With the recent teardown of the Raspberry Pi 500, there were immediately questions raised about the unpopulated M.2 pad and related traces hiding inside. As it turns out, with the right parts and a steady hand it only takes a bit of work before an NVMe drive can be used with the RP500, as [Jeff Geerling] obtained proof of. This contrasts with [Jeff]’s own attempt involving the soldering on of an M.2 slot, which saw the NVMe drive not getting any power.
The missing ingredients turned out to be four PCIe coupling capacitors on the top of the board, as well as a source of 3.3 V. In a pinch you can make it work with a bench power supply connected to the pads on the bottom, but using the bottom pads for the intended circuitry would be much neater.
This is what [Samuel Hedrick] pulled off with the same AP3441SHE-7B as is used on the Compute Module 5 IO board. The required BOM for this section which he provides is nothing excessive either, effectively just this one IC and required external parts to make it produce 3.3V.
With the added cost to the BOM being quite minimal, this raises many questions about why this feature (and the PoE+ feature) were left unpopulated on the PCB.
Featured image: The added 3.3 V rail on the Raspberry Pi 500 PCB. (Credit: Samuel Hedrick)
Looking forward to seeing the complete list of parts and a set of instructions on how to do this. I’d like to have a Pi500, but I’m done with SD card onboard storage. There are 100’s of makerspaces that have the solder level skills to help the average consumer make this a possibility.
I’ll guess that after the Holidays that TEMU will be full of NVME PI500’s
The author of that mod posted the BOM on his Twitter profile.
I’d rather insert the Pi500 widthways into an intimate orifice than use Xitter but I’m sure it’ll soon be available on places that aren’t sewers of hatred.
Thanks, that was my thoughts exactly. Not going to Xitter to get information. I don’t want anything enough to enrich the X owners.
They are already rich.
FYI chaninging the “x” to “xcancel” in the URL will give you a cached version hosted on an independent provider. In case you ever have need to view something. It also includes all the comments without needing to sign in, just like it used to.
Why those features were left unpopulated is pretty obvious: It raises the BOM and manufacturing costs enough that another $5 or $10 would have to be added to the price. And Raspberry Pi really does try to keep these things affordable. We’ll just hope that the plus/pro module does show up before much longer.
I don’t really see how it would even add $5 to the BoM. It’s the 4 pictured capacitors, a 3V3 buck IC, a couple more capacitors, and an inductor. At quantity that can’t be more than $3, even with a 10W buck PMIC.
I had already decided to stay away from rpi products, I’m just not their target audience anymore, but the unusable m.2 slot seems bizarre to me.
The DC/DC’s literally 18 cents at qty 1000 from a distributor, they’d be able to get it cheaper. The additional BOM cost is without a doubt sub-dollar. Maaaybe if they weren’t clever it’d be what, 4 extra line items at most (forgot the inductor and likely a feedback resistor that’s probably unique).
Could be testing issues, but I’d guess with that clip-together design it’s the extra effort needed to make it so that it doesn’t need to be opened.
I was feeling the same way about the regulator, but then remembered that some drives can pull shocking amounts of power. Anecdotally, it seems like 10W is roughly the most you can expect, and when I checked Digikey it looked like, through them, 3A 3V3 bucks cost just over $2 at quantity.
But wait! I was looking on my phone so filtering was clunky. You can get a fully featured (soft-start, high-frequency, PFM mode) TI 3V3 3.7A buck converter for 29c. So yeah, <$1 BoM cost. Obviously, that was enough to make someone flinch, but I think it’s ridiculous.
In the article, the AP3441SHE-7B is mentioned as fitting the footprint: not clear that’s the one that was finally used, but assuming it was, it’s 18 cents in qty for Random User.
I’ve gotta figure that it’s the combination of testing and not wanting people to frequently open it. Seems like they want it to be targeted more to “here’s something like that raspberry pi board thing except this is fully contained and pretty,” so populating something that can only be used by opening it would be against what they’re shooting for.
@Pat: They could solve the opening issue the same way some laptop manufacturers do: with a removable rectangle over the M.2 area just big enough to swap in/out an SSD without having to crack open the whole device.
According to Jeff Geerling, they did it this way because they plan to reuse the board form factor in future devices and it leaves them room to add a M.2 slot (and PoE+ at the other end of the board) for such future devices.
Thanks for your posting.
For the typical use case of the Pi 500 an X class high endurance SD card is OK. In a school scenario, most of the use probably won’t be disk-bound. Even in a programming course, students will probably be spending much of their time staring at their code on the screen.
A work-around could be to set up a RAM disk for writing scratch (not the app, Scratch) files. The Pi 500 has 8G of memory so it is feasible to set up a 2G RAM disk and mount it on /tmp or some useful point and still have decent memory for gcc or a similar tool.
