AWS vs Azure vs Google vs IBM vs Oracle - VMs Q1 2021

Go back to list

AWS vs Azure vs Google vs IBM vs Oracle - VMs Q1 2021
Compute IaaS Virtual Machine Performance Price Hyperscalers

The Cloud market is dominated by USA since its beginning. Despite none of them are a pure-player, they push the edge of technology through the their diversity of services and customers. Virtual Machines being one of the main component, we did a comparison of the Big 5, their VMs offering and more especially the compute optimised series

This study aims to compare a specific type of virtual machines, the balanced ones with a CPU/RAM ratio of 1:2 (ie 2CPU 4GB). Knowing that the provider's catalogs propose instance types with different focuses, we just picked products that match the best with the following specifications:

  • 2 CPUs - 4GB - 100GB of SSD Block Storage
  • 4 CPUs - 8GB - 200GB of SSD Block Storage
  • 8 CPUs - 16GB - 400GB of SSD Block Storage
  • 16 CPUs - 32GB - 800GB of SSD Block Storage

Custom characteristics

AWS' logo

c5a Designed for compute-intensive workloads with cost-effective high performance , powered with AMD EPYC
We discovered EPYC 7571 and 7571 with high variation in frequency, from 1,6 up to 3.3GHz.

Azure's logo
Fs v2 features a sustained all core Turbo clock speed of 3.4 GHz and a maximum single-core turbo frequency of 3.7 GHz. Our platform detected a Intel Xeon Platinum 8168 for all studied instance sizes.

IBM's logo

CX2 Best for workloads with intensive CPU demands. Different chip are used:

  • Unknown Intel Broadwell
  • Unknown Intel Skylake
  • Unknown Intel Cascadelake
Google logo

N2 has balanced price/performance across a wide range of VM shapes with Intel Cascade Lakes at 2.8GHz.
Google Compute being flexible in VM shaping, we are able to create the exact size we require outside of the predefined instance types.

Oracle's logo
E3.Flex lets customers specify the precise number of cores and corresponding memory to match workload requirements
Powered by AMD EPYC 7744, Oracle defines their compute unit as oCPU, where 1oCPU equals to 2 hyperthreads under OS perspective.