Can't you just plug a latest nVidia card to you MacBook with Thunderbolt when you need it? Also doesn't M2 have some sort of an accelerator built-in specifically for ML?
> Can't you just plug a latest nVidia card to you MacBook with Thunderbolt when you need it?
No, even when they shipped that support initially it barely worked and then support ended one OS release. Nvidia stopped shipping drivers so it hasn't been possible for around 6+ years now?
Looked into it extensively at the time and price and support wise I ended up having to build a PC to use Nvidia for 3D work (Octane Render).
> Also doesn't M2 have some sort of an accelerator built-in specifically for ML?
Yes, and it's good for a laptop but to put it in perspective M1 Max takes around 45-50 seconds to generate a Stable Diffusion image, 3090 takes around 5 seconds.
Very very sad news. I hoped to switch from a PC laptop to MacBookAir someday and use a GPU and other PCIe things over Thunderbolt whenever I would need them.
Apple used to support external GPUs on Intel-based Macs but they killed that when they switched over to ARM. It seems the new Mac Pro won't support internal PCIe GPUs either. You're just stuck with whatever GPU is built into the SOC now.
