Raspberry Pi Supply Chain Issues Beginning To Ease Up

[phoronix]

Phoronix: Raspberry Pi Supply Chain Issues Beginning To Ease Up

The Raspberry Pi team has a positive supply chain update with some good news ahead of Christmas and when they expect to reach pre-pandemic supply chain levels...

https://www.phoronix.com/news/Raspberry-Pi-Recovering-In-2023

[vegabook]

I'm tempted to say "when are we getting the RPi5", but actually, the relatively low performance of the Pi4 is almost a feature. It's just good enough (barely) to run a GUI on, and firefox is so painfully slow that it acts like a kind of digital-media-detox incentive, while still allowing for lighter-html sites like Stack Overflow to be perfectly usable. Also, I do high end quant dev and data engineering work, and I do a large proportion of my exploratory work on my Rpi4. That's because the Pi's modest perforrmance is brutally effective at showing up inefficient code paths. It's amazing how much perf you can actually squeeze out of this thing if you spend just a bit of time on hardware sympathy and even vectorization (Neon). I even got most of the wgpu examples to work on the RPI4's Ubuntu 22.04 built in Vulkan implementation, and that looks like only getting better. Most devops stuff even works too on Ubuntu 22.04, so you can prototype all sorts of mini devops environnments on it (or on a small cluster of them). A wonderful little device and the fact that the Foundation is trying to keep the prices right, even going as far as to point out which resellers are contracted to sell at MSRP, is very reassuring.

The only small chinks in the armour are NVME support (but I suppose that would be expensive), and also I hope their relationship with ARM is not so strong as to preclude them from exploring a RISC V version at some stage.

[M@GOid]

I think the best way to bring down RPi prices is simply to buy a option from the competition. ARM on Linux cannot be kept hostage of the RPi. Its competitors often offer more bang for the buck. And software support for those devices will never reach parity with the RPi if people only have eyes for one vendor. Just imagine what computing would look today if people never bought IBM PC clones.

[vegabook]

I think the best way to bring down RPi prices is simply to buy a option from the competition. ARM on Linux cannot be kept hostage of the RPi. Its competitors often offer more bang for the buck. And software support for those devices will never reach parity with the RPi if people only have eyes for one vendor. Just imagine what computing would look today if people never bought IBM PC clones.

problem is the competitors are not clones. As soon as you do something a bit out of the ordinary, and if you're not a greybeard, high risk of esoteric fails that can take days of obscure forum surfing to resolve. Always the same story. That costs me much more money than the rpi premium. I've run cubietrucks and beaglebones and jetsons (though the jetson was pretty damn good, but looks to be discontinued in hobbyist form).

Your point is a good one generally but we need proper, digital interface-identical clones. Do such things exist?

[stargeizer]

The problem is not the RPi "keeping ARM hostage". ARM doesn't give a %#&( about anything open source, and that's the main problem with the ARM ecosystem in general.

The RPi developer(s) and Broadcom have workarounded the ARM licensing by keeping the "trade secrets" of the design in the firmware, and opening what they can. They still do care about the bugs in the system and tend to fix the bugs in the software. Now compare that with the competitors: They just use whatever Rockchip/Allwinner can sell to them and they only cares about the blobs ARM/Rockchip gives to them, and that's it. You got a bug??, learn to live with that. Do you have a RK3288 based card? Kernel 4.4 is the only supported. Anything else is uncharted territory. (more modern RK chips only got 4.14 working.). Did somebody said "Librepotato"??. Well... expect a slower potato than the Pi4... but at least have support.

The difference with the IBM PC compatibles, is that the "compatibles" machines literally forced an "open" ecosystem, back then. And IBM fought them in the courts (and lost). ARM has not this problem, they just sells designs, and there is competition (bot nobody cares about MIPS nowadays) but the software side of the design is kept as closed software.

[ayumu]

As an aside, VisionFive 2 is starting to ship; there's some reports of people receiving them in the last few days.

[EvilHowl]

As an aside, VisionFive 2 is starting to ship; there's some reports of people receiving them in the last few days.

A Raspberry Pi competitor using a CPU with an open-source ISA like RISC-V has always been my dream. I hope it gets enough attraction.

[Kabbone]

The problem is not the RPi "keeping ARM hostage". ARM doesn't give a %#&( about anything open source, and that's the main problem with the ARM ecosystem in general.

