For a while now, I think about creating a carrier board for a CM5 module.
Now, since it is out and know the specs, we could maybe work towards a way to make it work.
Of course, not everything will work and it won’t be a simple weekend project.
Battery
Display
USB-C / Thunderbolt Ports
Audio Jack
Speakers
Input Cover
Finger Printreader
Keyboard
Trackpad
Wi-Fi
Bluetooth
Camera
Microphone
RAM (on board of CM5)
Storage (via PCIe)
Display
The Raspberry PI supports HDMI, maybe we can turn one of the HDMI signals to eDP using “LT6711” from Lontium Semiconductor.
Not sure the returns would be worth the effort. What are you getting from the RPI5 that you are not already getting in spades from the current available mainboards?
This is a regression in specs and that seems to be about it. You also won’t get any low power modes, like suspend / standby.
That said, have you heard of MNT? They are taking CM boards from all kinds of ARM manufacturers and putting them into carrier boards for their various product platforms.
Maybe that is something you want, but is already being done?
I’m not a gamer.
I know enough people who have a mindset of “why it would not work” instead “what can we do to make it work”. They don’t have joy in creating things, tinkering around and learning new skills.
The idea of Framework is that parts can be replaced by things that they don’t sell and we are here in the creator forum.
Buying a MNT notebook should not be the solution, especially not while Framework exists and we as DIY people are already invested in the eco system.
Might be worth chatting to FW or Deep computing to see how easy it would be.
I guess a stepping stone might be to duplicate all the USB PD, Battery and EC from a FW main board. That would give you the power on/off and charge functionality.
Then interface the EC to the CM5 and get the CM5 talking USB4 somehow or at least USB3 with DP/HDMI alt mode so it can use the FW Expansion slot cards.
Does the CM5 already have the equivalent of an EC or not?
From a marketing perspective, I don’t know whether there would be any demand for it though.
I really like the idea of this project. I think it could offer a lower cost and more power efficient (maybe) ARM version of the FW13 or even FW16 with far fewer features for those who like to tinker. It would also allow people to upgrade or even “side-grade” to RockChip based boards.
Like Deep Computing’s main board, I don’t think anyone outside the hobby/research community would buy it.
Saying all that, I would jump on this so fast.
Batch 1 CM5 carrier board LETS GOOOOOOO!
Last night in the #framework-discussion channel on Discord, we discussed the dimensions and whether the CM5 module would fit within the Framework case.
What we know:
The CM5, including its mating connectors, is 4.94mm thick.
There are two connectors for the compute modules (1.5mm and 4.0mm [mated])
The thickest part of the Framework mainboard is approx. 6.8mm.
A PCB typically measures 1.57mm in thickness.
We have to leave approx. 1.1mm clearance below the mainboard, as the standoffs dictate where the PCB can start.
We probably would have to rely on the case as a heatsink / no active cooling.
Based on these dimensions:
4.94mm (CM5 + connectors) + 1.57mm (carrier board) = 6.51mm, which is less than the 6.8mm limit of the Framework mainboard.
While this seems promising in theory, the carrier board would need to sit on top of the motherboard standoffs. This would result in a significant offset.
Option A
Instead of connecting the CM5 directly to carrier board, we connect it facing down to another PCB (let’s call it bridge) which then is connected to the carrier board.
For that, we’d need to reduce the PCB thickness to 0.8mm.
The chip in the pi5 isn’t all that efficient so I doubt it’ll beat the current offerings in pretty much any workload, and that is ignoring what hdmi to edp chipsets and stuff like that would also pull.
Something with an rk3588 may be a better option for that but the software support for that sucks a lot more than anything raspberry and still won’t beat anything else framework has so far in power consumption on anything you’d remotely do on a laptop.
On the other hand, the current mainboards are definitely not cheap. If you could get a CM5 on a framework-sized mainboard, for a lot of people that would already be enough power for a laptop.
On that thought..
