How we use Geekbench benchmarks

The Normalized Geekbench Score (NGS) labeled as “Performance” on many of our graphs/charts) combines three metrics—single-core CPU, multi-core CPU, and adjusted GPU Metal score—into a single number, providing a balanced indicator of overall system performance.

Because Geekbench 6 Metal GPU scores are in the hundreds of thousands, they can overwhelm CPU scores when combined (added up) with single-core and multi-core results. After analyzing the data with ChatGPT, we found that dividing the Metal score by 7.3 brings it into a comparable range/scale with CPU scores. In the future, in v1.1+, we plan to incorporate Cinebench, Speedometer Disk I/O performance as well as AI-related benchmarks to further enhance our scoring system.

Our goal is to help guide our community towards what is the best performance-for-value ratio for each person.

We’d love to hear from you if you’d like contribute to a future formula for calculating the best possible overall Mac performance metric.

Geekbench 6 is a MacOS app that benchmarks CPU and GPU performance using tasks that mirror real-world usage. Its four main score types — Single-Core, Multi-Core, GPU Compute, and ML Workloads — each reflect different aspects of device performance.

Single-Core CPU Score
What it measures:
Performance of a single CPU core.

Real-world relevance:
– App launch times, UI responsiveness
– Web browsing, email, basic productivity
– Games and tasks that don’t use multiple threads

High single-core scores suggest smoother performance in everyday, lightly threaded workloads.

Multi-Core CPU Score
What it measures:
Combined performance across all CPU cores.

Real-world relevance:
– Running many apps or browser tabs
– Video editing, software compiling, data crunching
– Background syncs or emulation

A strong multi-core score benefits pros and heavy multitaskers.

GPU Compute Score
What it measures:
General-purpose GPU performance using APIs like Metal, OpenCL, or Vulkan.

Real-world relevance:
– Accelerated photo and video editing
– 3D rendering, simulations
– Machine learning inference

Though it’s not a direct gaming benchmark, it reflects the GPU’s compute ability for creative and technical workflows.

In short: Geekbench 6 is designed to reflect real-world performance across modern devices. While synthetic, its tests map closely to tasks users care about, making it a practical benchmark for comparing Macs.

The “Bang for Buck” score is a helpful metric designed to measure the overall value of a Mac computer by balancing performance against its cost. To calculate this score, we first take the Geekbench performance benchmarks for Single-Core CPU, Multi-Core CPU, and GPU Metal tests, normalize the GPU Metal score – dividing it by 7.3, and sum these normalized values to generate a combined Performance (sometimes called the Normalized Combined Geekbench Score).

This Performance score is then divided by the device’s MSRP (Manufacturer’s Suggested Retail Price) or its current market price to determine how much performance you get for each dollar spent.

A higher Bang for Buck score indicates a better performance-to-price ratio, highlighting models that provide superior value. This scoring system helps users easily compare various Mac models, allowing them to identify devices that deliver optimal performance relative to their cost, whether looking at newer releases or considering value in the used market.

The Bang for Buck Grade is a curved A+ to C grading system to give you a general indication of the value of a machine. Less than a handful of machines get an A. That’s not necessarily a grade on the quality of the Mac itself.

For example, if you’re a high end video editor, you would get a lot of value of having a faster $7,000 M3 Ultra shave off hours of rendering times per month but using Bang for Buck grade that machine would get a C or D grade.

Bang for Buck is useful for those who are currently budget conscious. High end power users know that performance can be expensive but they also know their time is with a lot of money.