The fastest core to idle, ryzen Master or a Windows 10 pot?

Ever since a half-month-old release of a Windows power plan to improve the performance of three-generation Ryzen processors abroad, there has been a constant debate about whether Windows 10 or Ryzen Master is using the wrong core, causing the program to leave the fastest core to one side. Some people think it’s a pot of Ryzen Master, and at the same time some people think That Windows 10 is based on incorrect CPPC information.

The fastest core to idle, ryzen Master or a Windows 10 pot?

Ryzen Master

AMD has been studying this issue for some time, and today issued a statement explaining the problem. In short, neither of the above statements is entirely wrong. In fact, Ryzen Master sorts CPU cores differently from CPPC, which Windows Scheduler is exactly the sort of the latter. Ryzen Master will indicate a particular core that is objectively the fastest, but Windows Scheduler does not use which core is the fastest. Instead, it selects the two cores with the fastest average speed within the same CCX.

This may seem complicated and difficult to understand, but this happens because Windows Scheduler needs two cores to share a single-threaded task, so that it can freely switch tasks between the two cores, reducing the heat of a single core to ensure that both cores can be as frequency-efficient as possible. Therefore, the two cores selected must be in the same CCX, otherwise the high latency of switching tasks between the two cores can lose a lot of performance and offset the benefits of switching cores.

AMD recommends that if the user wants the best performance, it is best to do the first, the global C-states and CPPC in the BIOS are set to “auto-on” or forced on; microcode or higher BIOS version, and use AMD X570 chip driver safters after July 2019.

To address any potential distress, AMD is currently updating Ryzen Master, which will display information that matches CPPC2 data.

Add a Comment

Your email address will not be published. Required fields are marked *