@Aura_Text@Prakhar6200@Mr__Polygon@tgod34748 There are applications where CUDA adoption is high (e.g., training machine learning models), but those fall outside the scope of what Geekbench is trying to measure.
@Aura_Text@Prakhar6200@Mr__Polygon@tgod34748 CUDA usage in Geekbench 5 was low (less than 3% of GPU Compute results were from CUDA). CUDA adoption in consumer applications is low, too. OpenCL, OpenGL, and Metal are more popular for GPU Compute in consumer applications.
It's disappointing to see OnePlus handsets making performance decisions based on application identifiers rather than application behavior. We view this as a form of benchmark manipulation. We've delisted the OnePlus 9 and OnePlus 9 Pro from our Android Benchmark chart.
Geekbench ML, our new machine learning benchmark, is coming out in May for iOS and Android. Sign up to be notified when it's available: https://t.co/6TXsXVICjq
Geekbench 5.4, with preview support for Linux/ARM and Linux/RISC-V, is now available for download. Find out more on the Geekbench website: https://t.co/2qCsI9VFy3
Looking for a change in the new year? We're hiring for multiple roles (including for our new Machine Learning benchmark). Check out our current postings on the Primate Labs Jobs page: https://t.co/uvFNQHHIFQ
@BeirenXie The Huawei Mate 30 Pro 5G was added to the excluded devices list by accident in late December, and was removed yesterday.
We do not have any reason to believe the Huawei Mate 30 Pro 5G is engaging in any sort of benchmark manipulation.
Geekbench Browser maintenance is taking significantly longer than expected. We hope to have it back online within the next three hours. Thank you for your patience.