EC2 C7a Arrives in N. California With Up to 50% Higher Performance
Hey everyone, Shii-chan here! Today's update is about a compute-optimized instance family landing in a new region.
AWS What's NewWhat was announced?
AWS shared in its What's New feed that compute-optimized Amazon EC2 C7a instances are now available in the US West (N. California) region. C7a is powered by 4th Gen AMD EPYC processors (code-named Genoa) with a maximum frequency of 3.7 GHz, delivering up to 50% higher performance than the previous-generation C6a.
The story so far
Before this, anyone running compute-optimized workloads in N. California had C6a as their option. C6a was a strong performer, but it didn't support newer processor capabilities like AVX-512, VNNI, and bfloat16, and it topped out at 28 attached EBS volumes.
What changes
With C7a now available in this region, these workloads can run with more performance headroom:
- Batch processing
- Distributed analytics
- High performance computing (HPC)
- Ad serving
- Highly scalable multiplayer gaming
- Video encoding
C7a uses DDR5 memory with 2.25x more memory bandwidth than C6a. Combined with support for AVX-512, VNNI, and bfloat16, workloads that can take advantage of these capabilities should see real gains in raw compute performance.
Dive Deep
C7a comes in 12 sizes, from medium up to 48xlarge, including a bare-metal option, and it's built on the AWS Nitro System.
One detail worth calling out: the EBS volume limit. C6a topped out at 28 attached volumes, while C7a supports up to 128 — a meaningful upgrade if your architecture relies on splitting storage across many volumes.
C7a is available through On-Demand, Spot Instances, and Savings Plans.
Wrap-up
- Compute-optimized EC2 C7a instances are now available in US West (N. California)
- Powered by 4th Gen AMD EPYC (Genoa), up to 50% higher performance than C6a
- Supports AVX-512, VNNI, and bfloat16, with 2.25x more memory bandwidth
- Attachable EBS volumes jump from 28 to 128
- 12 sizes from medium to 48xlarge, including bare metal
If your team runs HPC or batch workloads in N. California, switching to C7a could make a noticeable difference.