Support for discrete cards (nvidia/AMD)

I know Intel GPU is good for everyday use with longer battery life. But, for somebody who likes coding, gaming, design and data mining etc want higher power to support their needs and internal gpu is not good. Is it possible for Framework to support GPU from nvidia/AMD?


I totally agree! I’m a student and I’ll be getting into University this year for computer engineering. A discrete graphics card would be great for all purposes!


iirc they said something about Thunderbolt 4 support so in theory eGPU support is a thing it could do.


Yeah egpu I think is an great combo. I’m going to college next year and I think this combined with an egpu will fulfill most people’s needs. I have an older gaming laptop and I rarely ever do hard workloads on it outside of my desk. I can’t imagine it’s too different for other people?


Absolutely a must for me for the Framework Laptop is TB4! The expandability brought with TB4 in terms of external graphics, networking and storage abilities (plus single cable docking) are something that would really help future proof my investment and allow me to really do everything I need to with my laptop by simply attaching things like an external graphics enclosure.


I did a quick check with a TBT2 eGPU I have and it is recognized. We can grab some newer GPU docks and test to confirm they work. But yes should work.


Remember that the expansion card interface is actually (AFAIK) USB-C. I don’t know what speed you need for an eGPU but the interface may be the limiting factor.


yes, tbt3/usb4 tops out at 32gbps, which is about the same as pcie gen 3 x4 - so will not be the same speed as a x16 slot on a desktop, but still plenty fast for a lot of use cases.


Good to know. Thanks @Kieran_Levin.


that is good for Framework want to Implement this feature. Honestly, I don’t mind which discrete card brand is better. But, in reality, Nvidia has a better positive review and performance than AMD. Although, AMD have better open-source driver support.

I think it is would be such a good idea. But maybe it requieres an especial expansiĂłn card interfaces because as @Ian_Darwin said, the interface could be a limiting factor. Also it could requires a ventiling system to do not overheat the laptop.

PD: greetings to everybody from Venezuela, dreaming to buy one of these in a future :wave:

Given the size of the laptop, an actual onboard dGPU would probably be limited to something like an MX450 for people who would find it useful to have CUDA, or maybe a 1650 with some clever thermal design.


yes, I think RTX series is not fit on current framework laptop because they support Tensorflow and raytracing which is quite complex and bigger laptop size is necessary in this case.

Yeah, it’s probably that we see something like that on a future model of the framework Laptop (maybe a gaming edition or something about it)


I would rather just have good integrated graphics than bad dedicated graphics - this size probably wouldn’t be able to handle dedicated graphics without clever thermal design


Great Work Framework Team! About GPUs, we shouldn’t forget.there are two.different user profiles: gamers and creatives (3d graphics engines, illustrators, Motion Graphics and Video rendering). (same for the CPU) maybe there would be a way to achieve something.


For starters, I think Framework should start with Geforce/Radeon series rather than Radeon Pro/RTX(Quadro) because the users’ amount is much higher such as gamers, AI researcher, data scientist, developers etc.

1 Like

Maybe, on the next future, there will be another great modular solution as a laptop dock station. I am not talking those to stand laptop closed on vertical axis while connecting all kind of ports.

I had been looking for something few years ago but it didn’t exist, yet. Maybe Team would think about it for their models.

A stand like the air-cooling ones for office (to elevate ergonomically the screen view) but with some extras (ports, extra-battery and eGPU) all modular, so you can change also change GPUs, battery… if needed it.

I am very keen to play an order for a DIY laptop, but I do need a CUDA-enabled card. I am trying to follow the conversation in this thread, especially @Kieran_Levin 's responses, to understand if I can indeed have such a card.

I don’t have much experience with hardware. Is the final verdict here that it will be possible to have an CUDA-enabled GPU, but with somewhat reduced performance?


Really, I don’t think, that there is enough space left in the current model. Exposing PCIe lanes for a dock is a possibility with H-series chips but I really do not expect that in a 13-14 inch form factor. I am actually impressed by what they were able to do with the current form factor, 28w is good for this size. Blowing that up to 35w-45w for the cpu and 25w-60w for the gpu does not sound like it is in the realm of possibility.