The Raspberry Pi 5 is the new hotness from the Cambridge-based single board computer vendor, thanks in part to its new wonder-chip peripheral that speeds up much of its interfacing with the world. The RP1 hangs of the CPU’s PCIe bus and takes on many functions previously in the SoC, and those curious about it now have a little bit of information. Eben Upton has posted an article about the chip, and there’s a partial datasheet and a video in which the engineers talk about the chip as well.
The datasheet is intended to help anyone wishing to write a hardware driver for a Pi 5, and they admit that it doesn’t reveal everything on the silicon. We don’t expect them to put this chip up for sale on its own because doing so would enable their competitors to produce something much closer to a Pi 5 clone. It does reveal a few nuggets, though; there are a couple of Cortex M3 cores for housekeeping, and alongside all the interfaces we know from earlier boards it has, perhaps most interestingly for Hackaday readers, a 12-bit analogue-to-digital converter. This has always been on our Pi wishlist and is a welcome addition.
So, if you read the datasheet and watch the video below, you’ll learn a lot about the interfaces and how to talk to them, but not quite so much about the full workings of the chip itself. They hint that there’s more to be released, but since the Pi people have a history of not letting go of the family silver, we won’t expect the keys to the kingdom.
Have a read of our Pi 5 launch coverage.
> GPIO pins are 5V-tolerant, and 3.3V-failsafe (they may safely have a voltage of up to 3.63V applied
when RP1 is unpowered
Very nice. There are also some comments in the blog about there being a spare M3 core which might be able to run PIO on the GPIO pins at some stage.
It’s a real shame that they seem set on not selling RP1 separately though, but I can see the point you’ve made.
They could for sure sell the rp1, and both get a nice bottom line boost and making it available.
I imagine that they pay a few $ a piece? So sell it in bulk at 25-30$, anyone trying to make a clone will have a huge disadvantage trying to compete, and they would get a big slice of profit either way.
Also a disadvantage for anyone wanting to make say a PCIe card with an rp1, but that would just raise the price a bit, and those could still go for maybe 100$. Acceptable if you have a usecase.
If you think the price is too high, then pick another suitable ic off the market. If there isn’t one, well then 25$ is pretty cheap isn’t it?
A too high price may well reduce their lead by adding extra incentive for a clone…
For 25$ in bulk an FPGA solution likely will win for pci-e IO card manufacturers.
Not selling them to prevent easier clones, isn’t that a bit against the open-source / open hardware-vibe?
I can assume they aren’t producing more than they need right now, but if demand is high enough I can imagine also providing them seperately like the RP2040. They don’t seem to mind the clone market there either?
The raspberry pi have never been open hardware. They’ve gone out of their way to keep it closed. Look for info about the history of the ODROID-W
The Raspberry Pi SBC contains a CPU that is licensed from Broadcom.Short of signing an NDA, you’re never going to find out exactly what’s making that particular chip tick in its entirety. I agree with the other comments too. No way is RPF going to give away the RP1 silicon as opensource. The cloners would have a field day and you’d probably get a lot of janky pi clones.
The raspi foundation has been extremely proprietary since they signed a deal to be the sole buyer of their SoCs/
Rpi has never been 100% open-source and (ideology side) that’s fine – as long as they are publishing enough info to do what people want while still making enough profit to reinvest in new products.
I’ve been SoC shopping a few times over the years and honestly Rpi’s offering is very hard to beat in terms of support, documentation, available OS’s and the sheer volume of users & hardware out there.
I still can’t figure out what sort of company that Raspi foundation is.
Are they a non-profit that is intended to help the educational market and spread information, or are they a commercial company that guards their own “ip” jealously to prevent others from learning their “secrets”.
Apart from that. I would like PCI-E to become a standard peripheral in generic microcontrollers. Those UsARTS, SPI and I2C channels are so ’80-ies. I don’t know if they still do it, but PCI-E used to be used in oscilloscopes to move the sampling data around (for example from an FPGA to the main processor).
The Raspi Foundation is the educational charity. Raspi Trading was, until the last few years, solely owned by the Foundation, presumably as a way of protecting the charity in case the company went under while trying to design and sell boards. As far as I remember another entity purchased a stake in Raspi Trading a few years ago, I assume that’s because of things like the cost of developing RP1. I guess it’s a fine balance of trying to help people to learn to program whilst not helping your competitors to trample you.
Even educators guard the answers to the tests…
Until you offer them a raise, if the students get the answers right.
Then they wright the answers on the chalkboard.
Managers of educators get what they manage for.
As with all PHBs with low pay/IQ staff the metric gaming is obvious.
I’m here to check your paperwork: it seems (checks clipboard), yes it seems you failed to fill out the standard Soap Box Permit application. Unfortunately we’ll have to confiscate it and assign you community service as a kindergarten classroom helper. Thank you for your cooperation.
(IFIB-PD)
P.S. “write”, “PHD’s”; also please edit for intelligibility.
🤣
dude they are both, and also they dont want other ripoff companies that are solely based off making profit. i agree with your pcie statement tho. they are developing their own silicon, and obviously they dont want people leaking them and using their own designs for profit.
A PCIe controller is enormous in size and energy consumption compared to UART, SPI and I2C. In nearly all embedded applications you don’t need that bandwidth. Oscilloscopes are one of the few high bandwidth applications. Other are perfectly fine at <1/1000 of the bandwidth of a PCIe lane.
The education thing was just good marketing, it’s more about making Eben richer
Eben has made himself fairly well off, but he’s not a billionaire. The education thing may be good marketing, but it’s also something the Raspberry Pi foundation is actually doing. I can’t see any reason to be negative about him other than your own entitlement.
Why not? Nothing wrong with getting richer. When you have a good idea and make it successful, you should get a cut of the pie. I don’t see a problem here. A lot of hard work has went into this project.
Neither do I see anything wrong with keeping your cards close to your vest so to speak. As long a good product is produced that you and I can ‘use’ at a decent price. All for it. Great!!!! Don’t see why some ‘criticize’ a successful project!!
Of bugger off is it… I’m sure Eben’s done OK for himself but Rpi are not exactly evil corporate overlords here, and their original intention / target audience (education + improving access for kids + getting kids into coding & electronics) is still very much in evidence and to be applauded.
Compare other tech companies forays into the educational markets that are far and away more cynical grabs at early userindoctrination & profiteering and it’s very odd to pick on the Pi foundation.
> “Are they a non-profit that is intended to help the educational market and spread information, or are they a commercial company that guards their own “ip” jealously to prevent others from learning their “secrets”.
Can’t them be both?
I mean if all the value they create goes to copycat how can they fulfill their educational mission?
People have such an entitled attitude – the Pi Foundation and Pi Limited don’t exist to give us free or cheap stuff AT ALL. It’s only a bonus and a privilege that we get to share in the cool toys they’ve made while working towards their educational goals.
But that doesn’t matter to the unwashed masses – it’s more important that they should be allowed to buy 100 Pi 4’s for $20 with overnight delivery and if Pi Limited goes broke and has to stop teaching IT skills to schoolchildren as a result then that’s fine by them. *facepalm*
Isn’t the “educational mission” just propaganda? Things may have started that way once, at least in words, but those days seem all but forgotten and they are just another company these days that happen to make one of the most popular single board ARM based computers. There are plenty of other competitors at similar price points, just not as popular.
“I haven’t bothered to find out what they’re doing, so they mustn’t be doing anything” is quite the take.
This entire thread underscores a fundamental
1. Misunderstanding of what a non-profit is. It’s at base simply a tax structure that requires reinvestment in the organization
2. A conflation of the altruistic missions of many non-profit organizations with their tax status.
3. A wilful ignorance that the UK and US are not the same country.
Reinvestment: verb. To pay all net cashflow to the principles and never show a profit.
Every consider WHY it is that the others aren’t as popular?
And here is why I hate the PI and have come to love the RP2040 on the Pico.
Documentation
The RP2040 is entirely open, but the big Pi is by no means open. I had always blamed this on Broadcom, but maybe I was wrong and this is all about decision by the Raspberry Pi Corporation.
Indeed, the documentation AND tooling around the rp2040 is amazing. Lots of things are well thought out and a pleasure to work with.
It is quite the contrast to the Pi where it is “run this image and do not ask questions on where it came from”
Pi has always had an EXTREMELY close relationship with Broadcom.
They’ve done a good job of making Broadcom suck less, but… There’s only so far they can/are willing to go to de-suckify the partner whose goodwill they are EXTREMELY dependent on.
Pi Foundation is just Broadcom’s way of trying to get the Linux community to hate them a bit less without bothering to fix the core reasons for them being so hated – since only Pi products are desuckified Broadcom, all of the rest of Broadcom’s product lines are “avoid like the plague that they are”.
The RP1 chip actually helps them cut ties to Broadcom. It allows them to have the I/O they are known for while having more freedom with the CPU it’s paired with. The SoC itself has a lot less responsibility to make boards ‘RPi compatable’. I hope this is a sign that they may be moving away from Broadcom, but given Eben’s position, it would probably take some fundimental management changes to make that process continue.
So far nothing remotely as functional, performant, complex and well supported as the Pi has been better documented. It would be nice if all electronics came with complete down to the logic gate documentation of the hardware for the folks that want it. But with the shear amount of work that has to go into just making the documentation fit for publishing and updated I’d suspect your favourite SBC’s would jump up a fair few quid each, and that is assuming you made all the silicon so are not locked behind any NDA anywhere…
I’d argue the Pi is and has been very supported of the concepts of Open hardware, eventually opening up bits you’d never expect to be opened. More is and always would be nicer, but at the same time you don’t want to become the troubleshooting and bugfixing department for all the cheap clone knockoffs of your product, that would kill your reputation and stop you making any money to pay the staff, let alone develop the next product…
> you don’t want to become the troubleshooting and bugfixing department for all the cheap clone knockoffs of your product
I suspect this is the primary reason. I’ve seen so many open hardware devs burn out because as soon as their product becomes even mildly successful someone else makes a near identical board yet cheaper and crappier board and then slaps their firmware on it, directing all help/documentation requests and complaints back to them.
