I’ve scoured the internet, and maybe my Google-fu just sucks today, but I can’t find anything comparing the performance of the AI accelerator on the Framework’s processor to Google’s Coral AI accelerator or Nvidia’s Jetson Nano.
Has anyone here done similar testing or have anecdotal comparisons they can make?
I don’t want to take a graph that doesn’t list the processor (or maybe I’m blind) as proof that it sucks, but this is good. Thanks!
If it is truly bad, I’ll look into integrating a Coral TPU into another custom expansion card. I want to make a MagSafe card (using @CJ_Elevated’s designs), but a charging-only port seems a bit wasteful, so if I can integrate a TPU into it as well, that’ll be great.
Update: Coral card much bigger than I thought, formulating plan B.
Yeah, I’m now looking into using this in an oversized slot (hopefully with some other functionality as well) if/when the framework team releases the CAD files for their oversized card design. I’m really hoping it’s oversized down rather than out; Even if it’s harder to fit components inside that way, having it symmetrical with two oversized cards that are taller (in the two backmost slots) would give the laptop more functional airflow, rather than being a bit of an eyesore if you had to have an oversized card that jutted out the side of the laptop for your daily workflow.
I saw that. Honestly lga soldering still scares the daylights out of me, and I want to try to stay within a reasonable budget ($50/expansion card), which PCB design probably wouldn’t accomplish, but that does sound like a fun skill to learn, and I may end up going that route if I can’t get the dual-tpu m.2 to work.
That’s a thought, since they do use 4 watts, which is 50% over baseline, but ideally I would like the option to use it while on the move at uni. This is a good idea I hadn’t considered though. Definitely plan B if the dual-tpu m.2 doesn’t work out.
On second thought, I want to be able to charge via MagSafe most of the time, which, to avoid misalignment, can only support USB 2.0 speeds. Also, I want to have the dual TPUs, which is only available as an M.2 or an LGA unit, and I’ve already expressed my trouble with LGA. Thanks so much for your help, but it looks like I need to do some CAD work to see how I can fit an M.2-2230 in a card.