2

Long story short: I haven't allowed Geekbench 4 to access the camera, which means that the application shouldn't be able to use the camera. However, in the benchmark result a category named 'Camera', with the unit 'images/sec', caught my eye.

Since Geekbench 4 is available on the App Store, I doubt that they're using any method to bypass system permissions. The only realistic possibility I see is that this is somehow simulated, but this begs the question: is there a way to accurately simulate this metric without actually using the camera?

Screenshot for reference:

Screenshot

4

https://www.geekbench.com/doc/geekbench4-cpu-workloads.pdf
Try going to that PDF. It's from Geekbench themselves and they explain the different methods of each workload.
In this case, Geekbench processes photos rather than take them. It applies filters etc. and decrypts, compresses etc.

From that PDF:

Camera replicates a photo sharing application like Instagram. Camera merges several steps into one workload
* AES decryption
* AES key generation using PBKDF HMAC SHA1
* SHA2 checksum generation
* JSON parsing
* JPEG compression and decompression
* PNG decompression
* Image compositing
* Image filters (gaussian blur, contrast)
* SQLite (INSERT the processed image metadata into a database)
All steps run on the CPU and are not accelerated by the GPU.

  • 1
    Good find! Although, ‘Camera’ seems like a misleading name for the benchmark, but maybe that’s just me. – xoudini Dec 22 '17 at 13:30
  • Yes, 'Camera' is misleading. – Neeron Bhatta Dec 23 '17 at 23:33

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .