Depends. There have been some leaks indicating that AMD might move to exclusively LPDDR5x for laptop CPUs in the near future.
LPDDR5x is around a third higher bandwidth (great for iGPU performance) and more power efficient (good for battery life), but muchore susceptible to signal degradation caused by SODIMM slots. So LPDDR5x is exclusively soldered or CAMM.
If those leaks are true then I wouldn’t be surprised if Framework moves to CAMM.
This, of course, always assumes the job the software is doing is amenable to having multiple threads. I occasionally use an electronic Spice simulation program, which although enabled to use many threads, will often run only one or two threads because the simulation cannot benefit from any more threads. From your description of use, it sounds like your process is similar.
Yeah the software has improved slightly over the past year but I have stopped paying for updates currently as they kind of soured the whole thing.
It is poor development. If the company tells you to get the best from their software you need a 3090 and a 13900K whatever and then it just uses 20% of both…you know something is wrong.
It’s not like 4 core+ CPUs have only been around 18 months or so. I cannot imagine anyone deciding to build AI based image software from 2019 onwards that can only address 2 CPU cores.
But as mentioned, things are improving. 
But can whatever process this software is for actually break down into processes that can be split across multiple threads in a program? Often the real world process is too linear to be able to do this, it doesn’t have things that happen in parallel without relying on each other, which is what is needed to be able to multithread an operation.
They may have ‘required’ a certain processor because that is what the compiler that has the features they want runs on. It may actually have nothing to do with threading capability.
Sure. 
Exciting! Having just purchased a 7040 mainboard, hopefully the next time I upgrade my FW it will utilize this CAMM2 standard. DIMM has been the mainstay since my very first build so such a change would be pretty unique.
I mean for the Framework 16 they are the first to adapt USB-EPR, so maybe they’ll be the first to adapt thist too?
This might be a bit harder since it relies on the modules being available too and those are a tad more complicated than a power supply.
I do hope they aren’t the last ones on the train though XD.
FW16 laptop uses SODIMM DDR5. It would need a new design motherboard to use CAMM.
The first one does yes. Don’t think converting the current design to camm was the ask here.
Well obviously, this wasn’t a “can we have it on this version?”, just a “they got to the 48V pd party first so can they be early to the camm one too?”
Camm is more of a chicken and egg problem though, for the 48v pd they designed their own psu and it still works with the old 20V PD ones too, for camm you do need camm modules and framework likely won’t roll those themselves. So in that regard it would probably be better for one of the higher volume players to move first.
For this reason I would expect it to be a couple of years at least before FW offer CAMM in their products.
It will take a desire to offer a new processor (why would you make a module with an existing processor where they already have working SODIMMs) and they have indicated they won’t be doing so with the recently announced Intel processor because that is just the same as the existing one but with chrome hubcaps.
I don’t see a new product as being a starter to offer CAMM unless there was some really serious performance advantage, or there is some really serious power consumption or physical size advantage.
So unless there is a new processor with super duper features coming I don’t see a case for development yet. With their small team they have enough on their hands getting the existing product lines out the door. I suspect that a proportion of the development team will be doing support for some of the more esoteric problems seen before they can embark on what would be a major new development, even if it “just” a motherboard to go in an existing product.
lpddr
lpddr again XD
That on the other hand is definitely an issue
https://investors.micron.com/news-releases/news-release-details/micron-first-market-lpddr5x-based-lpcamm2-memory-transforming
Latest update, looks like Micron has come out with a product.
TL;DR:
Goes up to 9600 Mbps instead of SODIMMS 5600 Mbps (Why can’t the industry just use MT/s, Mbps implies an incorrect speed for the entire module)
Reportedly 80% better idle power (aparrently due to the use of LPDDR5X ICs?)
My take: This solidifies the theoretical benefits of CAMM, and I think it’d be a good idea to move to this new standard in maybe a few years or sooner!
I’m not to sure what the difference is between LPCAMM2, and CAMM, are there already different form factors??
Let’s freaking gooo
Maybe not quite a fair comparison, you are comparing current sodimm with the projected 2026 lpcamm capabilities.
Not sure this solodifies anything we didn’t allready know, lpddr isn’t new but I definitely agree lpcamm should be used as soon as there is a reasonable guarantee that modules will be available from more than one source.
From what I read it’s the same connector and form-factor (the form factor is a bit freerer than sodimm iirc) but keyed so you can only plug in the right kind. I really hope it turns out that it is possible to support both in the same device but if it isn’t going with the lp version is almost a no-brainier (apart from the not being able to reuse existing sodimms issue).