My guess is the design was locked down in ~2020 and there just have been million other things going on. They know and their sorry but hopefully it won’t take forever to offer as a “plus” feature.
Considering that the drive will be operating at most with a PCIe gen 3×1 connection I doubt it will use much power. It’s not like the gen 5×4 drives that use lots of power and need heatsinks.
For making and selling a product, take the cost, multiply by 2, then multiply by π, then you are almost close to the end user price increase.
You could finesse these numbers either way, but a 5$ price increase is reasonable even with the bom that cheap. Still a better choice in my book since it’s a tremendous usability upgrade – it’s not worth buying without an nvme(to me). For “normal” RPI stuff, I already have ones lying about. For anything I’m directly interacting with, aka things that the 5 might do that the four is not fast enough for, it needs an nvme.
One of the reasons to depopulate an area on a board can also be because something did not work as intended and it was easier to just depopulate it. Raspberry Pi has to plan for stability. They may have intended to include it but removed it later because they found a case in testing that created problems that could not be resolved in time to make production.
Doesn’t need to be opened? Not for me, then.
If I bought the 500, it would be with the goal of transferring the main board under a (low profile) mechanical keyboard.
Maybe even one with a trackpoint, for portability.
Raspberry pi doesn’t compete on price. There are already more powerful products in the same price bracket or as powerful products in lower price brackets from competitors. If I’m buying based off cost, I am not buying raspberry pi to begin with.
What raspberry pi competes on is support and features. There is absolutely no excuse to not spend $4 more and give us a better product when features like this is the entire point of buying raspberry to begin with.
Raspberry pi doesn’t compete on price. There are already more powerful products in the same price bracket or as powerful products in lower price brackets from competitors. If I’m buying based off cost, I am not buying raspberry pi to begin with.
What raspberry pi competes on is support and features. There is absolutely no excuse to not spend $4 more and give us a better product when features like this is the entire point of buying raspberry to begin with.
“It raises the BOM and manufacturing costs enough that another $5 or $10”
Yeah, no. The parts in total are sub-dollar in large quantity and it’s two line items.
I think it’s more they don’t want to encourage people to take it apart, and avoiding that would require an opening of some sort, and that would likely add to the price enough.
A dollar on the BOM is multiple dollars on the sales price.
It’s also more than two lines, there’s a regulator, the ancillary parts (6 if they’re using the datasheet example), the socket, the coupling caps.
They already have the socket on the BoM.
“A dollar on the BOM is multiple dollars on the sales price.”
If you’re accounting for the parts and assembly cost, what else is a BOM item costing you that you have to mark up for?
Additional testing costs is a possibility, I should’ve added that. Tend to think with a snap-on case it’s just that they didn’t want people opening it, clips’ll probably break easy.
Your margin, wholesale margin and retail margin.
They didn’t set “$90” to be the cost based on a fixed multiplier of the production cost. A sub-dollar production cost increase that improves demand wouldn’t’ve touched the price. A lower percentage margin doesn’t matter if you make more money.
The problem is that adding this doesn’t just do that, it also requires you to support it (testing, but also including supporting opening the thing) and the extra support costs kill you.
Recommend not feeding the trolls. This person is going to argue about costs he can Google 12months after the design was completed. Let alone the sw, documentation and support costs he failed to bring up.
Its all the support costs, that’s what I was saying. The parts/assembly themselves are nothing, it’s having to deal with “I broke this thing trying to open it” and “why doesn’t my Axcaf SSD I bought from AliExpress work.”
They already have the NVME hat, why should this be any different? Software isn’t an issue and the documentation just needs modified from the HAT documentation.
Why does it matter if the design was completed 12 months ago? They kept the footprint for a reason and it isn’t like voltage regulators have significantly dropped in price during that time. You can keep arguing if you want but it is a fact that in bulk quantities populating those components would add very little to the cost.
The case is also very obviously not designed to be opened by the user, so either they’d have to have an assembly step with multiple SKUs for SSD size, or they’d have to redesign a more expensive, user-openable case. Both add cost.
The case could be modified pretty easily. All of would really require is a snap in hatch which would just require another injection molded part and a relatively small and simple one at that.
The most expensive part both by purchase power and P&P is the connector, and that is already populated.
I don’t know where people got this idea but at least the earlier post clearly showed that it’s not populated.
But it’s also like 50 cents in quantity and a simple surface mount part, so it ain’t gonna drive anything.
yeah they do such an amazing job on price while putting a ton of components on every board that our initial “wow” gets replaced with a wrong-headed “therefore they should be able to add even more components at no cost.”
myself, i’m still full of wow when i look at, for example, the high parts count of the $4 pico board
How are you amazed at the price and part count of the pi pico? Have you never seen any other development boards? The pico is nothing special.