Hell, you could even have a dedicated expansion card with just feedthrough GPIO!
Although you are limited to using USB-C from the mainboard side I guess, as the FW shell is prepared for that… So yes, technically, you can repurpose all the USB-C-connector connections to GPIO signals, but that might also be asking for trouble if you put in the wrong expansion card…
That proposal of mine reminds me a bit of this article.
You would need an (empty) FW laptop shell though to repurpose it, otherwise it doesn’t make sense anymore and I think empty shells are harder to come by than mainboards, as these get replaced more often.
An Idea, if you used a FW16, and put the CM5 in place of the dGPU, it might have more height clearance and also room for a heat pipe cooling the CM5 cpu.
For one the mainboard to make it work in the framework is likely going to be more expensive than the cm5 itself and the the other it’s going to get blown out of the water in pretty much all aspects by even the lowest end 11th gen board.
Honestly not sure about that, especially since you can get much more usable stuff cheaper.
You can get full blown laptops with n100 chips for like 200 bucks most places which makes for a much more usable platform.
The framework platform itself is just not that cheap, that’s like trying to put a lawnmower engine in a lambo, it’s still going to be too expensive for the common man and just worse in general but can still be a fun project.
Raspberry Pi CM4 or CM5 carrier boards doesn’t cost that much.
The followings are two examples of CM4 and CM5 carrier boards I’ve designed.
For prototyping, I used JLCPCB and it was ~$600 for 5pcs, including shipping.
It’s equal to $120 per board.
But for 100+ bulk PCBA, the unit cost drops down to $35 ~ $50.
I think if we make a CM5 motherboard for Framework, the price would be similar.
It’s definitely more expensive than CM5 itself in terms of 2pcs or 5pcs prototyping.
But if someone suessfully create it and resell after bulk PCBA, it can be cheaper or similar priced to Raspberry Pi CM5.
You don’t think adding hdmi to dp converters, chage/battery controllers and pd controllers and maybe even an ec and stuff would make it more expensive, especially in low volume?
Right now it just looks like you’d be paying slightly less to about the same as getting an 11th gen board (or a whole ass n100 or something based low-end laptop) for something significantly worse but it would be pretty cool.
This project isn’t about price or performance – there are plenty of other options for that.
If you want the most bang for your buck, you might want to consider a used notebook from eBay
It’s about exploring what’s possible and seeing how far we can push the Framework ecosystem.
I want to see if we can make the CM5 work in a Framework laptop, identify the roadblocks, and figure out how to overcome them.
The Framework platform already gives us a head start with its modularity and standardized dimensions.
For those who share the excitement of building, experimenting, and pushing boundaries, I’d love to collaborate and hear your ideas.
Let’s turn this idea into reality and have some fun along the way!
On another note, any ideas how we can source some of the mainboard connectors?
On a typical Framework mainboard there are:
Name
Part
Comment
USB Type-C
S CONN LOTES AUSB0534-P003A 24P 3.41 USB TYPE_C P0.25
I really like this idea. The lack of suspend/sleep would be almost a dealbreaker for me, but I would still probably get one if something like this hit the market.
As long as the positioning is right any usb-c connector should do right?
Worst case you could do your own audio board with whatever connector you can source more easily. Might even be able to put the codec straight on there which would make it arguably better than the regular mainboards XD.
If it turns out to be necessary something good enough can likely be easily 3d printed.
Pretty sure any low profile enough switch should do just fine, hell 2 jumper holes would probably even do (you can trigger those with tweezers).
Suspend to idle could probably be made to work 2-3W isn’t order of magnitude more than a current intel or amd framework draws during sleep.
Just had this exact thought this morning. Looking at the CM5 datasheet you could probably drive the edp display with a SN65DSI86 MIPI DSI to eDP placed somewhere on the mainboard. The thought would be creating a carrier board in the same footprint of the framework mainboard. If height would allow it the CM5 connectors to be placed directly on the board.
