Understanding Hyper-Threading in Computational Resources

Hyper-threading is a CPU technology that allows each physical core to handle two threads (tasks) simultaneously. This increases the total number of tasks your CPU can process at once and improves performance for certain workloads.

What is a vCPU?

A vCPU (virtual CPU) is a thread exposed by a physical CPU core.
With hyper-threading enabled:

• 1 physical core → 2 threads → 2 vCPUs

Example:

• CPU with 2 physical cores → 4 vCPUs
• CPU with 8 physical cores → 16 vCPUs

How Hyper-Threading Works in Cloud Machines

The cloud machines we make available at Inductiva often expose all available vCPUs by default.
Example machine types:

Formula:

vCPUs = Physical Cores × Threads per Core

Controlling Hyper-Threading

You can disable hyper-threading with Inductiva at the moment you create a Machine Group. This is done by setting the parameter threads_per_core=1 in the constructor:

cloud_machine = inductiva.resources.MachineGroup(
    provider="GCP",
    machine_type="c2d-highcpu-4",
    threads_per_core=1)

This forces the machine to expose only physical cores:

Note:

• In the cloud, you are charged based on the total number of vCPUs defined by the machine type, not on how many vCPUs you actually use. This means the cost is the same whether you set threads_per_core=1 or threads_per_core=2.
• You can not change the number of threads per core at run time. This needs to be done when initializing the machine.

Summary

• Hyper-threading increases parallel processing by exposing multiple threads per core.
• vCPUs = physical cores × threads per core.
• Cloud environments usually expose all vCPUs, but this can be controlled.
• The cost is based on the machine type, not on the actual vCPU usage, meaning, using threads_per_core=1 does not reduce costs.