Will overclocking the axoloti work?

#1

to put it simple: i found a article on overclocking the stm32f4 mcu series, they can go up to 250mhz max. (source: http://stm32f4-discovery.com/2014/11/overclock-stm32f4-device-up-to-250mhz/ )
(i know overclocking is unsafe)

now i am wondering if this is possible to do with the axoloti core for more processing power.
or will it have alot of problems with the usb, dram and other periphals at 200mhz core speed (or more)?

has someone tried overclocking an stm32f4 mcu before? and willing to share thoughts on this part?

#2

It is possible, I have tested overclocking to verify that there is margin on the design on sample boards.
But how do you plan to test if the processor is still doing its job correctly in all circumstances? Some patches may run, others may crash, as we don't know where the critical paths are that will fail first, it's hard to make a conclusive test.
I would not recommend...

#3

the preenfm2 firmware is available in an overclocked (192mhz) version. it is widely tested and considered stable. xavier (the preenfm dev) just put a warning when you download it. maybe you could compile a second overclocked firmware, for people to use at their own risk as well...just put a warning sign somewhere, unsupported or whatever.

#4

Sorry, I will not do this, the processor workload in Axoloti can be quite a bit more diverse than on the preenfm, making it much much harder to validate stability.

#5

i understand, and i don't :slight_smile: is there a risk of bricking the axoloti? i thought overclocking could just make it freeze. "releasing" a overclocked firmware (with the warning and no warranty of course) would allow for much broader testing, since you don't have to do it on your own...

#6

I think it is unlikely to be destructive, but would only know for sure afterwards...

Practically no amount of functional testing can conclusively exclude all sorts of subtle hardware errors. There are too many parameters involved: ambient temperature, the processor fabrication process variation, and different patches can stress different potential critical timing paths. It's a very long time tunnel I do not want to enter or support entering.