AWS announces general availability of Amazon EC2 G7e instances powered by NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs for generative AI inference and graphics workloads.
- •Delivers up to 2.3x inference performance vs G6e, with 2x GPU memory (96 GB/GPU) and 1.85x memory bandwidth
- •Enables running models up to 70B parameters with FP8 precision on a single GPU
- •NVIDIA GPUDirect P2P provides 4x inter-GPU bandwidth over PCIe vs G6e, supporting up to 768 GB GPU memory per node
- •4x networking bandwidth vs G6e; multi-GPU instances support GPUDirect RDMA with EFA and GPUDirectStorage with FSx for Lustre
- •Available in 6 sizes (g7e.2xlarge–g7e.48xlarge) in US East (N. Virginia) and US East (Ohio)
This summary was automatically generated by AI based on the original article and may not be fully accurate.