With the current memory shortages making the MicroSD cards 2-3x the usual bulk cost I took a deeper dive into our situation.
If you are unaware, most class10 a1 cards don’t work well in our boards and screens, they never have. Sometimes they do not show up at all.
It never really made sense to me that a card that CAN go faster did not work in a device that goes slower. Felt like saying you can’t drive a Ferrari in a residential area. Turns out I think there is a simple explanation. Class 10 cards are SDHC fat32, Class 10 A1 are SDXC exFat. If it is that simple I am going to be irritated I did not take the time to look into it more.
If anyone has any info to offer I am all ears. I am buying a few cards to test this out. Also will formatting them to fat32 fix the issue (with the 32gb limit, but who really needs that many gcodes).
Wow, I haven’t looked at SD card prices in awhile and they are bonkers.
My understanding is that it’s not just the filesystem, it’s how they connect. Faster cards want to connect via SDIO where we are connecting via SPI which is a slower fallback.
It seems that ESP-IDF 5 supports exFAT but I don’t know if FluidNC is compiled against that version. Even then, I think it would require code changes.
So, all that to say I don’t think it’s that simple and supporting a broader range of SD cards probably requires both hardware and software changes.
It does seem that some of this could be a timing issue so perhaps some minor FluidNC changes could help.
This is a mix of what I think I know and a conversation with AI.
The real issue that needs to be addressed is someone needs to sort out an implementation for FluidNC that can use either SPI or SDIO, worst case via config.yaml directives. Best case automatically try both and use SDIO if it is detected or SPI if not.
Also, EXFAT support for embedded stuff is long overdue. Not because we need 1Tb .gcode files (!!) but because that is what is on the commodity cheap hardware we can get.
I wonder if you can just add SPI flash memory on the Jackpot itself and not have to worry about SD cards at all. I never remove the SD card but I know some do. With the WebDAV support in 4.0, maybe that is less necessary?
I just bought an 8G card for the laser engraver to make sure it was the older/slower card type. Everything else I had in my house was newer/faster and didn’t work.
But regarding the question that you are really asking: the AI bubble is going to implode some day and firms like Crucial who took a step back from the private sector because of AI will have a rude awakening… My guess.
I have been playing with ESP32-S3 dev boards and they utilise PSRAM, which is effectively on board accessible flash memory.
As an aside I ended using flash memory in the S3, for the opposite reasons on this thread, I was struggling to get some of the prototyping sd card boards to work at all, and really struggled with the older cards to get them to work.
