The benchmarking platform GeekBench released the 6.1 version update today. After all platforms include Clang 16 factors, it says it can more accurately reflect the multi-core performance of the device.
Compared with version 6.0, GeekBench 6.1 has a 5% increase in single-core scores and a 10% increase in multi-core scores. Therefore, it is officially recommended not to compare the scores of the 6.0 and 6.1 versions of the device.
Geekbench 6.1 updates are as follows:
Include Clang 16 arguments.
Geekbench 6.1 extends the workload gap (the pause between workloads) from two seconds to five seconds, minimizing the difference in results due to thermal throttling on new smartphones like the Galaxy S23 series.
Support for Scalable Vector Extensions (SVE) instructions in several image processing and machine learning functions, as well as support for AVX512-FP16 instructions in various image processing tasks.
For systems without FP16 (floating-point 16-bit) instructions, Geekbench 6.1 uses fixed-point arithmetic (fixed-point) to consider based on multiple image processing functions.
Specifically targeting high-end desktop processors such as 12- and 16-core AMD Ryzens, AMD Threadrippers, and Intel Xeons, Geekbench 6.1 enhances multi-core performance for background blur and horizon detection workloads.