Upgradeability of framework combined with low cost of SOC's

Before I tell you my idea, I want to clarify some things. This is IN NO WAY intending to imply: One, that I am creating this myself, and two, that I am the absolute most knowledgeable in this regard. I am simply suggesting features that I would like to see in some future products or revisions of the framework laptop (both 13 and 16). With that out of the way, here is my idea! Processor companies like AMD have made custom chips for certain devices before, like with the steam deck. The problem with SOC’s is that they can’t be upgraded. Or can they? Ok, so hear me out. The base spec model of a revised framework laptop could have the components like CPU, RAM and even SSD, be integrated into a custom SOC. Then, if you wanted to upgrade your laptop, the NVMe M.2 slots, in addition to the SO-DIMM slots would be empty, since the base specs are inside the SOC. So, you could add more RAM, or a bigger drive, and on the 16-inch framework, you could still add dedicated graphics! The benefit of this is that by reducing the base model’s specs down to a single chip, the base model could be a lot cheaper. The best part is that such a revision would only change the mainboard, and not by much! So, these new, SOC mainboards could be put into your already existing machine, if you have one. Another benefit of this is that the existence of SOC mainboards doesn’t necessarily make the already existing mainboards obsolete. When you are configuring your laptop, if you don’t want to start with a top-of-the-line machine, but you think you might upgrade in the future, you could buy the SOC mainboard laptop to save money on the upfront device. Doing so also wouldn’t affect the cost of future upgrades in any way! On the other hand, if you want to start with higher specs, you could buy the standard chip mainboard, and just have its specs be higher! That is my idea for a new feature that framework could implement. If you have anything to add, let me know! Just know that I will probably take a long time to respond. Thanks!

1 Like

I suspect the main problem will be the RAM. I doubt it will be technically possible to have it on-package as well as in a socket.

I don’t know whether the cost savings of such a base model would be worth it in the end. But I would like to be proven wrong, since it is a nice idea.

It definitly is technically possible, there are µC with internal and external RAM. But framework does not have the volume to run custom silicon - so factoring in development costs and tooling each of these SoC would cost several 1000$.


Thanks for the replies! In regard to framework not being able to produce custom silicon, I would like to clarify that what I meant was that framework could make a deal with AMD or intel to have them produce these SOC’s similarly to how valve had AMD produce a custom chip to fit the needs of the steam deck.

I don’t know that it would save initial cost. You’d be spending the extra cost to get a CPU with integrated RAM and SSD, along with all the controllers for them, PLUS the controllers for external RAM and SSD along with the slots/sockets for everything. I think the main reason integrating it all into the SOC is cheaper (potentially) is because you get rid of the sockets and external controllers and support components for all that stuff. If you have to have it all integrated into the SOC and also have all the sockets and support for external stuff too, I don’t know that it would offer a savings initially.

I’m certainly no electronics engineer or designer of such systems. So take these opinions with the grain of salt they deserve as just the speculative thoughts of a guy on the internet, lol.

1 Like

Even a small amount of integrated ram and storage – e. g. 4GB / 64GB – would be a great feature, enabling the mainboard to be used as a standalone computer. Most people would still want to add their own to complement these, but in theory you wouldn’t have to, and you would have an obvious upgrade path later.

That said, the amount of space on a mainboard is limited, and if the team suddenly found itself with a little extra, I’m not sure that this is how I would prefer it to be used, instead of, say, another nvme or ram slot, or even a(n internal) microSD port.

Laptops do this already; as an example the System76 Lemur has 8GB integrated and a (by default empty) ddr5 slot. That’s why the options available are 8, 16, 24, 40.

1 Like

So it already exists in other laptop’s huh? That is amazing! As @danso said, 8GB would probably be a good amount to have in these SOC’s. It’s not just enough for basic web browsing, you could potentially play some minor games like Minecraft or Roblox with that!

Are you sure this isn’t simply LPDDR soldered to the board? I’ve certainly seen that before (with a SODIMM), but that’s not on an SOC.

More parts soldered on Mainboard is definitfly not the phiolosophy behind framework.

That’s really the worst idea i’ve ever heard of for framework laptop. It’s basically saying make the laptop less repairable.

Sorry man, super bad idea.

The mainboard from the 13" laptop can already be used as a standalone computer. You just have to leave the RAM and an SSD in the sockets on the mainboard. But it can be used standalone. It wouldn’t have to have those things integrated into the SOC or soldered to the mainboard to be used this way. Unless by “standalone” you mean with no components in the sockets, but I don’t know why that would be a problem. Framework even teamed up with Cooler Master to make a case specifically for using the FW13 mainboards as a standalone computer.

It’s definitely ddr – in particular it’s claimed to have four of these things at 2 gb each.

While that might not meet the technical definition of system-on-chip, would that technicality make any difference to the experience of using a Framework mainboard with some memory built-in?

