There is still a massive cost difference so if it actually catches on (that’s a big if) I can see occulink or something similar sticking around, at least for lower cost gaming focused systems, the re-drivers for at least the early gen tb5 interfaces are going to be brutally expensive so even if te interface part came for free with all the cpus (which is the case wit most intel and amd laptop cpus and usb4 right now) a lot of manufacturers would still leave it out in all but the high end options.
@Adrian_Joachim No worries. And thank you for the clarification, my “process” was to think EGPU.IO is generally trusted (by me) and that the page was very recently updated (which I took as “continuously fact checked” as well, so here we are). In any case, thanks for the background clarification.
These chats have helped me put the puzzle together on questions I had for a long time. I was some years ago really impressed with Razer engineering, and now framework took this place for me.
Given that the Framework CEO himself says this should work somehow and that he is a bit active apparently in some of these chats is a sign that really gives me hope for what’s to come. And of course many others here have similar plans.
I now believe I am not asking too much from this notebook, so hopes are up. Let’s see!
Everything before and after that blurb looked solid even in that article, that one sentence almost gives you whiplash with how nonsensical it is. Still a great resource but nobody is perfect.
Definitely.
Kinda bummed they didn’t break out a few spare pcie lanes on the 13 but I understand why.
1 Like
Yes, and that is why I said 1 or 2 years.
USB 4 version 2 (why not usb 5?..) has already been announced and it looks like the rumours said Thunderbolt 5 would be, just with optional parts (exactly like usb 4 is to TB4). So TB5 announcement has to be around the corner.
But even if Intel announces TB5 in a few months it might take up to a year to be seen in computers and then it will take even longer to get TB5/USB4v2 hubs, cables, eGPUs…
So yeah, Oculink has some time to fill the gap. But after TB5 is popularized I suspect Oculink will disappear again from commercial electronics.
@Name2
Nah I think my argument still stands.
If occulink was to catch on for the egpu use case enough that at least some manufactureres actually have built in occulink ports it will continue being a thing even after usb 4.2(or whatever) or tb5 comes out because of the cost difference.
Just look at all the amd 6000 series laptop out there and check how many of them have usb4, it’s not a large proportion even if the chips come with most of the required parts already included. All you need is a re-driver and hooking it up properly. Looking at how bonkers the signaling rates for tb5 will have to be those re-drivers are going to be a lot more expensive than the already expensive usb4 ones.
Just hooking up some pcie lanes from the cpu to an occulink connector is not free but a lot cheaper and even if you wanted to do it nicely and add a (much cheaper but still kinda expensive) pcie re-driver to the port (which is not required but helps, especially with longer cables) it will still be a lot cheaper to implement.
If however the egpu use-case remains a very small niche there is a good chance it’ll indeed “disappear” again once tb5 shows up. By disappear I mostly refer to return to the state it is now with a few small entrepreneurs using it.
The coolest outcome would be if pci sig actually did design a standard for consumer external pcie (with authentication and maybe even power and hotplug) but I highly doubt that. PCI sig is very much server first and trickle down from there.
1 Like
This is exactly the reason for my plan forward.
Shared on Reddit:
That is the rated durability of an off the shelf connector from the same supplier. Two notes on this:
- We’re developing our own semi-custom connector with the supplier specifically to make it better for end-user handling.
- The cycle life in datasheets is rarely comprehensive. We’ve had instances where the datasheet on a connector (I think it was a pogo connector) said 100 cycles, and we asked the supplier to retest to 2000 cycles and found the connector passed that too.
25 Likes
Will the connector be available for purchase in the marketplace as a stand-alone piece or is it built into the mainboard?
1 Like
The only connector is a flex cable with some pins to interface between motherboard and expansion bay, it also shouldn’t wear out.
1 Like
Pre-order already done ;). If the expansion bay even has a good connector durability, man, this notebook is getting flawless.
2 Likes
@nrp Thank you for the update on this, truly love having this information available and what you all are doing for the tech community!
3 Likes
Is there any way we could get a statement on if frequently changing expansion bays is a possibility? It seems implied in the configuration page when pre-ordering the laptop that it’s possible:
But it may put some of our minds at ease if there was an official statement on if it’s safe to be changing these frequently - I’m personally getting the 16 to use in university and was considering getting both in order to change for portability when taking the laptop to class, but it’d be nice to know whether or not changing the expansion bay multiple times a week is something I can do without risk of damage.
@Owen The Expansion Bay System is on the list of planned deep dives. We may get more information on swapping then. Past deep dives can be found here frame.work/blog/category/hardware
People have asked for more information on Expansion bay swapping (as well as other things), but it does not seem Framework is ready to provide more at this time. The deep dives have been how they are releasing more information.
@MJ1 I didn’t think about the deep dives! The ones we’ve gotten so far have been very detailed, so I expect we’ll get some really nice information in that one. Hopefully we get all of our questions answered!
I would honestly not be surprised if the deep dive on the expansion bay comes last. The good sign here would be that it shows FW is aware how essential it is for the entire product. My feeling is the FW 16 is a bit like a formula 1 race car being still tested, discussed and checked by busy engineers/suppliers/team etc. until shortly before putting it on the track ;).
1 Like
Because they are likely not completely finalised by the fact that there has been GitHub commits very recently in the Expansion bay repo
1 Like
There’s also a pending deep dive labeled “Connectors”. 
3 Likes
Somewhat dead post, but the connectors deep dive post does talk about the difference between the FXBeam connector and the connector Framework is using.
We then worked closely with Neoconix to build our own customized version of FXBeam (on the right in the image) that is compatible with the same Mainboard interface, but is substantially more robust.
However, Framework fails to mention the rated cycle life of these connectors, only stating that “the interposer is safe for handling and repeated cycling”. Whereas on the input module, they specified “They are rated to 10,000 cycles and are difficult to damage even if you try to.” So it is probably safe to assume they either do not have a good estimate of the cycle life of the custom FXBeam or it is not a number that is high enough that they would like to share it publicly? At least not as impressive as “10,000” certainly. Still would be nice if they could provide more information on that.
Framework might not yet have final numbers, but have enough data to tell that it’s good for how they intend for it to be used. Understandably, they may not want to release preliminary numbers, as people may repeat it and fail to note that it’s preliminary, that the true cycle numbers might be much higher.
1 Like
did framework release a cycle number for the connector yet?
1 Like