Say you have a cloud server with 1 vCPU. Say it’s able to serve 100 requests per second. Now you upgrade to an instance with 2 vCPUs. How many will it serve — 200?
No, the answer is closer to 100 than to 200. This is because vCPUs are typically just hyperthreads running on the same core 1. When you “upgraded” to an instance with two vCPUs, you now have two threads running on the same real CPU, competing for the same resources. It’s closer to cutting the pie in two than to baking a second pie. The v stands for virtual. That should give you a hint not to take it seriously.
vCPUs are not stable performance targets. Ignore the marketing, and evaluate servers by the number of real CPU cores.
Except for Graviton2 processors on AWS.