I use SW4STM32.
I noticed that when I generate a project with SW4STM32, the speed of my STM32F401VBT6 is twice faster.
For example, HAL_Delay(1000) actually awaits 500ms.
USART1 transmits data on the speed 230400, although 115200 is specified in cofiguration.
But if I generate a project using CubeMX it's all ok.
I transfered all the source code from the project generated by CubeMX into the project generatd by SW4STM32, but the problem remained.