Processor speeds?

jim lee

jim lee

Well-known member
Back in the day of Teensy 3.2 compiling. If I accidentally got the processor speed set to 96 mhz, it would cause all sorts of runtime errors and oddness. Setting it back down to 72 MHz would solve all of this, and everything would go back to rock solid.

Now with this 4.0 machine, should 600 MHz be the "solid speed"? That seems to be the highest non-overclocked speed choice.
 
Found this in NXP PDF:
Code: 
The i.MX RT chip is a Cortex-M7 based chip that operates at speed up to 600
MHz to provide high CPU performance and best real-time response.
From first PJRC Beta of the 1062 the default speed has been 600 MHz IIRC. And that is the default speed for all releases and 'common' use unless otherwise specified for some special case. So, it is the upper bound of specified usage, but has been the norm for the 4.0 and 4.1.

Less fully tested would be the alternate Optimizations options away from the default. That could vary with the toolchain in use at the time as optimizations evolve and refine and as they apply to the code at hand. That may be the case with the T_3.2 observed 'errors and oddness' as it the T_3.2 defaults to 'Overclock 96 MHz' where most use would have occurred.
 
600 MHz is the officially rated speed by NXP.

600mhz.png
 
One note that probably isn't applicable for most teensy applications, the expected life of the part is significantly lower at 600MHz.
It's only a couple of years of run time at full speed in a warm environment.

If you want something that will run 24/7 for multiple years then you probably want to drop the speed down one step (528MHz from memory). That increases the expected run time to over 10 years.

Probably not something most people will care about. But if you want to build some sort of home automation system that will be sitting there running in the background all the time then it may start to matter.
 
It's only a couple of years of run time at full speed in a warm environment
Click to expand...

I think that is overstating the issue a bit.

If you run the die temp all the way up to 95C which is blistering hot and the point at which the CPU will go into thermal shutdown, then the expected life is derated to 3.2 years @600MHz which is obviously very short. 528MHz is 8.7 years at the same 95C temp, which is obviously much better. Thing is I don't think anyone is actually running their Teensy at 95C on a continual basis, at least not on purpose.

The life derating charts only go as low as 76C in which case it shows a derated lifespan estimate of 10.8 years @600Mhz and that is still running hotter than most applications will see and a fairly long life in the microcontroller world.

If you just want to maximize your life time in a 24/7 application or your application involves sticking it in a small sealed box out in the Phoenix sun, then running it at 528MHz or lower makes a lot of sense if you don't need maximum performance.
 
You must log in or register to reply here.
Back
Top