You can get a Khadas Vim Pro which has very similar hardware. And like the C2 it's one of the few ARM dev boards that has partial mainline Linux kernel support. It has opensource Mali drivers thanks to BayLibre, but some outstanding issues with USB unfortunately. (the Khadas team is really small so unfortunately progress is slow :( )
Dram and nand manufacturers have been under producing. The logic being that last time there was an undersupply everyone upped production and flooded the market. They see this as a massive negative as it dropped prices and sunk everyone into losses.
Dram and nand fabbing does not use the same process as cpus. Thus it has its own seperate supply and those machines can only be used to make that supply. Thus there is no simple mechanism for the market to self regulate. Unless a maker is willing to idle multi million dollar machines supply is fixed. Which is just another way of saying that a significant portion of the cost in dram and nand manufacturing are fixed costs and thus even if you lose money overall because you make money on the marginal unit you are forced to keep making more units even as you lose money.
Its a tough industry but for now they have a sweet situation going. Provided no new makers enter the market and no existing makers up supply they will keep printing cash.
I work for a tech company in Taiwan who package DRAM and SD cards.
The factory takes silicon wafers, grinds them down to the right thickness, cuts them up (dicing), bonds wires on, puts them into plastic chip packages, tests them, and puts the finished chips into pretty cardboard boxes with plastic windows that you see for sale in stores.
The testing and manufacturing machines are really expensive. The parent company invested too much in x-ray machines to test BGA chip packages, and almost went bankrupt in the dot-com crash. Management wanted to fire the expensive programmers in the IT department/call centre/automation.
My boss (the founder's son) just finished his Masters in Artificial Intelligence in the US. He wasn't going to let the programmers just leave. So he made a startup company in the same office, spun off, refinanced the loans, and now both companies work together closely but the IT department can also do contracts for other clients to help stay afloat.
Automation tasks are fun when I get them - lots of coding drivers for old hardware in the testing area. Most of the time the factory just keeps running though, so I can do data analysis on the log files to help optimise the scheduling.
Most cards are manufactured at high capacity (e.g. 512GB SD card, 400 GB microSD). When they fail a test, they are "reworked" - the die is cut in half with an electron microscope, and tested again. Thus we get many small cards and a few expensive big cards.
The clients are the people with their name on the SD card, and they're the ones designing the silicon. If they order more, the factory has the capacity. But if the purchase orders are down, we can't make more unless they sell.
> When they fail a test, they are "reworked" - the die is cut in half with an electron microscope, and tested again.
Are they actually slicing NAND flash dies, or just unstacking them and disabling defective blocks? (The largest NAND dies so far are 512Gb TLC, which is 64GB, so large SD cards need several dies.)
They're not literally cut in half; the arrays are designed with "jumpers" to disable blocks. But the capacity is halved, so that's why the 64GB dies become 32GB, 16GB, or 8GB cards. (the dies live again!)
On top of that SBC makers are competing with smartphones for parts.
And while possibly a bad example in this regard, Apple has a history of reserving whole yearly factory outputs to ensure they have the parts available.
Not so much that they are competing for parts with smartphones so much as they are straight up picking up the leftovers. Nobody making SBCs is really shipping enough units to really have any leverage with suppliers, not even the Raspberry Pi Foundation. Any shortages are gonna hit them first.
Kinda. At least with the RPi it is successful enough that their SOC supplier has been making custom chips for them to maintain various compatibilities while boosting performance.
Many knowledgeable people in this thread. I have a related question. I want to compile AOSP and run on hackable hardware (i.e. some dev board). The official instructions are for some slightly pricey board called HiKey/LeMaker. I was surprised about lack of RPi support. Any comments?
If you are ok with using older AOSP, Olimex Allwinner A20 based boards are a good bet.
If you want a recent version of Android, I would look at Snapdragon 410c.
Odroid-w will never be re-stocked because Broadcom stopped selling them the SoC (presumably because the Raspberry Pi Foundation didn't like the competition). I also grabbed two of them before Zero came out, for tenfold(!) price.
If you really like that SoC, the Egoman media dongle is using the same one. (disclaimer: they also make the red Raspberry Pi, and I worked for them for 4 months in summer 2013).
I was about to leave here a pointer to elty.pl where I bought one a while back, and they were in stock just a few days ago, but they've become unobtanium there as well.
Now until they become available again the closest thing to the C2 seems the NanoPI K2, which is close enough but not identical, so compatibility has to be tested.
Kind of highjacking the thread, but are there any SBC that support 4K HDR under Linux? I bought a OrangePi PC2 as the hardware does, but it only works at that resolution under Android.
