STM32F401VCT6 works fine on 42MHz clock, but crashes at 84 MHz

I'm programming a custom board built around the STM32F401 MCU and I'm having a hard time configuring the system clock.

I'm using a 8MHz external crystal and I'm trying to set up the PLL to get the maximum frequency given by the datasheet, i.e. 84MHz. I've double checked the values with the cubeMX Clock Configuration tab (i'm not using it to generate code though) and the reference manual, the resulting setup being like this:

pll_m = 8,
 
pll_n = 336,
 
pll_p = 4,
 
pll_q = 7.

the formula for SYSCLK frequency value is

SYSCLK_freq = HSE_freq/pll_m*pll_n/pll_p

which gives us

8MHz/8*336/4 = 84MHz

which is a maximal frequency the MCU can run on. There are some additional constrains on the clocks' values, but they are all satisfied with this setup.

The problem is, it doesn't work. The software crashes even before the main() function - the clock setup function is called from the initialization code in reset handler, and the mcu crashes shortly after setting the sysclk source to PLL.

Interestingly, everything works fine if I set the pll_m value to 16 instead of 8 - the MCU then runs on 42MHz.

The thing is, it shouldn't - the smallest input clock frequency tolerated by the PLL is 1MHz, but it's easy to see that in this case it's HSE_freq/pll_m = 8MHz/16 = 500KHz.

I've checked the HSE frequency with a scope and it indeed equals 8MHz, also checked the frequency of the working setup - indeed it is 42MHz.

The high clock frequency is not that critical and my application can run just fine on 42MHz, however I'm just curious and a little surprised as to why this happens.

I'm using the StdPeriph library to set up the clock and it takes care of setting the flash wait states to 2WS, which is appropriate value for 84MHz clock running at 3.3V, so it's not the flash issue I think.

Any ideas about what could I be doing wrong would be greatly appreciated