How do they do such a good job on price anyway? There are often better products for cheaper, the main thing that raspberry pi has going for it is support and community, not value for money.
Wouldn’t this just mean that they couldn’t get the nvm slot work reliably in all circumstances? That’s what I would first think of when it’s not populated at all.
Seems logical.
The article seems to imply the SSD has its own regulator so I doubt it’s run off the main 3.3V rail
I bet there are some devices that pulled enough power to brown the rail out or something pedestrian like that. They notice it late in the product dev cycle, after they’ve already put the footprints in, and they just don’t populate and leave the feature out for now.
They managed to make the NVME HAT for the pi 5 fine (including running it at PCIe gen 3 when it is only rated for gen 2) so if they couldn’t get it to work reliably in the pi 500 then that is a bit concerning.
Is it possible that there is not enought power on the 3V3 rail for some additional hardware?
Based on the number of times I’ve seen power warnings in the logs when running off a 45W PD adapter, I’d say this is likely.
Because your 45w PD adapter is mostly wasted on a device that only has 5v input. Hazarding a guess your adapter like mine puts out around 2.5A max over 5v
It doesn’t use normal PD and it doesn’t use anything near 45 W. PD has set voltage and current levels, a lot of PD adapters max out at 3 A, which might make sense for yours which could be capable of 5, 9, 12 and 15 V at 3 A, the 45 W would only be available in the 15 V 3 A output. 20 V 3 A which is 60 W is the max for normal PD. There are higher power but you need cables with e-markers in them to say that they can handle higher power. Then there is a 100 W limit, which comes from a 5 A current limit and max voltage of 20 V. Then there is further extensions up to 48 V 5 A. The main thing is that the device, power supply and cable all need to agree on the output and the lowest voltage and current supported by all components is chosen.
The pi 5 only accepts 5 V input, so on your 45 W charger which is likely limited to 3 A then the charger output will likely only be 5 V with a max current of 3 A, so only 15 W which would be why the pi complains about low power. The official pi power supply is capable of 5 V 5 A which gives a total of 25 W. The pi itself will generally use less than 10 W but it also needs to power anything connected to the USB or GPIO so that is why they specify 25 W as what they consider enough power.
There are ways to switch the warning off though which is useful as it only uses the USB input for determining how much power it can draw, if you power it through the GPIO it has no way of knowing how much power it can draw. If you power it with your 45 W charger it likely tells the pi that it can only draw 15 W which is lower than the 25 W threshold. Your charger should have something written on it or in the manual telling you which voltage and current configurations it supports.
They clearly built this thing to a specific sales price and margin.
at $90 , it’s enough under $100 to be ‘not $100’
at $90, it’s ‘only’ a bit more ($20 / 22%) more than the Pi 400…
if it were $100, people would think it’s LOT more than the pi 400 (where as it’s only 30% more)
ALSO….it’s clear there’s going to be some form of ‘PRO’ / higher priced version of the keyboard released in the future, so this is a clear product differentiation strategy…
also don’t forget, they need to sell at a price that doesn’t cannibalise other combinations of product (Pi 5 + POE hat + NVME Hat, etc)
lets say they’ve sold half a million Pi 400…
if they sell half a million Pi 500’s, a $2 increase on the BOM at the same sell price is a MILLION bucks gone for ‘no reason’
if they create a ‘PRO’ version, they could add $10 to the BOM for all the unpopulated parts, sell the device at a premium (say $140) and target the tinkerer market, the small percentage of people that dismantle everything they get their hands on to ‘add’ stuff…….
I mean, it wouldn’t be no reason: it’d increase the functionality and therefore demand. But yeah, you make more money by clearly differentiating products: the people who would want it with an M.2 are likely to pay quite a bit more for “nice features” supporting it, like, for instance, a clear list of supported SSDs.
No audio jack. Disabled this and that. Well, screw Raspberry Pi. Perhaps another company will try addressing the needs of hackers.
Raspberry Pi should have never been given open-source hardware status.
“Raspberry Pi should have never been given open-source hardware status.” Good news! It wasn’t.
Lots of open source lovers love the Pi because they are pretty open, but at no point have they ever been really OSHW or claimed to be. However being probably the most open SBC at least in the performance/price bracket and having excellent long term support as well…
The best option you have even if it isn’t perfect is worth celebrating for being so many steps closer to what you want – might just inspire a rival or create enough noise that the company making that option takes yet another step in the direction you really want.
Seems fairly obvious they want to test the need for it, and if they think there’s demand, release a higher-priced model with it next year.