From the context it really should be obvious to the reader that I was talking about using a Framework mainboard without adding any ram or storage devices.

Sure, I just don’t know what you’d really gain. I was just pointing out that since you can already use the current setup standalone, literally the only thing you’d gain is not having to populate the RAM and SSD sockets. So I was just curious if that was really your main reasoning, and if so, why. Just making conversation.

You’d have the extra expense of an SOC with integrated RAM and storage, but still have the extra space and cost required for the sockets as well. I don’t see the benefit vs. just putting RAM and SSD in the sockets that are there. But that’s just my two cents. Everyone has their own ideas of what would work best for them.

I imagine the SOC memory would be faster! But yeah, your point is a fair one, it’s effectively the same thing and I agree it suggests that two different RAM interfaces can be used, though I do wonder if soldered + module is the same as SOC + module. Is it really all the same?

1 Like

What you gain is not needing to have memory and storage devices.
Do you have any experience with single-board computers such as a Raspberry Pi or any of the ones sold by Pine64?
They’re fun, and useful in a variety of situations, and it’s not hard at all to imagine a single-board Framework mainboard with built-in memory and storage and also expandable slots.

But as I also said, I’m not sure that I would prefer it over a second nvme slot (for example).

Na, pass. If the onboard memory fails, the board fails. And if you’re pairing the memory SODIMMS having a soldered-on one sorta flys in the face of whatever the user wanted to install.

It’s really common among other OEMs, and those definitely wanted the user to buy the OEM memory modules only - and as spare parts they remain full retail price forever on sites like EMPR, from what I’ve dealt with at least. Anyone wanna pay $400 for a DDR3 2gb module today?

@BarriBurt , that has to be the single strongest argument I have seen against this. I like seeing the discussions that are going on! Though, I am still going to have to counter argue what @BarriBurt said with this: with how well engineered modern chips are, realistically, what are the chances of the onboard memory failing? Probably about the same as the CPU failing on the current designs. So, it’s not really gonna make that much of a difference in that regard. Still, though, it’s good to look at arguments for both sides. And as I said before, the existence of these proposed new mainboards doesn’t necessarily mean that they need to get rid of the old designs, so if this wasn’t your type of thing, you could still use the tried and true, designs. Although, to play devil’s advocate here, I could see why having an SOC combined with these slots, could actually make this design more expensive, so in the end, (at least for now) it seems like with what I have gathered so far, that this might not be as good an idea as I originally thought. Even despite what I have just stated, I would like to be proven wrong, so I am still open to see anything else you’d like to add. Thanks!

Maybe i am completely missing something here, but what’s the point of the proposed new mainboard design? The three arguments i could find here are

  1. lower cost, which has been counter-argued with higher development cost and probably higher cost for having soldered + socketed components
  2. being “easier” to use (especially stanalone outside the laptop), because… you don’t have to stick RAM and SSD in it?? With guides and countless tutorials and even an enclosure for the mainboard available I really don’t see how we would win anything here…
  3. upgradeability, which from my understanding wouldn’t be greater than what we currently have, because you can already put more RAM or storage in your Frameworks later on

I mean absolutely no disrespect to OP or anyone else, just trying to understand the idea.

I just discovered a technology that would make this concept actually superior to the previous design. UCIe, a way, for modular chiplets to be brought together to make your own SOCs. So, for instance, you could make an SOC with an intel CPU, AMD graphics, Qualcomm Wi-Fi, Google AI accelerator, etc. The complete chip could be changed on the fly by swapping out the chiplets. This would truly make this concept work, and in addition, it would mean that the SO-DIMM slots would no longer be neccessary, because you could just add a better RAM chiplet to your SOC. Same thing with storage. With this technology, you could also theoretically add more SOCs, for more power. This could be made as the new purpose of the Framework 16’s expansion bay. More SOCs. UCIe really seems to me like the new PCIe. Allowing you to add more chips for more performance, without needing to add an additional board, such as a graphics card. UCIe could not only make the framework laptop better, it could also allow for other laptop makers to make upgradeable designs with a lot less effort. Although, The framework laptop would still be superior with its modular IO. Also, think about this. With a modular SOC, comes the possibility of something truly great. Upgradeable smartphones! I firmly believe that UCIe, is the future, and a very bright future at that.

From what I understand, UCIe is not meant to be user modular, it uses very high bandwith interconnects, which basically nessecitates all Chiplets be interconnected on the same substrate?
UCIe is great for further expansion of Moore’s law and computer density though.

UCIe as you describe it is no different from current sockets such as PCIe and DIMM, it’s just that manufacturers have abandoned these standards due to the “need” for faster soldered I/O (with the obvious added benefit of sooner obselence).

Edit: Just wanted to say I appreciate the discussion going on here though, it’s these types of posts that I found were most educational to anyone reading/participating in them when I started learning :slight_smile: