Oculink is a connector, that is used by some parties but it’s just one option and I hope if there is an official one it would be better.
1 Like
There are connectors that connect to m2 slot (including 2230 sized slots) and expose an Oculink connector at the end of a cable. In the Framework 16, if they don’t explicitly block the ability to route the Oculink cable from the m2 2230 slot to the expansion bays, it should be “easy” to have an external Oculink header possible.
I hope this is possible in the 16, as it dramatically extends the life of the hardware, allowing for async piecemeal upgrades, which fits the ethos of Framework.
@Matt_Hargett yes but in the 16 it would be even easier to have an oculink port in the back, in the expansion bay that has 8x pcie.
The issue is how to get it in the 13 and it does not look like there is an easy solution.
2 Likes
Based on the discussion of a potential Framework 16 Expansion Bay converting it’s PCIe x8 connector to Oculink, is there a way two ports on a Framework 13 (Specifically TB4 ports on the Intel platforms) can be linked together into a recognizable x8 connection to then adapt to Oculink (and whatever new products the 16 would also be able to use based on Oculink)?
my work laptop is a Dell Precision mobile workstation that uses a dual thunderbolt/Type-C connection to a dock (primarily for power since I got this before Type C got a boost to is maximum power delivery) so dual connectors have some kind of precedence, however basic.
Intel didn’t share many details, but the demo consisted of dual Thunderbolt lanes, with each running at 40 Gbps (for a total of 80 Gbps). That’s twice the performance of Thunderbolt 4, which tops out at 40 Gbps in aggregate.
TLDR: probably not
On the intel you got 4 tb4 ports but the left and the right side each share one 40gbit uplink so you would need to plug into both sides to get the full bandwith. On the amd just the 2 top ports are USB4 but also 40gbit per side so same issue.
Great now you have 2 pcie3x4 links (if you put expensive thunderbolt/usb4 chipsets on both and even then it’s not full bandwidth thanks to overhead). While there are ways to split and or share pcie connections I don’t know of a way to combine them so unless you want to run 2 gpus in sli or something that won’t be of much use.
But hey at least you don’t need to reboot your machine to unplug the gpu XD.
Got one of those too and I prefer that implementation over hp’s cursed usb-c+barrel jack connector but not by much.
1 Like
I’m late to the party on this thread, but I am planning to try this on the 7640u as soon as FW starts shipping to Korea. Hopefully that happens before I move back to the US in a year. As is well documented, TB3/USB4 looks ok on synthetic benchmarks, but severely limits gaming performance. Oculink is a better solution, but making it work on a Framework 13 would take some effort.
Anyway, here is what I’m thinking: There are pcie 4 x4 to oculink adapters available on aliexpress. The one I’m looking for has a flexible cable attached to the pcie connector. I plan to route that to the bottom left expansion slot and mount the oculink connector inside a hollowed out USB-A expansion module. I’ll use a kit like the Osmeta GK01 to mount my GPU and plug in the oculink cable to the port now exposed in the expansion module.
For my OS, I plan to use one of the storage expansion modules in the bottom right slot. For the game drive I can use an external USB4 40gb/s NVME enclosure. Hopefully I can find one that’s part of a dock so I can connect the drive, KB, Mouse, and ethernet with one cable in the top left port.
The advantages of this setup could be near full performance on an eGPU, no cables hanging out of your laptop, sufficiently fast Windows/Linux OS, and a fun DIY project. Also it would be completely reversable with no holes needing to be cut in the case. Possible cons include the external drive module disconnection issue (not sure if this is still a problem on AMD?), and possibly other unforeseen issues.
Once Framework starts shipping to Korea I’ll get all the parts and document my experience. In the meantime, maybe this can give you some ideas on how oculink on a FW13 might be feasable.
1 Like
where is this documented again. i’d like to see a spec diagram of ports, wiring and throughput.
Could I just ask what the benefit of this setup is over something like a mini-ITX build…? Like I understand the “fun” aspect of buying gear but at the end of the day isnt it just better to do a real build at this point?
1 Like
By this you mean what exactly?
Personal preference, I guess. My current thin and light laptop is 6 years old, and I won’t be able to bring my whole desktop PC in my upcoming move. So I’m in the market for something new and would like to have one device that’s thin and light enough to bring with me for work, but can become a good gaming pc when docked. Gaming laptops tend to be too big with poor battery life, and building a miniITX would require investing in a whole second machine. A setup like I’m planning would only require a GPU, PSU (which I already have), an external kit, the laptop and a bit of jiggerypokery. Besides that, I enjoy doing this kind of thing.
