gpt4all
gpt4all copied to clipboard
Incorrect tool tip instructions
Documentation
The tool tip for Application General Settings => CPU Threads states, "Amount of processing threads to use bounded by 1 and number of logical processors". I believe this should read physical rather than logical processors.
I conducted a test on a CPU with 32 logical processors to demonstrate that when using only CPUs (no GPU) with localdocs turned on, the amount of time between sending the prompt and the start of the response decreased until 14 CPUs were reached, then rapidly increased once the 17th CPU and beyond was added. (Test was stopped at 22 logical processors.) See chart below.
📉