Some of you may have heard of the new upcoming GPD G1 eGPU that can be connected with OCuLink. Thin, sleek, totally doesn’t look like a 2nd PC.
As of now, OCuLink provides a much better performance for external GPUs. Yes, Framework’s mainboards don’t have OCuLink to begin with. That’s where the NVMe to OCuLink adapter comes in. I was kinda bumbed that, from what I see, the mainboard has only 1 NVMe slot. Could’ve been so much easier to have 2 slots, one for storage and the other for OCuLink. Guess I could just use an NVMe enclosure for storage instead.
My point is, with the new GPD G1, you can get an extremely portable gaming setup going on ETA Prime has videos pairing it with mini PCs and they absolutely rock. And DIY Perk’s video using the Framework mainboard to create a 3-screen portable beast shows the amount of things you can do with it.
Can’t wait for the future of Framework! Also, we want 2 NVMe slots, please!! And if I weren’t asking for too much, an OCuLink port too haha!
You didn’t mention, but I have to presume you are talking exclusively about the Framework-13. Since the Framework-16 has 2 NVMe slots and also exposes PCIe with its expansion bay, making OCuLink easy via an expansion bay adapter, which at least two people are working on seperately.
Oh, I was speaking about only the mainboard without the other components. Can’t seem to find a mainboard only that has 2 NVMe(s), or am I not looking hard enough… and the catalogue filter under “Framework Laptop 16 - AMD Ryzen™ 7040 Series” says 0.
The goal is to 3D print an enclosure that accommodates an OCuLink adapter through NVMe. Think of it as a mini-pc to eGPU to external monitor
The Framework Laptop 16 is upcoming, now available under preorder. Currently, expected to ship late Q4 2023. Production will initially go towards full laptops, with separate mainboards being made available sometime later. It has 2 NVMe slots and an expansion bay connector that has PCIe, as well as power and USB 2.0. If you can wait I’d suggest using it instead.
Still only 64gbit for the pcie link (vs 32gbit on tb4), the 120 is for the non symmetric mode (120 in one direction and 40 in the other instead of 80/80), probably mostly for displays. The pcie4x8 of the framework 16 is still quite a bit faster than that and the occulink solutions are going to be a lot cheaper but well no hotplug.
OCuLink is not necessarily incapable of hotplug. Was there ever a confirmation that the Framework 16 does not support PCIe hotplug on the expansion bay slot?
Occulink with pcie hotplug would be a holy grail on the 16 but unfortunately egpus are a pretty small niche and occulink ones are an even smaller niche withing that so unfortunately I doubt framework will commit the engineering resources to do that anytime soon, especially since they outsource the bios, would be awesome if they did though.
PCIE is not hotplug unless it explicitly is, it needs special consideration (in hardware and software, some of the hardware side for the 16 could probably be handled on the expansion card though) and is relatively rare even in servers these days (especially since the shift from replacing parts in running servers to replacing whole servers at a time).
Also, I just want an extra nvme slot for the oculink converter, no hotplugs. My whole plan (if I were to go for it) is to just 3D print my own enclosure for the motherboard, 1 nvme slot for storage and the other for oculink. It’d be a fun little project.