On STM32F429, why DMA2 MemoryToMemory transfer can go up to 96MHz???

[waclawek.jan]

Can it? What's the proof? What are the circumstances?

Do you find it too little or too much?

Why the three question marks?

JW

[NBogd.1]

Hello.

From my previous question about getting data from AD9226 with an F407, in the end with help from a smart fellow, I did managed to get

data at 54MHz.

The same person told me the fact that the F429 can be overclocked to go up to 96MHz DMA2 transfer rate,

when transfering from GPIOC port to memory, using this configuration:

 /* PLL_VCO = (HSE_VALUE or HSI_VALUE / PLL_M) * PLL_N */

#define PLL_M   4

/* USB OTG FS, SDIO and RNG Clock = PLL_VCO / PLLQ */

#define PLL_Q   12

#define PLL_N   288

/* SYSCLK = PLL_VCO / PLL_P */

I connected the ADC to a Core4X9I board, from Waveshare, that has an F429 uC, and it goes well.

And I want to know how is this possible?

Thank's!

[waclawek.jan]

At room temperature and ideal power conditions and the proper alignment of Mars to Jupiter, any integrated circuit can be run at higher than specified frequency, with outward success. Sometimes, if you are lucky not to exercise some critical path (which you have no idea which one are they) that "higher frequency" may be surprisingly high, even twice as high as the manufacturer's specification.

What the manufacturer gives you are *guaranteed* values. In other words, if you overclock even a "little bit", the chip may fail under any unexpected circumstance, in the most funny ways.

One war story for many, once I've tried to help somebody who claimed his programs run perfectly except when a certain command falls onto a certain address in FLASH... needless to say, it turned out the chip was severely overclocked, running absolutely perfectly at the specified clock.

JW

[NBogd.1]

Thank you for your time to clear the things for me!