Primate Labs, the creators of Geekbench, have released the first version of the artificial intelligence benchmark Geekbench AI. The software is based on Geekbench ML (machine learning), which has been available for some time.
AI algorithms work differently on different hardware. Geekbench AI is a cross-platform test, so users can see very different results. For this reason, the leaderboard will be divided into three groups: CPUs, graphics cards, and NPUs (neural processors).
The Geekbench AI test will use proprietary algorithms to test the speed of new processors. This indicator can be compared between devices in each group, although nothing prevents you from comparing a phone with 450 W video card.
Primate Labs explains that AI testing is much more complex than GPU testing. It is not only raw computing power that needs to be measured. Over time, GPU vendors have added new frameworks and API support, which should also be reflected in the benchmark. The Geekbench AI benchmark focuses on various subtests, but it is difficult to compare them with gaming benchmarks that take into account scaling and frame generation. The developer says that it is even more difficult to take into account features for AI tests.
The Geekbench AI results will be divided into three categories: full accuracy, semi-accuracy, and quantized scores. The solution aims to provide a better understanding of how developers or hardware vendors interpret tests by offering more dimensions, similar to how CPU testing centers around single- and multi-threaded metrics.
Users will be able to find out about the fastest AI hardware. The official ranking may change somewhat slowly as new devices are added Geekbench AI is available for free as a trial.
Source: Geekbench
Spelling error report
The following text will be sent to our editors: