I'm tired of information being vague, under-specified, or only available under NDA (if you're lucky). I'm not stupid enough to hop on this ride again.
Are there any fully open (in terms of schematics, firmware) RISCV rpi-"compatibles" out there? I'd be happy to pay triple the price of this thing for a power-efficient linux-capable sbc that is open.
Genuinely curious: why does their announcement upset you so much?
Most of the tech world announces products, executes a marketing strategy and then releases stock into the market.
You make it sound like someone left you standing at the altar. You didn't know it existed an hour ago. If you were on a long vacation, it might have released before your return. Why get angry?
Not only that, they wont even release all the information needed to it for certain applications. The x86 platform is far more open than raspberry pi ever was.
Why do you assume this specific announcement is what upset me? Why do you assume this has anything to do with their product marketing strategy? Why do you assume I wouldn't have gotten angry even after they released the product?
Any board based on StarFive's JH7110 is currently best in this regard, I think. Datasheet & reference manual for this SoC is available.
Especially their VisionFive 2 board. I've even downloaded a schematic for it (although older revision than actual board I have). And they're pretty good about upstreaming drivers.
That said: what you probably care about is documentation for integrated peripherals (esp. GPU), and existence of open source, mature drivers for those.
RPi is very good in this regard. Afaik the only binary blobs there is some GPU/SoC firmware, and (maybe) some boot code.
RPi's in general are very well supported & documented, and its software ecosystem is very mature compared to anything RISC-V based.
Could you pinpoint what you think is lacking there?
Other ARM based boards may offer more bang/$. Likely at the cost of documentation or driver support (Beagleboard being an exception).
"Raspberry Pi 5 is faster and more powerful than prior-generation Raspberry Pis, and like most general-purpose computers, it will perform best with active cooling. The Raspberry Pi Case for Raspberry Pi 5, with its integrated fan, is one way to provide this."
They pose a question themselves, and don't even answer it. Of course something will perform best with active cooling. Does it need it?
I don't need this wishy-washy marketing language from Raspberry Pi.
I hate it when someone answers a yes/no question with something other than yes or no.
Here’s my proposed edit:
Q: "Does Raspberry Pi 5 need active cooling?
Original A: "Raspberry Pi 5 is faster and more powerful than prior-generation Raspberry Pis…”
Better A:
"For modest workloads, no. For heavier workloads, you will get better performance with active cooling. Raspberry Pi 5 is faster and more powerful than prior-generation Raspberry Pis…”
It literally tells you that you don't need cooling, but if you add it you'll get more performance.
I don't find it hard to understand what this means: the soc limits it's core performance based on thermal conditions and will throttle when hitting limit temperature. That's standard behavior on every computer or smartphone or GPU out there.
Make temperature lower and it will clock and run at higher speeds without throttling.
They can't tell you how the board will behave on any random case you decide to put it on.
The uncased requirement is exactly what I'd expect to see there. Other than that, only if they decide to get really technical (they should) and tell you dissipating power / °C and temperature limits.
The answer is nuanced. It can run workloads somewhat faster and cooler than Pi 4 without active cooling. But it also can't reach close to its peak performance without active cooling.
"The combination of a newer core, a higher clock speed, and a smaller process geometry yields a much faster Raspberry Pi, and one that consumes much less power for a given workload."
pi5 with active cooler is about 1.2-1.5x faster than pi5 without cooler for most workloads that care than without.
Pi5 with active cooler is about 2-2.5x faster than Pi4. So Pi5 without cooler is probably about 1.5x faster than pi4, depending upon workload. (And more than this for quick bursts where thermal mass wins).
And if you dont have a heatsink and fan of sorts just use alternativing fingers on the cpu, they can still absorb about 10-15 DegC off the cpu temp and thats overclocking a 3b in the 1.35Ghz range. Surprisingly robust. Sadly cant get it to idle below .6Ghz yet, that needs more work.
But it makes wonder how much more phone manufacturers could squeeze out of their phones, although Apple are definately overclocking the 15.
It depends on what "need" means. I'm pretty sure you can take a 400W TDP Threadripper and run it without active cooling. It will throttle down to run at whatever speed (well, TDP) that doesn't fry it. The Raspberry Pi does the same thing.
If your goal is to get the highest score on every benchmark, then yeah, you need active cooling. That has been true on every Raspberry Pi, I think. (I don't remember if the 1 needed active cooling. I did not have any. I also remember it taking over a day to recompile Linux! Still faster than setting up a cross compiler at the time ;)
And how hot does the bulb get? Googling yields "The surface temperature of incandescent light bulbs varies from 150 to more than 250 degrees" (https://www.pacificlamp.com/temperature-of-a-100-watt-bulb.a...). Which, googling says, is likely hot enough to damage a CPU.
150F is like 65C which loads of CPUs run fine at. Going past 212F (100C) is usually where you'll start having problems with a lot of common desktop processors.
