Geekbench 5 is primarily a cross-platform benchmark for main processors, which runs through various tests in order to finally output the processor's performance in the form of a numerical value. With this numerical value you can check if the processor is achieving the performance it is supposed to achieve or you can see how much faster/slower another CPU is approximately.
In addition to the CPU benchmarks performance, Geekbench can also perform various graphics benchmarks in the form of a numerical value. There are a total of 4 different benchmarks that can be run with Geekbench:
CUDA Benchmark
CUDA, or Compute Unified Device Architecture, is an API (Application Programming Interface) developed by NVIDIA to use the graphics processor to calculate processes that the main processor is normally responsible for. The API is mainly used in technical and scientific calculations. Since the API was developed by NVIDIA, this benchmark can only be run with NVIDIA graphics cards.
OpenCL Benchmark
OpenCL was originally developed by Apple in collaboration with AMD, IBM, Intel and NVIDIA. It works similar to CUDA, but is an open system and can therefore be used by almost any graphics processor.
Vulcan Benchmark
Vulkan is a cross-platform open source programming interface for graphics cards. This API focuses on 2D and 3D graphics. In contrast to OpenGL, you can or must specify all the details in the code yourself, while with OpenGL the driver does a lot on its own.
Metal Benchmark
Metal is a programming interface developed by Apple and is comparable to Vulkan in that every detail has to be specified in the code. Since Metal is a proprietary interface, the benchmark is only supported by Apple GPUs and AMD GPUs.