And for the rest of us, we can just boot off an external SSD drive (I use Samsung T5/T7s) for ‘faster’ disk access rather than an SD Card…. Worked on the 4, now way better on the 5. :) .
Oh and you don’t have to pop open the case either to break something, warp something, bend something, lose something, etc :) .
do either you know where you are?
That’s some pretty lame reasoning, especially on a site called Hackaday.
Pretty sure most could handle opening some plastic and putting a square peg in a round hole.
The Pi has never been aimed exclusively at the mature hacker though – the target audience includes children etc! So requiring a careful disassembly that will almost certainly fatigue those clips to failure if you do it more than once or twice no matter how carefully you do it…
Really for a Pi product they would need either a hatch or some other case redesign IMO, especially as this model really is targeting the kids getting into robotics and coding more than anything else.
Because carrying a loose USB SSD is way more convenient than throwing an NVMe drive into the unit and having it all together.
The Pi 400/500 form factor is very minimalist and clean, and it’d be a lot nicer to have everything in the case, IMO.
Then you have something detachable and relatively fragile hanging off the pi 500 which is not what most people want.
Ri 5 is overpriced enough as it is. There are now many more practical alternatives.
RPi is just another company now.
Pretty crappy how they left out the m.2 slot. It literally costs less than a dollar in bulk prices to add. Honestly poe wasn’t necessarily imho. Just nvme would be enough
I’d go the other way entirely and say PoE is much more useful for what the thing is myself – between netboot and the easy to exchange SD cards you have your flexible hackable tinker platform that can easily be deployed at scale for the conference/classroom nailed. Where a tricky to get at slow to change M.2 slot…
But both would be very nice of course, I just don’t see personally as many uses for the M.2 when its so awkward to get at, and retooling the case to fix that would add, probably a fair bit to the costs, that would probably end up reflected in the sale price. Pi products are expensive enough already, not overpriced by any means considering the quality of support and long term availability etc, I just don’t see it being worth it to the Pi folks to populate those features in the 500. But as an option on the PCB worth developing at least – as that doesn’t cost much and then for the industrial or education markets as a more special order or in a more ‘pro’ version with matching price tag… Plus for the few folks that desperately want a fairly cheap thin keyboard with a built in computer and M.2 or PoE they must know the community would figure it out and be able to do add the right components back in for themselves.
There are three good reasons they didn’t populate this:
It would add EMI and might make shielding harder or more expensive.
It would increase the BOM cost and therefore the price.
It would require retooling the case to make it possible for users to swap out the drive.
A reasoned list on a Hackaday Thread. I salute you!!
We are awash in a sea useful possibilities. RPis were not the first but they have been the standard bearers and, in their way, have helped grow a ton of users and projects that might not have come to be.
But sausage making is messy and the vocality of nay saying goat throated gatekeepers have to bleat the same old chorus. “why didn’t they do t EXACTLY the way I want it done”
Feh
Rock on RPi and all the efforts like it.
It’s kind of obvious. The board resides in a case which was not designed to be readily opened.
The reason given by Rasp Pi is credible. They intend to use this board in other devices so have designed in the capability to add the storage.
I expect to see a product in the near future that takes advantage of this, eg laptop with removable drive.
It’s forward thinking, not regressive money grubbing miserliness.
The USP of Rasp Pi400 or 500 for me is with a stack of SD cards and one Rasp Pi I can have a dozen different computers simply by swapping out the card. There are other ways of doing this, but this way is simple, cheap, and fun.
For anything more serious, I’d want something with a better keyboard. If only Rasp PI had a product like that…
(Yeah, I know, Rasp Pi5 with a HAT. Got one :) )
It’s not a price thing. It’s a “this device is made for children and people kept complaining that companies were hogging all the other pis” thing.
the m2 connector is on the wrong side, only 1 lane & the trapdoor is missing.
if made for childen is the argument, it would be better with:
1) a trapdoor with a screw (for parrent upgradable storage), not easy to swallow micro-sdcards
2) a cartridge type connector for an nvme expansion cartridge (in a case/shield)
3) headset jack + tv/rf/dvb output and fullsize hdmi, would not need special new cables
4) drop the gpio port and sell some working pico boards/electronic kits
there is a difference in shorting gpio pins on a 35$ vs a 90$ computer or an 5$ microcontroller.
if made for companies/coporations: change to full-size keyboard w/two nvme and two full size hdmi.
if made for hackers: add everything and the kitchen sink .. or at least connect all soc pins to pads.
Just because it doesn’t make sense to us, doesn’t mean it doesn’t make sense to somebody.
Most of you miss the point entirely :-) The real reason is that, after educating us on computing, the guys at Raspberry will have us educated on hacking and soldering of tiny parts. I find the project challenging and very interesting for the young audience.