I think they answer your question in the PSU section “Raspberry Pi 5 consumes significantly less power, and runs significantly cooler, than Raspberry Pi 4 when running an identical workload.”
That's not documentation of any reasonable level though.
An MPU designer expects to see something like "200mA draw from the 1.2V power-domain when running at 400 MHz" or "10mA draw from the 1.2V power-domain when in first level of sleep". (Maybe not this small since Rasp. Pi is a more powerful chip, but... you know... actual specifics).
"Will my Raspberry Pi 4 power supply work with Raspberry Pi 5?
"Raspberry Pi 5 is a higher-performance computer than Raspberry Pi 4, and you may have problems using an under-powered supply. We recommend a high-quality 5W 5A USB-C power supply, such as the new Raspberry Pi 27W USB-C Power Supply."
The question asks about power requirements, but the answer is about performance?
The first time I read that I thought the 5 needs more power than the 4, not less.
But then you get 5V 5A ... 27W that is clearly incorrect too. So my guess is nobody is proofreading the technical specifications, and everybody that cares was kept away from that page.
You don't specify a PSU by its energy consumption. You specify it by output.
Also, that 92% efficiency, is believable, but a bit high for a 5V 5A PSU (this is a difficult combination). I would expect any such unity to be marketed as high-efficiency.
If you frequently work it really hard, it'll have larger temperature swings and may fail earlier, but it'll still probably last quite awhile. The failure is not likely to be catching on fire.
Beaglebone (from Texas Instruments) is more open, but still not as open as you'd probably like. Still, its a better balance than what Rasp. Pi organization has (more documents are available on AM335x, open-hardware for Beaglebone Green and reference designs, full chip specifications and the like). Beaglebone isn't really "more expensive", as much as its just "lower specs at the same price" though.
The "most open" are MPU chips and their associated "System on Module" boards. This isn't quite a SBC, but its easier to use than a BGA. These SoMs are very poor from price/performance perspectives, but instead serve as reference designs and/or prototypes to the $8 or $9 chips. The overall expectation is that you're "supposed" to be building your own PCBs eventually, so the SoM are kind of just a prototyping aid.
Most SoM provide 100+ pins from the chip as well, meaning you absolutely have to build a PCB to use them. However, 2-layer boards solder very easily to a SoM with castilliated edges (even with a hand-soldering iron)... albeit with a bit of flux and technique and practice. Its just the easiest way to deliver the most-pins of customization in the smallest space. So a relative beginer should be able to boot an SoM. The most difficult routing and Power-Delivery-Network details are already solved on an SoM, you just gotta apply power and build out the final interfaces / connectors.
Take the ATSAMA5D27-SOM1, 104-pins in a 40x40mm form factor. $50 from Mouser for 500Mhz and 128MB RAM (though fully open source and fully documented at linux4sam, and processor manual, U-boot process and everything). But the underlying SiP (MPU + DDR2 RAM) is like $15... while the MPU alone is like $8 and 128MB of DDR2 RAM is only like $3.50 in practice. Since in mass-production, you'd probably have a custom PCB anyway, that's the most expected use case. https://www.linux4sam.org/bin/view/Linux4SAM . I'd say that Microchip / Atmel's MPUs seem to be the best documented that I've found, but are unfortunately the lowest specs. Still, they also have some of the lowest power-consumption (like 200 mW or something), so really they're in a low-power class of their own. Still Linux though.
------------
STM32MP1 is the MPU from ST Micro. Like the Microchip SAM-MPU series, the STM32MP1 is available in SOM, SiP, and "raw" MPU form. Except the SOMs are like $100+, the SiP is like $50+, while MPU is $10ish.
-------------
I know NXP has a huge line of MPUs. I haven't researched them yet though.
---------------
I think all the hardware designers at this level just "assume" that their customers, if they care about "open source", are probably making their own PCBs.
If someone "just" wants a SBC (like the Rasp. Pi), there's not much point in publishing a ton of documents. People can just boot the Rasp. Pi and start messing with Linux.
-------
I got no experience with this yet. I'm just curious and am thinking of a simple MPU layout project ever since I discovered that OSHPark has 6-layer boards and KiCAD supports BGAs in practice. Overall, these lower-power lower-end MPUs fill a different niche than a Rasp. Pi ever would. But I feel like there's enough overlap that these might scratch your "open source" and "fully documented" itches.
I always thought the beaglebone had a better hardware design. The thing I first noticed was the female header pins - why would the pi have pins that can be shorted out?
the beaglebone pru is cool too.
But all of that pales in comparison to the huge mindshare the pi has, which makes all the difference.
Are there any fully open (in terms of schematics, firmware) RISCV rpi-"compatibles" out there? I'd be happy to pay triple the price of this thing for a power-efficient linux-capable sbc that is open.