They are claiming 100% compatibility to Cortex-M4. And their lead partner is ST! The first new STM32F7 looks like a faster version of the STM32F429. So in theory, it should not be difficult at all. No idea about practice, though
There are other vendors with good Cortex-M families. But I’ve always been comfortable with our choice of STM32, and today more than just comfortable
The new core is faster thanks to many improvements, looks like almost a factor two compared to the existing M4 cores at the same clock frequency. Double precision FPU is new. Haven’t looked closely at the other details yet.
The first STM32F7 seems to run at 200 MHz, with plans to go to 800 MHz eventually (don’t hold your breath though, this could take a few years…).
If someone is interested in the details (in German, though):
Right now all I see is the STM32F756NG but that seems like a 216 pin package (my custom boards use the 144pin F4). Do you think they’ll release a 144pin version soon or should I start working on a new board?
Also, do you think GHI will be producing a board with these chips?
The G400 is great, but it’s not just a single chip. And the projects I’ve been working on are STM32F4 specific (not that it’d be hard to change but I’m doing some register manipulation). So having something that’s “binary compatible” would be better for me.
I have no inside information here. I guess that eventually they will. Samples of the first new chip have reportedly been shipped already, mass production in March next year. I’d expect smaller packages later next year.
The point is, of course, that GHI doesn’t have to become experts in more than one manufacturer’s chips. Right now, they’re working with three; Atmel’s (Hydra), ST’s (Cerb*), and NXPs (G120, G400). It seems to me they did that because they needed a new SoC to replace USBizi, and Mountaineer helpfully provided the port, but for the bigger SoMs, the STM32 wasn’t ideal, because it didn’t have a DRAM controller or LCD interface. That’s changed, now, with the STM32F427, and with the new M7 parts, they could do everything on one architecture.
That means less churn in the port, and fewer bugs, because they’d only be working on one port, instead of three.
That is true, but by the time M7 is out and fully tested, GHI will have their firmwares polished, besides, they are already offering a wide range of industrial modules covering just about anything, so I don’t think there’s any point to move to the same architecture. That would be a move for the sake of moving. My guess they will skip M7 and rethink their strategy when M9 goes out