The RPi developer(s) and Broadcom have workarounded the ARM licensing by keeping the "trade secrets" of the design in the firmware, and opening what they can. They still do care about the bugs in the system and tend to fix the bugs in the software. Now compare that with the competitors: They just use whatever Rockchip/Allwinner can sell to them and they only cares about the blobs ARM/Rockchip gives to them, and that's it. You got a bug??, learn to live with that. Do you have a RK3288 based card? Kernel 4.4 is the only supported. Anything else is uncharted territory. (more modern RK chips only got 4.14 working.). Did somebody said "Librepotato"??. Well... expect a slower potato than the Pi4... but at least have support.

The difference with the IBM PC compatibles, is that the "compatibles" machines literally forced an "open" ecosystem, back then. And IBM fought them in the courts (and lost). ARM has not this problem, they just sells designs, and there is competition (bot nobody cares about MIPS nowadays) but the software side of the design is kept as closed software.

I have a Rock64 with a RK3328 and I would say it's quite well supported by mainline. In general I feel pine64 is doing a great job to gather community efforts to mainline as much as possible of their boards.
Mostly you are screwed if you want to use gpu acceleration, but this is not my use case normally

[schmidtbag]

I'm tempted to say "when are we getting the RPi5", but actually, the relatively low performance of the Pi4 is almost a feature. It's just good enough (barely) to run a GUI on, and firefox is so painfully slow that it acts like a kind of digital-media-detox incentive, while still allowing for lighter-html sites like Stack Overflow to be perfectly usable. Also, I do high end quant dev and data engineering work, and I do a large proportion of my exploratory work on my Rpi4. That's because the Pi's modest perforrmance is brutally effective at showing up inefficient code paths. It's amazing how much perf you can actually squeeze out of this thing if you spend just a bit of time on hardware sympathy and even vectorization (Neon). I even got most of the wgpu examples to work on the RPI4's Ubuntu 22.04 built in Vulkan implementation, and that looks like only getting better. Most devops stuff even works too on Ubuntu 22.04, so you can prototype all sorts of mini devops environnments on it (or on a small cluster of them). A wonderful little device and the fact that the Foundation is trying to keep the prices right, even going as far as to point out which resellers are contracted to sell at MSRP, is very reassuring.

The only small chinks in the armour are NVME support (but I suppose that would be expensive), and also I hope their relationship with ARM is not so strong as to preclude them from exploring a RISC V version at some stage.

Well, considering the RPi series was never meant to be used as a cheap desktop computer, it's relatively low performance isn't a bug but it's not necessarily a feature either. You're not using it the way it was intended (the SoC in particular). The RPi was originally intended for educational purposes and hobbyists. So, it's kinda like using a pipe wrench as a hammer and saying "this helps me take my time and plan my strikes carefully" as if the wrench was supposed to be used that way. While a RPi4 is capable of basic desktop usage (at least with well-optimized software), a home server, or a HTPC, you can spend a little more money on other ARM platforms that are better suited for the job. Though I get the impression Broadcom's GPUs are probably in the best shape for Linux, so perhaps RPis are still a solid option for HTPCs, at least if the Ethernet isn't a limiting factor.

[vegabook]

Well, considering the RPi series was never meant to be used as a cheap desktop computer, it's relatively low performance isn't a bug but it's not necessarily a feature either. You're not using it the way it was intended (the SoC in particular). The RPi was originally intended for educational purposes and hobbyists. So, it's kinda like using a pipe wrench as a hammer and saying "this helps me take my time and plan my strikes carefully" as if the wrench was supposed to be used that way. While a RPi4 is capable of basic desktop usage (at least with well-optimized software), a home server, or a HTPC, you can spend a little more money on other ARM platforms that are better suited for the job. Though I get the impression Broadcom's GPUs are probably in the best shape for Linux, so perhaps RPis are still a solid option for HTPCs, at least if the Ethernet isn't a limiting factor.

the analogy of the pipe wrench leaks like soggy paper cup (as is often the case in your posts, if I may venture). _For my use case_, namely coding mathematical algos, clearly stated in the post ("quant"), or exploratory devops (also clearly stated), the rpi has perfectly identical capabilities to the end deployment hardware, but with an order of magnitude or two less bandwidth and performance. It's not using a hammer for a wrench. It's a prototyping device.

But I know you like to argue for any ol' reason, with consistently sloppy logic, and all too often with minimal experience of the actual domain you are arguing about. The above won't be the first time, nor sadly, the last!