Apart from the height issue mentioned, soldered chips are a little speedier as all components are more physically close.
The issue is ‘is it worth it’ I can’t ever say I’ve had a CPU failure in 30 years since my Sinclair ZX Spectrum 40 years ago. I’m pretty sure any failure I’ve had has been another component.
But yes having all components DIY seems like fun, so way back 60 years ago every resistor was a DIY part. Every switch (vacuum tube) was a plug and play, now there are billions of switches in CPUs and RAM etc.
The move towards using less space, and less energy and getting quicker response is the mainstay of electronics, except for the DIY diehard who wants to be hands on.
So thinking backwards to old style individual components is not overall popular, except for movable storage. Framework have taken a step towards modularity, but . . . well who knows
@amoun thanks for filling me in, I was not aware that there was a height constraint issue that would prevent framework from doing this sort of thing. I suppose the chance of a chip other than the CPU failing is also very low, so I guess it would not be worth it to not make the CPU removable. Thanks again for the info!
When I first trained for this job, one of the things I was repeatedly told was that the CPU would almost certainly be the last component to be at fault. For the most part that was true - I very rarely replaced processors, and only after… well, pretty much everything else. In the days before soldered CPUs that meant laptops didn’t usually get new ones either.
A few years back, it seemed like I was replacing about one or two a month - an unheard of frequency IME - and always with a similar pattern: multiple bluescreens with random titles, successful diagnostic test results, and no other explanation. The mainboard, RAM, drive, OS - everything - would be replaced to no avail. But pop in a new proc and suddenly everything was fine.
It’s since died down, although since nearly all of the laptop motherboards I replace now have a soldered CPU, and occasionally that pattern pops up, I sometimes wonder…
I created (completely in my head - not for human consumption) a conspiracy theory to explain it:
To counter the CPU hardware bugs discovered back then (Spectre and Meltdown, I believe?), the industry conspired to send out BIOS updates that would fatally fault the processor, forcing an upgrade - one way or another (for example, on-site repair).
Just something to keep my mind occupied while I ran the diagnostic after replacing the processor I was told I’d never have to replace…
Just a note, there is no requirement for posting ideas here, this is a user forum, so it is totally okay to post ideas like these that you have. I agree that it seems like a good concept, but others have brought up concerns about height, and I will also add that Intel (and now AMD) do help Framework design the mainboards, and likely have guidelines that will not permit them to put processors on separate boards, as they don’t want lost performance from that to seem like a fault of the chips themselves (this is speculation though)
I’d personally love to see an idea like this since I also want an easier way to replace the CPU and also a potential for adding a CM4 powered Framework, but it’s also important to remember that Framework likely thought of a lot of things like this on their own, and that some/all of the constraints mentioned could be the reason why they weren’t able to implement it.
Removable CPUs aren’t unprecedented in laptops, hell I saw a video on one where the guy upgraded the GPU, CPU, RAM, and Storage in one go. Even managed to get SLI over Mini-PCIE Modules (MXM dGPU’s) Framework 16 comes closer in that respect.
But I would really love it if the CPU Socket/Chipset were on a separate daughterboard. More so if the CPU itself was replaceable. At the minor expense of a bit chonkier laptop due to adjusting how the cooler mounts on, or even minor & theoretical performance differences.
If we’re keeping the rest of the board separate, let’s say someone wanted a 10th Gen Intel Framework for whatever reason… Then they could downgrade the UEFI, swap out the 11th Gen Daughter-board and try it out as a troubleshooting test or experimental project. Similarly if they wanted to try a different CPU in the same family/socket or dare I say an upgrade that supports the same socket.
Now ofcourse I know this won’t mean we can suddenly change an Intel board to an AMD one due to universal firmware not really being a thing here, but I can see interesting use cases for it already. The raspberry pi of laptops
Exactly! I am glad to see there are other people that are on board with this idea! Maybe, if it gets enough traction here, it might be noticed by the framework team, and potentially implemented in a new updated version of the mainboard! Also, this dream may be getting closer to reality as we speak! In a not-so-recent-as-of-recently video from Linus Tech Tips, where they were reviewing a new mini-PC from a company called: minisforum, I saw Jake put a laptop CPU into a real-freakin-small socket in the devices motherboard! So, these manufacturers might already be bringing socketed mobile chips back, which is amazing!
The framework 13 mainboard already kind of is a cpu daughter-board, you can’t strip it down a lot more and only at the cost of a lot more complexity which for most cases is not worth it.
Would be neat if they could expose some of the unused io on the cpu though but I can also understand why they didn’t do that.
I remember the old Clevo desktop replacement laptops could drop in a lower TDP desktop CPU in. If my memory serves, the earlier generations there were socketable laptop CPUs too.
Maybe the 16 inch could potentially have enough Z height but that would really make it thick/bulky (which is not the trend nowadays).
Another issue would be the pinout of different generations (see Intel) is probably different, so might break compatibility too, else the R&D process would really take a long time.
Being able to use a CM4 would be very interesting.
There are multiple different CM4-compatible modules available, from rpi, rockchip, nxp as well as riscv and fpga options.
My idea would not be to replace the internal CPU, but being able to use a different architecture with the same screen and keyboard and seamless switching between them.
This would require a HDMI to display port IC some means to expose USB devices to the CM4 socket.
If USB and HDMI are exposed as well on ports on the back, this daughter board could also be used as an external monitor/keyboard for servers or anything else.
I was searching for a laptop stand for my framework 16 (no, I haven’t received yet, I’m batch 15), and though: what about retractable foots as an expansion bay?
Like many keyboards have small foldable “foots”, I think of “long” (~20cm) legs, one for both sides, allowing the laptop to “stand” with the screen nicely raised at eye level.
I’ve considered designing legs that could be installed in the expansion card slots but I can’t think of any good designs that won’t be too wide when folded away yet stable enough when extended. I think it’s more than possible, though.
Maybe they could be “folded” under the laptop, hidden near the small riser foot already installed on the underside of the dGPU module? This way it won’t stick out more than the module already does.
Ok, this might be incredibly stupid, but hear me out…
A small i3 or Pentium computer in the expansion bay that can be controlled with a built-in KVM. It could be a little development environment for testing apps, kernel drivers, researching malware, or whatever you want. Would a VM probably work better?? Yeah, but that’s not as cool.
On a serious note, it could be very useful for developers working on risc5 support because all the risk5 boards available aren’t bigger than an RPI anyway. Just slap one in there with a KVM and maybe some debugging hardware. It would provide a way to run code on risc5 hardware directly while leveraging your normal laptop to run your heavy development environment and other apps (Spotify, Google, Discord, Teams, etc). Then use the debugging hardware to flash and run the code on the risc5 board when you want to test.
Make your own RISC-V board with the new PIC64 chip from Microchip. Contains a quad core RISC-V processor plus a superviser core (If I am reading the info correctly). Has multiple PCIe channels and uses external DDR4 memory.
It was more of a fun idea. If something comes of it, great! If not, I wasn’t expecting anyone to design this for me. I am going into my second year of college studying to be an Embedded Engineer, so while I don’t have the knowledge to design this idea now, I totally would be down to designing it sometime after I graduate. I think it would be really fun and it’s the type of thing I want to be able to do for a living.
Then they could downgrade the UEFI, swap out the 11th Gen Daughter-board and try it out as a troubleshooting test or experimental project. Similarly if they wanted to try a different CPU in the same family/socket or dare I say an upgrade that supports the same socket.