Geekbench 4 was a big update for the company and it was released back in August of last year. XDA’s own Steve Zimmerman gave us an in-depth look at the changes that had been made for that version, and even took the time to conduct a thorough interview with the CEO of Primate Labs, John Poole that you can check out here, here and here. The company just recently announced the release of the 4.1 update for their benchmarking software, and it comes with a number of changes as well.
Since there have been changes made to how overall scores are calculated when compared to Geekbench 4.0, it’s recommended that you do not compare 4.0 scores to 4.1. While some results could be the same (especially when it comes to overall scores), individual workload scores are typically showing different scores in 4.1 when compared to 4.0. So definitely be aware of that when you start comparing benchmark scores for older phones to your brand new one.
The builds of Geekbench have been updated to newer versions for the different platforms (Clang 3.9 for Linux, Android NDK 13b for Android and Xcode 8.2 for iOS and macOS). The company has enabled AArch32 cryptography instructions in the Android ARMv7 build and they’ve changed the Memory Latency workload to avoid cache hits on Cortex A72 and A73. The update will now report Memory Latency workload performance in nanoseconds, and has also added AVX512 implementations to FFT, GEMM workloads.
The Geekbench 4.1 update disables SQLite cache statistics, as well as LLVM runtime assertations, so that it can improve multi-core scalability. We’re told that you should expect about a 2% increase in single-core scores and at least a 5% increase in multi-core scores because of these changes. However, do note that the increase in multi-core scores will depend entirely on how many cores you have in your device.
Source: Geekbench
from xda-developers http://ift.tt/2oRlunC
via IFTTT
Aucun commentaire:
Enregistrer un commentaire