I don't think a SATA controller makes sense for them. Keeping the price low is an important design goal, and adding more chips goes against that. An external hard drive with USB3 is probably the best option for a while.
I doubt SATA is that expensive, and honestly they could alternatively throw in PCIe instead of SATA if that's less expensive, and it would also be way more general. I'm considering getting a Rock64Pro because it's reasonably well supported and has a PCIe port, which means I can buy a SATA add-on board.
I wish Raspberry Pi had better support for fast storage.
Remember that the new Pi 4B has USB3, and according to the benchmark posted on their site it can read from an SSD over USB at 363 MB/s, and write at 323 MB/s. That's pretty good, I don't think you'll get much better than that with a SATA chip connected to the same SoC, so a regular external hard drive with USB3 is probably the best option for most Raspberry Pi customers.
The Rock64Pro is probably a great board too, and it also has USB3, so it can do both options. There's some Odroid boards that have built in SATA as well.
I'd start looking at actual volume prices for controller chips first, before dismisssing this, but that's just me. I have no idea what the cost is, but I could imagine it to be like maybe 5-10 cents by now. Just a guess. It's that kind of old but still very useful tech by now.
Btw: Since we needed to interface to SATA reliably via USB (from a raspi CM.)
We ended up identifying a particular Taiwan-designed JMicron-based SATA/USB2 controller that worked really reliably with Linux, unlike, say every single PRC-designed SATA/USB2 controller chip we tested. It was a pretty painful experience going through all of that testing.
We bought 120 retail packages of that JMicron-based product, directly from the Taiwanese producer, just to be sure, to guarantee our first-gen production run.
I think many RasPi users want a reliably high-speed, low-cost storage device. It doesn't make much sense for all of these users to have go through this very annoying process. It should just work. Since the chip cost for this by now is likely very small in the raspi kind of volume, why not?
Nothing of that complexity is 5-10c in volume (even huge volumes). I would guess somewhere in the 1-2$ range, which _is_ a big issue for them. For reference, 5-10c won't even buy you a middle to low performance op-amp.
The Pi became possible because of integration, where the SoC supported all the required interfaces internally so very few external ICs were required. That's what allows them to keep the price so low. Additional chips add significant additional cost. That's why there have always been compromises like the USB-connected ethernet. They can only do what the SoC supports natively.
Keep in mind that this is a 10+ year old problem by now (talking about USB2-to-SATA). They've gone through all of the cost cutting methods possible by now.
I can buy that chip that also happens to come with a PCB, a very nice extruded and anodized alumunium case, plastic end-pieces + screws, indication LEDs, voltage regulators, 10-20 pieces of small passives (redistors/caps), a screwdriver, 2-3 pieces of instrution manuals/warranty notices, two layers of very nicely designed paper packaging + shrinkwrap... all for like $5 in retail. The cost to the retail store for these items is probably closer to $2.
(Oh, and there's also the assembly cost for that PCB inside the case.)
This is my job. I design consumer electronics that ship in volumes larger than the raspberry pi. You've probably used something I've designed before. The backseat engineering by software devs in these threads is annoying and presumptuous. The pi designers aren't incompetent, they've probably rolled up dozens of feature BOMs to find what they can afford to ship.
Now, I wouldn't be surprised if it was 50c instead of 1$, but _nothing_ is 5-10c, so you're off by an order of magnitude. Even a single transistor in a package is a few cents. Additionally, if they add the IC they still need a connector, and those are some of the more expensive components in a device like the pi, probably 25c-1$ depending on quality.
> I can buy that chip that also happens to come with a PCB...
You're right that the economics don't seem to make sense, but you made several mistaken assumptions:
1) The price of that device is _almost entirely_ driven by the cost of the IC. Those designs have almost no supporting electronics, and the PCB itself will be 10-20c.
2) You're overestimating the margins. It's probably not delivered to the retailer for 2$, closer to 3$. Generic electronics are not high margin.
3) In cheap, commodity devices the manufacturer can use grey-market ICs, but the pi designers can't get away with that. That easily doubles the IC price.
> This is my job. I design consumer electronics that ship in volumes larger than the raspberry pi. You've probably used something I've designed before.
Fair enough. Thanks for your explanation.
> The pi designers aren't incompetent, they've probably rolled up dozens of feature BOMs to find what they can afford to ship.
That's a borderline straw man argument. You're seem to be implying that I either wrote (or think) that they are incompetent, while none of those things are true.
> Now, I wouldn't be surprised if it was 50c instead of 1$, but _nothing_ is 5-10c, so you're off by an order of magnitude.
Okay, let's say it's 50 cents. I would still think it makes sense to include this feature. Also, that's "half a magnitude off" compared to 10 cents, not a magnitude off.