1 Like
I kinda wish for the same usage, zen 4 is pretty good, storage can be fixed.
Issue is getting from M/2 to oculink to an expansion card basically
1 Like
I’m sorry. I was on my phone and distracted by kids. 
I am looking into adding Oculink to a Framework 13 Intel 11th Gen mainboard, and I am trying to study the detailed specs, and diagrams for the Framework 13 series.
I remember a while back seeing few PDFs that detailed the bandwidth distribution, ports, and chip logical layout.
I’d love to know what the bandwidth, form factors and supported protocols are for the 4 USB-C ports, the wifi m.2, and the ssd m.2 nvme pcie port, and any other relevant ports that are available.
I think there is a display port for the mainboard to the laptop display, but on on other threads it has not been clear how to use that port for an hdmi or dp display.
I’m also not clear with the ssd m.2 pcie port what lengths it supports. Does it only support the 2280 length? or are shorter lengths supported?
I see on amazon that there are a number of short oculink m.2 adapters, but i wasn’t sure if it would work with the F13 mainboard, or if I needed an 80mm one.
If i can find enough specs, i might modify printable case, or design my own to work with oculink, and maybe a few other projects.
Pcie wise it’s simple, you got 4 lanes in the m.2 m-key slot for the ssd and 1 lane in the m.2 e-key slot for the wifi card. That’s it as far as exposed pcie goes.
Having the right length certainly makes mounting easier but anything shorter works too just needs a bit more creative mounting.
I can personally confirm the GPD Win Mini with the same chip as my Framework 13 (7840U) works fine over Oculink, would be cool to see the FW16 be able to add the connector.
Is this correct?
( * top is the fan blow out position, * bottom is towards the battery )
Intel Framework 13 - 11th Gen?
- SSD M.2 ==> NVME PCIe 4x (4.0 || 3.0?) - M Key Slocket
- WiFi M.2 ==> PCIe 1x (4.0 || 3.0? ) - E Key Socket
- USB-C (Top-Left) ==> USB 4.0 || Thunderbolt (v?)
- USB-C (Bot-Left) ==> USB 4.0 || Thunderbolt (v?)
- USB-C (Top-Right) ==> USB 4.0 || Thunderbolt (v?)
- USB-C (Bot-Left) ==> USB 4.0 || Thunderbolt (v?)
- Laptop Display port ==> ?
Not sure what pcie gen the 11th gen has, gotta look that up yourself.
the 11th gen already have tb4 though bandwidth wise it’s basically the same as tb3, keep in mind the ports on each side share a 40gbit link each and you can’t use the full bandwidth for pcie tunneling. Not that attaching occulink to thunderbolt would be particularly reasonable since you just get the drawbacks of both without any of the upsides.
It’s 4 lanes of edp but that really won’t help you with occulink.
Yeah but i fear the 16 to be too bulky, guess we can ask for a framework 14 ? 
Ideally we could do a kit like they did for the thinkbook 14+ / 16+, going directly from the existing pcie x4 to one of the expansion slot, and just have it there
I wonder how hard it could be since it’s essentially just direct pcie, traces in a pcb is doable
Edit : thunderbolt 5 is seemingly confirmed with arrow lake
Intel Barlow Bridge Thunderbolt 5 controller for Arrow Lake-S CPUs detailed, enabling PCIe Gen4x4 for external GPUs - VideoCardz.com
1 Like
It is not confirmed in Arrow Lake. It may come to the HX CPUs but the H will likely come with TB4.
Honest, this thread was created back on May 2023, if we had an Oculink bay connector baked to work with Framework support this would had my money. Even with TB5 released latency might be worst than with Oculink, Lenovo managed to get the hotplug to work with their TGX (Oculink) Thinkbook laptop. I don’t see why Framework couldn’t come with a solution like that and it would not incur in the patent cost that Thunderbolt has which transfer the cost to customers.
A small 13 factor laptop with Oculink is what some may be looking, not a super niche like many may think. Once people start realizing the befit to carry a light laptop around, sitting in front the TV, booking their vacations and then docking the laptop in an eGPU to play a demanding game with new gen GPUs in PCIe 4 or 5 they will admire this setup so much that won’t be a point of return. This would spread like fire.
The mainboard has at least 2 usb4 ports.
Maybe with some driver magic one could somehow join them to look like a pcie x8 lanes using the usb4 in pcie thunderbolt mode.
Kind of inverse bifurcation.