Intel Designs Custom Chips for AWS’ New C4 Instances

Intel has designed custom Xeon processors for Amazon Web Services that will power the cloud provider’s new server instances optimized for high-octane computing. The chips will provide the highest level of CPU performance EC2 has ever seen.

Amazon previewed the new type of instance, which is not yet available, at its re:Invent conference in Las Vegas Thursday. Called C4, it comes in five different configurations, ranging from two to 36 virtual CPU cores and from 3.75 Gigabytes to 60 Gigabytes of RAM.

The instances will use hardware virtualization, which is as close to bare-metal cloud as AWS gets, and run within Virtual Private Cloud environments only, Jeff Barr, chief evangelist at AWS, wrote in a blog post.

The custom AWS CPU, called Intel Xeon E5-2666 v3, is based on the chipmaker’s Haswell architecture and built using its smallest-yet 22 nanometer process technology. The processor runs at base speed of 2.9 GHz, but with “Turbo boost” can go up to 3.5 GHz, according to Amazon.

This is not the first time Intel has customized a processor for a big customer. Making tailored chips for cloud service providers, Internet companies, and hardware vendors has grown into a big business for the company in recent years.

Another recent custom job was for Oracle’s massive database machines that came out in July.

Diane Bryant, general manager of Intel’s data center group, described a new approach the company had taken to tailoring chips for hyper-scale customers using Field-Programmable Gate Arrays.

An FPGA is a reconfigurable semiconductor typically used to give a user the ability to test different configurations before they commit to a volume purchase of non-programmable chips. Intel plans to include an FPGA in a single Xeon package and offload some of the CPU workload to the FPGA.

The chipmaker gives the customer the option of testing different configurations and then order static System-on-Chips that would use the configuration that works best for them. Another option is to deploy Xeon packages with the FPGAs at scale so you can reconfigure them in the future for different workloads.

When Bryant talked about the offering in June, it was not yet available, and she did not say when it would hit the market. It wasn’t clear whether Intel used the approach in designing the latest custom Oracle or AWS CPUs.

In June, Bryant said Intel had designed 15 custom CPUs in 2013 for different customers, including Facebook and eBay. More than double that amount was in the pipeline for 2014, she said.