> We don’t expect them to put this chip up for sale on its own because doing so would enable their competitors to produce something much closer to a Pi 5 clone.
That’s the POINT. Release the chip, Ebon.
What I would absolutely love is a cheap PCI-E addon beard with a RP1.
Hell, do it in m.2 as well please.
Yes, for at least the M.2 version you would need a cable to run to a breakout board for the actual IO, but still…
There is, essentially no current way to get truly low latency IO into a modern laptop. And few for a lot of desktop systems.
Just look at what it takes to get to the point of taking in GPS PPS.
RP1 could give that, and so very much more.
This was exactly my thought when I first started reading about the RP1…
One of the best things about an rpi is an easy way to make GPIO be able to talk to a computer and higher-level protocols in a sane way.
Having something like this that I could drop into an x86 computer or thin client to have the same rpi interface would be phenomenal!
That flexibility also goes a different direction…. this could be the first steps to offering something like RISC-V for the processor.
Wouldn’t that be an interesting proposition?
Would probably require someone with decent PCB designing skills and mad rework skills to make a PCI-E card with the chip and showing off what that enables for the momentum required for that to happen
Going by one of their posts about it, they actually made a PCI-E board for testing purposes.
You have some commercial m.2 FPGA boards for around 100$, eg: https://rhsresearch.com/products/litefury
If you just want to read PPS an pci-e serial port card might be a good option.
Looking at the header image, i cant stop to wonder how the memory size select resistor array should be working. One line going to four resistor placements, all having one pin grounded. How should that work? Analog divider where higher resistance to ground means bigger memory?
Yeah, different resistances to ground would be my bet. The memory is presumably soldered on, so this isn’t intended to be user-selectable. They just arranged the silkscreen around the sense resistor to conveniently indicate which of four different resistor values was populated and which memory size that corresponds with. My take at least, YMMV.
I have long admired the BBB (Beagle bone black) for its excellent documentation — and have chosen it over the Raspberry Pi for that very reason. I don’t see that providing documentation details has harmed the bottom line for the BBB in any way. But as the article says, we are seeing some documentation now on this new chip, and maybe a full datasheet is on the way.
I have a BBB acquired quite a few years back when I was vacillating on which I should concentrate on to ‘play’ with and learn for projects. I recall the OS was a bit more painful to setup at the time compared to the RPI. There was something funky about using it that I don’t recall right now… Anyway, I don’t use the BBB at all now and have been ‘all in’ with the RPI group of products ever since. Really liking the Pico for most little automation projects and the RPI-4 for Linux required projects. I have a pile of Zeros just sitting there now as Pico boards have taken over their role. Looking forward to the RPI-5 just because :) . Another nice line of boards I use is the Adafruit Metro boards. The Grand Central was a super find for one of my projects. With a RPI-4 talking to a Grand Central via USB I had lots of I/O available for the project.
Android box much cheaper with better specification, but no I/O connector. The Android can be replace with Linux Armbian. If you need some I/O you can add mricrocontroller that connected to USB or Wifi/Ethernet.
I really don’t understand the perspective of people saying that the RPi is well documented, the only thing that I can say is that the documentation is readily available, but that is not the same thing.
Large parts of the raspberry pis are undocumented and rely on binary blobs. whereas there are vendors that have far more comprehensive documentation available with fewer dependencies on binary blobs, although less accessible for many due to language barriers.
If you are factoring in community supplied documentation and findings in that then that’s not unique to the Pis either.
the RP2 is probably the closest thing I would argue is well documented from the foundation.
Does it really matter??? As long as it 1) Runs Linux, Ultibo, RTOS, home grown OS, etc…., 2) One can access all the exposed interfaces (documented), who really cares about binary blobs? A bunch of ado about nothing in my mind.
Be nice if you could have more openness as there is a reason folks went through the effort to make stuff like Coreboot a thing. But I mostly agree – when dealing with a computer, which is despite the microprocessor overlap what a Pi is you really can’t avoid closed stuff easily and a few bits of secret sauce is very much par for the course.
I think RP1 is an amazing development and if it’s made available it creates exciting opportunities. I’m kind of thinking of the possibility of embedded Pi. Instead of buying a Raspberry Pi board, you’d get an M.2 board for a laptop or NUC device. Another possibility that comes to mind is industrial control requiring multiple Pis. Instead, one could develop a multiple IO card with multiple RP1s driven by one AP, or a cluster of RP1 boards sharing one source of control. This for me could be a new Pi game changer beyond the regular board iterations.
That bad headline is just gonna live on forever in the url slug I guess.