That’s only a x4 link, not an x8 link like the one from Josh. If I’m going for this, I want to go all the way and have maximum bandwidth available.
Although it’s nice to hear it works well. Although you’re running Windows? Probably runs well for Linux, but less certain.
But currently, just theoretical to me.
The expansion bay by itself has nothing, it’s empty. And the non-GPU expansion bay module just has the fans.
Last I recall, Josh was looking into whether it would be necessary to add a retimer.
Nope. Framework made things easy. There is a back panel that is removable. Josh said he plans to have the ports fit through there & include a 3d printed replacement back panel with cut-outs for the ports.
See here for what one needs, on the FWL16 side, depending on what one already has: OcuLink Expansion Bay Module - #431 by MJ1
Nice, thanks.
Although I wonder for hotplugging if the retimer is needed, there was some mention that it might be needed in some cases. I dunno though, I’ll go by what Josh things, at least initially.
I do wonder…can he just release those 3D models so we can print them with our own filament in our own colors/whatever?
And I presume it’s the standard SF-whatever cables for the 8i link that he’s using? I figure it will be.
Someone please confirm, or correct me
For the connectors, standard OcuLink.
I recall something about maybe two versions, I think it was one version with an 2280 SSD slot and a 4x OcuLink, the other with 8x OcuLink.
I wouldn’t count on hotplugging, at least initially. We don’t know if the current FWL16 EC firmware supports that.
~edit~
But the 8x, might be in the form of two 4x connectors so that people can more easily use either 4x or 8x. There are dual 4x to 8x cables.
I’ll have the models uploaded alongside all of the other design files for it.
Correct on the versions, but with the 8x link, it would normally be splittable, but due to FW16 limitations, it will only act as a 8x link, that is not splittable.
Heh, that’s an interesting thought, although I think most of us don’t particularly want to split it up.
However, at the destination board, does the FW bus support PCIe bifurcation? So if there’s bifurcation hardware on the end of the 8i cable, could that be done?
I was able to build a project on my own with each component costing no more than $50. I had fun doing it.
It’s not quite the difference you describe. The UT3G is converting USB4 to PCIe. My setup is doing the same. Oculink is essentially just cabled PCIe.
If I get the OcuLink bay for the Framework 16, I can still use my setup and have a USB4 NVMe enclosure.
nice to see you again josh, is everything ok? hows the rest?
I’m also using a usb4 egpu with the ASM2464. I noticed that you were able to flash a different firmware that allowed for higher pcie bandwidth. How did you do that? I can’t seem to find the details on that process.
For further Info write a PN its offtopic in here…
With that Tool you can flash another Software (they are provided with the Tool) and you can check the current Link Status. If its TBT3 its in 22Gbps Max mode, only with PCIE Tunnel its running with the MAx 40gbps. You don’t need to change the Firmware, the stock one is the most compatible.
No rush or anything, I know personal time and rest is important, just curious if there are any new update?
How can I check the link status and speed in Windows 11?
I linked the Tool for the asm264 based usb4 Adapters with screenshot, just run it.
But its not an oculink topic.
It would be cool to have a OCuLink i8, that could be switched to an OCuLink i4, because using a 2x i4 to i8 OCuLink cable would be annoying, but I do understand that it would be difficult.
I think there’s a huge number of “cool” things that could be done for this project, but it’s a matter of how much time Josh has to implement things. This is the first version, with a a 4x and an 8x option, I’m thinking about getting both honestly. I’d love to see it with an 8x PCIE socket internally and a card you could plug in to get a variety of M.2/Oculink/etc ports, but for now a nice working version will be great, without gold plating the entire thing, even though that might be cool.
Any ideas on what OcuLink eGPU docks are available to use with the 8i version of the expansion bay once Josh produces them? He mentioned above that the 8 lanes can’t be split, which I take to mean that regular 4i docks can’t be used with the 8i bay. I looked far and wide but have yet to find any available in the U.S. Without a compatible dock, it’s hard to pull the trigger on the 8i bay, even though I have no need for the extra M.2 slot on the 4i bay and just want full eGPU performance.
I personally bought this OCuLink 8i dock (Previously posted by Aaron_Baff) with this OCuLink 8i cable and tested it using a PCIe to OCuLink 8i adapter on my PC. It worked as expected on my PC: As if it was plugged in to the PC normally.
However, I have not yet found a good enclosure/case for it to put the eGPU dock in.
Hopefully someone knows of a good enclosure for this eGPU dock.
I’m thinking of something like this made of 2020 extrusions (looks like) that’s adjustable for all the poorly inconsistent Chinese clone boards as everyone copies each other.
If nothing else it’s adjustable and you can build off it yourself with some extrusions from amazon and a hacksaw.
@bleuthoot, Question, are you using the .5M or 1M cable? I’m curious if it works with a longer 1M cable ok or really need to keep it short?
Anyone using longer or any better cables that can be gotten to do so (ie. shielded)?
I understand electrical path crosstalk and things like retimers, but just wonder realistically how far you can push that distance of these components before you start seeing signaling problems or just dropping for lack of proper modulation on the bus with the random kit of the week from alibaba.
