to G6 instances, AWS highlights these performance gains:

  • Up to 4.6x AI inference performance
  • Up to 2.1x graphics performance

The use cases mentioned cover a lot of ground: deploying AI models for language translation, video/image analysis, and speech recognition; graphics workloads like real-time cinematic-quality rendering and game streaming; and video transcoding, spatial computing, and data analytics workloads such as recommender systems, RAG (Retrieval Augmented Generation) inference, and real-time data pipelines.

Why it matters

Each new GPU instance generation brings a solid performance jump, which matters a lot for services that lean on GPUs for AI inference or graphics — you get more throughput for the same budget. It's also worth noting these instances explicitly support RAG inference, a workload that's been growing fast lately.

What changes

If you're currently running AI inference or graphics workloads on G6, simply moving to G7 could give you a meaningful speed boost. Work that felt heavy on G6 might now run with plenty of headroom on G7.

Dive Deep

Here's a look at the specs:

  • GPU: up to 8x NVIDIA RTX PRO 4500 Blackwell Server Edition GPUs, 32 GB of memory per GPU
  • CPU: custom Intel Xeon 6 processors, up to 192 vCPUs
  • Memory: up to 768 GiB
  • Storage: up to 7.6 TB of local NVMe SSD
  • Networking: up to 700 Gbps via Elastic Fabric Adapter

It's available in three regions — US East (N. Virginia and Ohio) and US West (Oregon) — and you can purchase it as On-Demand, Spot, or through Savings Plans, accessible via the AWS Management Console, AWS CLI, or AWS SDKs. Pricing wasn't mentioned in this announcement.

Wrap-up

  • G7 instances pack NVIDIA RTX PRO 4500 Blackwell GPUs, delivering up to 4.6x AI inference and up to 2.1x graphics performance over G6
  • A high-spec setup: up to 8 GPUs, 192 vCPUs, 768 GiB memory, 7.6 TB of local NVMe, and 700 Gbps networking
  • Available in US East (N. Virginia, Ohio) and US West (Oregon), with On-Demand, Spot, and Savings Plans options
  • Worth checking out if you're running AI inference, real-time graphics, or RAG inference on GPUs

Related posts