I do not think ST will ever release it publicly. My assumption is simply that negotiations around publication rights or licensing permission did not conclude successfully.
Nothing to worry about — it appears that ST’s partnership with Think Silicon, the Greek company behind the NEMA GPU, has run its course. If this is the case, I do not blame ST. It is fair to say that negotiations with companies operating within the Greek business and legal framework can be more complex and time-consuming than expected, even for arrangements that might otherwise seem straightforward.
Notably, ST’s latest announced “game changer”, the STM32V8, seems not to include a modern GPU at all. The prior flagship STM32N6 only excluded vector graphics support. Of course, there is no free lunch — a vector graphics requires an additional dedicated SDRAM framebuffer. Realistically 4MB would be a more practical minimum for fully utilizing a vector GPU at resolutions like 1024×768. With 3MB, the headroom is tight if the goal is to run more complex graphics or unlock the vector graphics GPU’s full potential.
In my view, a much bigger concern is that STM32N6 have launched without MIPI-DSI support. This trend is disappointing, as reliable driving of a 1024×768 LCD over LTDC (or even 800x600) is not feasible — bandwidth, pin count, and signal integrity all become limiting factors. The interface also demands too many pins, making high-resolution display designs unnecessarily difficult.
If this trend continues and the STM32V8 also launches without MIPI-DSI (which I do not know), the STM32U5G9 may become the last STM32 microcontroller that even partially meets my design requirements (with the note that its FDCAN implementation has been functionally constrained). It currently appears to be the final device in the portfolio to offer native MIPI-DSI support. Without this interface, the feasibility of reliably driving 1024×768 displays in future STM32-based embedded designs becomes highly uncertain.
