Hello world!
Today, a somewhat strange idea crossed my mind: why not add emmc memory to the uConsole?
So I bought a few components to try a little experiment:
Yes, I know, it’s a little simplistic, but I can’t wait to compare with a classic microsd!
eMMC performance will be capped at the SDIO maximum transfer rate, but other performance characteristic may be better. I’ve wondered why manufacturers haven’t socketed the eMMC on SoMs to enable replacement. There isn’t much cost differential at this point.
Looks like a great idea. I’d be interested to see your benchmarks.
I have a CM4 w/ eMMC (not CM4 lite) in my uConsole. I got it very cheap (2x for the price of 1). The CM4-eMMC was a bit of a pain to set up, but it is about twice as fast the SDcard-based systems I have. So I am optimistic about how your setup will work out. I think that SD interface forces the chip to use the 4-bit transfer mode of SDIO rather than the EMMC 8-bit mode.
Some simple bootloaders and microcontrollers won’t work with EMMC either, as you can’t use SPI mode (1-bit) that SDcard normally supports.
Some pine64 devices use socketed emmc, as do orangepi’s boards.
Thank you for these interesting remarks! It is true that I did not address the problem of the speed of the transfer. Above all, I am looking for better memory durability, in order to make data storage more reliable. We’ll see what it looks like with the benches! 
Yes, I hope that experimental assembly can work! Theoretically, the adapter work with a Pi, so…
I now use industrial grade microSD in all my projects. SanDisk 16GB is around $60 for a 5-pack. Relatively expensive for such a relatively low capacity, but after having multiple systems fall over dead over the years, I learned my lesson. I used to use an RPi1 on a static IP as my domain’s DNS server. The SD card’s failure mode was that it completely locked up on any reads - such as on boot.
High-quality, Industrial grade μSD cards use NAND die that are “binned” as higher quality when the wafers are cut and packaged. I’ve visited the SanDisk manufacturing facility in Shanghai. The manufacturing process technology is fascinating.
While it isn’t advertise/documented, the cards may also run the NAND in MLC mode (2 bits per cell) versus TLC or QLC (3 bits per cell and 4 bits per cell, respectively). These cards are relatively low capacity because of this lower bit density.
I have just received the same Radxa eMMC to uSD adapter and Orange Pi eMMC card for 256Gb. Just tested it to be fully working with my Raspberry Pi 400. My guess it must work with uConsole too (cannot test because have not received it yet). The speed is better comparing to regular microsd card, and for sure it should be much more reliable in long-term perspective.
Thank you for all these very relevant details. I can’t wait even more to receive the equipment to test! 
Hi all!
The stuff has arrived. I’m currently testing 3 configurations :
Of course, I will give measurement results, but the first tests indicate that the on-board memory is the fastest.
The slowest is the sd and the emmc is somewhere between the two, more on the side of the integrated emmc.
I have an unexpected problem with the microsd flat extension cord. It works perfectly with an sd card, but not with the emmc + adapter. I suppose this is due to a small electrical resistance that the emmc cannot tolerate.
If someone has a suggestion, I’m interested!
The sequel to the next episode 
I have got similar results. Emmc+adapter is a bit faster than sd card, but slower than internal emmc. I think it is due to microsd bus limitations. But anyway it is more reliable than simple microsd card.
I tried to use a UUGear RasPiKey 32GB eMMC in the TF slot and initially had no success.
Where the key’s PCB fanned out beyond the microSD size to accommodate the eMMC chip, it was preventing me from pushing in the key deep enough to engage the slot’s lock. The key would just push itself out.
I then tried the RasPiKey with a microSD card extender PCB, and that worked, but obviously looked a bit precarious, protruding out the top of the console. So what I discovered is that I can engage the lock in the TF socket holder using the extender, take the extender out, and simply put the RasPiKey in, which now butts up snugly with the uConsole, doesn’t fall out, and has been working nicely for the last day. Using a normal microSD card again afterwards allows the lock to be disengaged as normal.
Using the simple agnostic tool for an SDcard speed test, the log reports:
Sequential Write 43487 KB/second
4k Random Write 7571 IOPS
4k Random Read 5515 IOPS
