> oversized
Having "too much performance" is usually a good problem to have.
Is the existing system running on a "named" operating system, like Linux (or uCLinux) or one of the usual embedded OS suppliers? Is it running an in-house OS with complexities like threading, or is it running "a dumb event loop"? Answering that is a big part of the selection criteria.
The suggested parts (i.MXRT1170) run at 800MHz or more. Very powerful.
These are very capable, but extremely complicated chips. We've used an i.MXRT1060.The Reference Manual is over 3600 pages long, and that doesn't include the CPU (that's in the ARM manuals). That is nowhere near long enough as it is mainly "lists of registers" without saying how they work. The IOMUX chapter alone (to set up the I/O pins) goes from Page 317 to 1008. Just the list of IOMUX registers alone goes for 27 pages. You have to set them up before the chip starts working. There are 28 pages just listing all the clocks in this chip (and they're all configurable). Every peripheral is like that.
It took us about 2 years to get our OS working on it properly. We had to rewrite all the drivers as the peripherals in these chips are unlike the previous ones, and a lot harder to use. The drivers had bugs, the chip had lots of complicated problems we had to work around, and the sample code had problems we had to fix, or in most cases, fix by complete rewrites.
This cost us a lot of time, and still hasn't been addressed:
https://community.nxp.com/t5/i-MX-RT-Crossover-MCUs/i-MXRT1060-SEMC-SDRAM-Data-Corruption/m-p/182549...
Just getting working SPI code took me months. The LPSPI module is very complicated, and the NXP code didn't handle most of the options. The supplied initialization code performs an "M*N search" through all possible baud rates using the dividers to find the "best one". That took milliseconds! I made it 150 times faster.
I'd suggest you look for the simplest chip you can find that will do the job, or get one with a proven and working operating system you can use without having to rewrite it.
Tom