Raspberry Pi Sees Their First Price Increase Due To Supply Chain issues

[ed31337]

You'd do even better with a SoC made on a more modern process. Pi v4 uses 28 nm, while ODROID N2/N2+ is made 12 nm. It's both faster and runs much cooler. It's so much more efficient that the N2+ is basically a factory-overclocked N2 and still runs cooler than a Pi v4. Beyond that, it has a better GPU, as well.

Very nice. If they ever release one with 8GB or (ideally) 16GB of RAM and dual head HDMI, I think I would have to consider buying that as an upgrade from my current RPi4.

Even better if they ditch the audio DAC and RTC to lower the price. My HDMI TV used as a monitor does a very fine job of acting as my audio DAC (via HDMI) and I have tons of Android phones laying around that I can use as an RTC via USB/adb.

[smitty3268]

A decade ago? No, they wouldn't have been using 28 nm that soon.

You're right, I think they're on even older tech than that. I didn't intend for 28nm to be taken at face value, I was just throwing out example numbers. Same when i said a decade. I think 65nm (2005-ish nodes) is what I've heard is in use, but I'm sure there's a mix of different nodes involved for different chips and different manufacturers.

[coder]

I've managed to put together my own power supply and I'm pretty happy with it: a $5 DC-DC step down converter from eBay/China, a couple big-ass AGM batteries, and a couple more $5 DC-DC CC/CV step down power supplies to keep the AGM's float charged via old laptop wall bricks. The power could go down for DAYS and I'd have a functioning computer, although I still need to hack my monitor to run off the AGM's.

You'd do even better with a SoC made on a more modern process. Pi v4 uses 28 nm, while ODROID N2/N2+ is made 12 nm. It's both faster and runs much cooler. It's so much more efficient that the N2+ is basically a factory-overclocked N2 and still runs cooler than a Pi v4. Beyond that, it has a better GPU, as well.

[ed31337]

RaspberryPI is a really bad HW platform regarding stability.

USB power supplies limited to 2A and too thin USB cables can cause voltage drops that lead into instabilities, lock-ups and even damaged SD cards.

Hmm, I beg to differ. RPi4 has been rock solid regarding stability for me. Yes, it was crap when I tried attaching a bus powered USB SSD and a bus powered USB HDD at the same time. I just did a little modification of my USB to SATA adapters and now I backfeed 5V directly into the USB HDD. No more stability issues, whatsoever. This is the quietest, most stable computer I've ever had.

Then, once you fix voltage (with a thick cable and a soldering iron), you get the issue of their own Linux distribution writing huge amounts of data to SD card every day and eventually wearing out flash chip on the SD card.

Mmmkay... But it's Linux, so you can fix that pretty easily yourself.

They should provide a board with an on-board power supply (ideally with a battery connector for internal UPS) and an m.2 port for a proper SSD with proper wear levelling.

You know, you can probably get a CM4 carrier board that does all that you're demanding, but of course, then you're likely to blow past the $35 price.

I think the RPi4 is really pretty good just the way it is. I'd like RPi5 to be faster, snappier video, lower power consumption, and maybe I'd be okay with giving up the USB 3.0 ports for an M.2.

I really don't need or want them to put an expensive power supply circuit on it. I've managed to put together my own power supply and I'm pretty happy with it: a $5 DC-DC step down converter from eBay/China, a couple big-ass AGM batteries, and a couple more $5 DC-DC CC/CV step down power supplies to keep the AGM's float charged via old laptop wall bricks. The power could go down for DAYS and I'd have a functioning computer, although I still need to hack my monitor to run off the AGM's.

[coder]

The factories with the old nodes are maxed out on production, but nobody is building new 28nm fabs.

I think some older fabs are transitioning to relatively newer nodes. Also, China (SMIC) is definitely bringing online new 28 nm capacity.

https://www.anandtech.com/show/16568/smic-to-build-a-new-fab-in-shenzhen-production-to-start-in-2022

the auto industry has kept their 28nm chips for a decade without updating because they didn't want to have to validate them again.

A decade ago? No, they wouldn't have been using 28 nm that soon.

[coder]

I suppose there are some applications that will actually make use of all that memory if you run the RPI as a desktop computer, but I haven't ran into one yet!

It's not uncommon for a C++ codebase to use > 1 GB/job, during compilation with gcc >= -O2. I've seen people complain that they've had to restrict the number of jobs to < 4, when building Chromium on a Pi v4 (4 GB, I think).

[coder]

wen rpi5

Don't get your hopes up for a massive improvement. Given their price constraints and prior plans, they're likely to stay on a 28 nm node, which means no massive improvements should be expected.

I think they're likely to improve the GPU more than the CPU cores, because the GPU is the most outdated aspect of the Pi v4.

[coder]

you get the issue of their own Linux distribution writing huge amounts of data to SD card every day and eventually wearing out flash chip on the SD card.

If your root partition is on a sd card, then you should really get a high-endurance model, like those made for professional and industrial use cases.

They should provide a board with an on-board power supply (ideally with a battery connector for internal UPS) and an m.2 port for a proper SSD with proper wear levelling.

Get an ODROID device with eMMC storage. I have a N2+ and it's pretty awesome.

[smitty3268]

I heard it slightly different: the car industry embraced "just-in-time" manufacturing decades ago. So when demand stops, there is no stockpiles of unused components. With covid, car sales dropped hard, so they pull the plug on components orders. But demand restarted too fast, and suppliers (in theory) couldn't ramp-up production fast enough. So it is car executives fault for keeping margins too thin. And now they are having a hard time explaining that in the investors meetings.

It's both, actually.

The 28nm (or whatever old nodes they use) production is maxed out, but the car manufacturers (or rather their suppliers) did have a chunk of it reserved.

During the shutdown they cancelled all their orders, and their production got sold off to others instead.

Now they realize they need a bunch of supply real fast, but those plants have contracts with other companies and can't just cancel them to give the automakers priority. And they've been attempting to bring on additional resources as much as possible but nobody even makes that old equipment anymore so it's fairly limited what they can do once they're running at 100% capacity.

If they were on a newer node, it'd be easier for them to bring up new capacity by throwing money at it.

[M@GOid]

They don't, and that's precisely the problem.

The factories with the old nodes are maxed out on production, but nobody is building new 28nm fabs.

All the new production coming online is 7nm (ish, depending on the company) and smaller, but the auto industry has kept their 28nm chips for a decade without updating because they didn't want to have to validate them again.

I heard it slightly different: the car industry embraced "just-in-time" manufacturing decades ago. So when demand stops, there is no stockpiles of unused components. With covid, car sales dropped hard, so they pull the plug on components orders. But demand restarted too fast, and suppliers (in theory) couldn't ramp-up production fast enough. So it is car executives fault for keeping margins too thin. And now they are having a hard time explaining that in the investors meetings.