DEV Community

Jacob for AWS Community Builders

Posted on


New EC2 Models re:invent 2022

AWS Released 5 new EC2 models in Re:invent 2022. Hereby an overview.

Model Processor vCPU Memory Network EBS
C6in 3rd Gen Intel Xeon 128 256 200Gbs EFA 80Gbs / 350K iops
R7iz (preview) 4th Gen Intel Xeon (Sapphire) 128 1024 DDR5 50Gbs 40Gbs
C7gn (preview) Graviton3E 200Gbs EFA
Hpc6id (preview) 3rd Gen Intel Xeon 64 cores 1024 200Gbs EFA 25
inf2 (preview) Inferentia2 12 (up to 2.3 Petaflops) 384 accelerator NeuronLink

C6in Network Optimised

First with the new 200Gbps EFA network connection.

Target: Virtual applications (firewalls/routers), Telco 5g user plane functions, data analytics, HPC and CPU AI/ML
Region: US East (Ohio), US East (N. Virginia), US West (Oregon), and Europe (Ireland).

R7iz Memory-optimized (preview)

First 4th generation Intel Xeon Scalable processors (code named Sapphire Rapids up to 3.9Ghz
Highest performance per vCPU among x86, 20% more then z1d
First model to use DDR5 with 2.4x higher bandwidth

Including 2 bare metal sizes

Target: RDS or other workloads that have per-core licensing model or require high compute and memory

C7gn (preview)

Latest Generation Graviton3E.
Up to 200gbs Network and 50% higher packets transfer then C6gn.

Target: virtual application , data analytics


Highest CPU code count with large local NVMe up to 15.2 Tb local NVMe

Target: HPC Clusters
Region: US East (Ohio) and AWS GovCloud (US-West

inf2 Deeplearning (preview)

Up to 12 AWS Inferentia2. 3x higher compute, 4x higher throughput and 10x lower latency then Inf1
Up to 2.3 Petaflops DL , 384 GB accelerator memory with 9.8Tbs bandwidth
NeuronLink , intra instance ultra highspeed interconnect
50% better performance per wat to GPU based EC2. Integrated in PyTorch and Tensorflow

Target: Deeplearning.


Top comments (0)

Want to help the DEV Community feel more like a community? 🥹

Head over to the Welcome Thread and greet some new community members!