Possibility of new STM32H7 with larger RAM (for display double buffering)

[decalvus]

Anyone have some insider knowledge or just plain rumours of what chances are for ST considering releasing a STM32H7 with larger RAM in the near future?
Like 3MB or larger, as some STM32U5’s have (I also just noticed some competitors having large RAM devices, like i.MX RT533 / RT555 with 3MB and 5MB respectively)

The reason I ask is because this would make it possible placing two full-size frame buffers in internal RAM for displays with larger resolutions, doing away with the need for an external SDRAM.
The SDRAM makes PCB layout quite more complicated and needs more layers. It also requires a larger pin count device for the SDRAM connections, often making use of a BGA package necessary.

If only internal RAM is used a smaller pin count LQFP package can be used, making PCB layout much easier. The internal RAM is also much speedier than SDRAM, which is helpful for graphic applications with large and high FPS animations.

The STM32U5’s with 3MB is great for many graphic applications, but is possibly a bit underpowered at 160MHz for some applications with larger screen resolution or demanding animations.

Using two full-size frame buffers (double buffering) helps avoid tearing effects, but requires a lot of RAM. A 5” display with 800x480 resolution, and 24bit (RGB888) color depth, requires 2x800x480x24bit = 2.2MB for two frame buffers. A 7” 1024x600 display requires 3.5MB.

And then there must be memory left for the remaining tasks (perhaps 500kB or even 1MB, but I’m new to all things display related so am just guessing). If the goal would be to support resolutions up to say 1024x600 (which I guess covers a lot of display territory), then about 4.5MB is needed.

Well, just wanted to show my wish list for next years devices.

-Pete

[TDK]

Consider using QSPI RAM. Only 6 pins needed. You can get something around 50 MB/s which is plenty for most things. 8 MB capacity in a SOP8.

APS6404L-SQN-SN AP Memory | Mouser

Probably not going to be new H7s. They've already released quite a few variants and are on to newer families it seems.

[decalvus]

Hello TDK

Thanks for the PSRAM suggestion, I will look into this.

Best regards,
Pete

[Tesla DeLorean]

One of the mass market issues here is die size, contributing to wafer yield and cost. A part retailing for $20 is going to have few takers wanting a $2 part.

You're going to make the die rather large, and that makes the minimum package size follow.

Some of the faux STM32 have two die and bond a memory die to the processor. On the original RPi, Broadcom mounted a Samsung SDRAM DDRx die a-top the processor in a double stack configuration.

You can hide quite a large board behind a 7" display panel

Looked at SoM?