How to benchmark CPU performance on Ubuntu?

Benchmarking CPU performance on Ubuntu involves isolating your system environment and running standardized utility tests like sysbench or Geekbench to measure processing speeds under load. This article walks you through preparing your system, executing tests for both single-core and multi-core evaluations, and interpreting the metrics to gauge your hardware’s capabilities.

System Preparation

To get accurate and reproducible benchmark results, you must minimize variables caused by background activities or software configurations.

• Close background applications: Terminate active browser tabs, development servers, and non-essential GUI software.
• Configure power settings: Ensure Ubuntu is set to prioritize performance over power saving. You can install cpufrequtils to set the governor manually:

sudo apt install cpufrequtils -y
sudo cpufreq-set -g performance

• Monitor system temperatures: Thermal throttling skewing your benchmarks is a common issue. Install lm-sensors to track heat metrics in real time during execution:

sudo apt install lm-sensors -y
watch -n 1 sensors

Option 1: Quick Command-Line Benchmarking with Sysbench

sysbench is a lightweight, widely utilized command-line tool that evaluates CPU capabilities by calculating prime numbers.

Installation

Update your package repository and install the utility directly through standard APT:

sudo apt update
sudo apt install sysbench -y

Running a Single-Core Test

To evaluate single-threaded throughput—crucial for apps that do not scale natively across multiple cores—run the execution flag while forcing a single worker thread:

sysbench cpu --cpu-max-prime=20000 --threads=1 run

Running a Multi-Core Test

To test aggregate computational throughput across your entire system architecture, substitute the thread count flag with your total logical CPU cores using the nproc variable:

sysbench cpu --cpu-max-prime=20000 --threads=$(nproc) run

Analyzing the Output

Focus primarily on the events per second metric under the CPU speed summary. A higher value denotes superior computational speed. Multi-core events per second should scale almost linearly with your system’s core count compared to the single-thread metric.

Option 2: Full-Scale Industry Testing with Geekbench

For deep-dive architectural comparisons against public databases, Geekbench offers a cross-platform testing suite suited for modern multi-threaded workloads.

Download and Extraction

Since it is not hosted in standard Ubuntu repositories, pull the official Linux binaries directly:

GB_VERSION="6.3.0"
wget "https://cdn.geekbench.com/Geekbench-${GB_VERSION}-Linux.tar.gz"
tar -xzf "Geekbench-${GB_VERSION}-Linux.tar.gz"
cd "Geekbench-${GB_VERSION}-Linux"

Executing the Test Suite

Launch the binary. If you are operating on a headless server or wish to suppress automatic browser prompts, use the --no-upload parameter alongside local save arguments:

./geekbench6 --no-upload --save results.txt

Alternatively, running it without options uploads your raw system performance score directly to the Geekbench online browser via a temporary URL, enabling swift performance comparisons against identical CPU models worldwide.