Using an Intel GPU on Macbook Pro

Hey Julia Community,

I’m only just starting to learn about GPU programming (mainly for deep learning with Flux, but I’m curious if/how I could use it for my other projects).

One barrier to getting started, though, is I can’t seem to find a solution for using the Intel GPU on my Apple laptop. I see that Metal is designed to use Apple’s M-series chip, while oneAPI uses Intel but requires a Linux OS. Other than paying hourly for JuliaHub’s GPUs, is there a way I can write/run Julia code using a GPU on my own machine?

Grateful for any help with my novice question!


What do you mean? Your Apple laptop does not have an Intel GPU. (Edit: unless I misunderstood, and you have an old pre-M1 MacBook, but then forget about running any kind of GPU computing on there). There is no way to add a graphics card internally or externally to any recent Apple computer. You are stuck with the integrated Apple Silicon graphics system.

In theory, the Apple Silicon graphics should be quite capable for GPU computing, but Apple isn’t very eager to expose the relevant APIs. I’m not sure what the current status of support within Julia is.

The de-facto standard for GPU computing is NVidia. If you want to play around a little, the easiest thing to do is probably to pay for access to GPU cloud computing instances. If you’re more serious, consider building a Linux workstation with whatever GPU is in your budget. Either way, you’ll be connecting via SSH from your MacBook to a Linux machine and run your Julia GPU code there.

1 Like

@goerz Thanks for the reply and advice!

My MacBook claims to have a “Intel UHD Graphics 630 1536 MB” card. It’s a 2019 (hmm, a little older than I thought—maybe the M-series chips are standard now and my info is out of date). Either way, it sounds like cloud computing is my best option, at least until I become more serious and have the budget for a new machine.

Yeah, sorry about the misunderstanding… The M1 MacBook was introduced in 2020, so you still have the older Intel-based architecture. The on-board Intel graphics definitely aren’t able to run any kind of GPU computing, unfortunately, so you’re pretty much stuck with cloud computing. Or, if you have some kind of university affiliation, they might have some computing resources that you would be able to use.

1 Like

Metal does support Intel IGPs, but Metal.jl doesn’t (we currently only target M-series GPUs).


You can sign up for Intel Devcloud

1 Like

Also if you are a commercial or academic customer and are wanting to evaulate kit, all vendors have demonstration centres.
Dell has Customer Solution Centres (CSC) in many locations, and development/benchmarking clusters.
Reach out to your sales team from whatever vendor your company or university like sto deal with.