Mar 13, 2020 07:00:00

Spec comparison table of AWS instance 'M6g' adopting Amazon's original ARM processor 'Graviton 2' will be released

The next-generation ARM processor Graviton 2 for cloud computing, announced by Amazon in December 2019, is used in

AWS M6g instances. AnandTech , a technical media, has released the result of comparing the specifications of the M6g instance with other AWS instances that use Intel and AMD processors.

Amazon's Arm-based Graviton2 Against AMD and Intel: Comparing Cloud Compute
https://www.anandtech.com/show/15578/cloud-clash-amazon-graviton2-arm-against-intel-and-amd

AWS Graviton (Providing Best Price and Performance for EC2 | AWS
https://aws.amazon.com/jp/ec2/graviton/

The 'Graviton 2' developed by Amazon is an ARM architecture processor that uses the Neoverse N1 core and is manufactured on TSMC's 7nm line. You can find out more about Graviton 2 by reading the article below.

Amazon Web Services Announces Next-Generation ARM Processor 'Graviton 2' Originally Designed with 7nm Process-GIGAZINE

The instances compared this time are M6g that adopts Graviton 2, M5a that adopts 'AMD EPYC 7571', and M5n instance that adopts 'Intel Xeon Platinum 8259 CL'. At the time of writing the article, M6g is only available in preview, and M5a and M5n are the highest performance instances of AWS instances that adopted x86.

Amazon EC2 M6g Instance (Best Cost Performance for Common Workloads) | AWS

https://aws.amazon.com/jp/ec2/instance-types/m6/

Amazon EC2 M5 Instance | AWS
https://aws.amazon.com/jp/ec2/instance-types/m5/

Below is a table comparing the specifications of M6g, M5a, and M5n.

Unlike other instances, M6g does not employ

SMT , which has been debated for rationality. This allows the number of vCPUs to be treated as is the number of physical cores. You can understand the discussion about the rationality of SMT by reading the following article.

It is pointed out that the technology 'SMT' that executes multiple threads with one CPU is becoming unreasonable --GIGAZINE

In M5n, when using 64 vCPUs, the virtual machine straddles multiple CPU sockets, but in M6g, the virtual machine supports a single CPU socket. Also, since M6g's ' NUMA node ' is one node, there is no need to worry about performance degradation when crossing NUMA nodes.

As for the CPU cache, M6g is the highest in the L1 and L2 caches, and the number of physical memory clocks is also the highest in M6g. Due to the low power consumption ARM processor, M6g has the lowest TDP.

The usage fee for M6g is as low as $ 2.464 (about 246 yen) per hour, compared to $ 2.752 (about 275 yen) per hour for M5a and $ 3.808 (about 380 yen) per hour for M5n. You can see that it is available.

AWS M6g instances have been announced to be generally available in the near future.