Hello! So I am pretty into graphics programming. I have been doing it for awhile but mostly with triangles using OpenGL and Metal. This next quarter I am taking a course that is going to have a larger focus on raytracing and I want to be ready to do not only the class but also some relevant tutorials and books that have been on my to do list for awhile now.

My question is essentially if I should be buying any hardware or virtual machines before entering the quarter. Ideally if computing tasks I put my gpu up to are going to take more than a minute I will probably be wanting to set up a remote machine to handle those tasks.

I also want to play around with Cuda programming.

Some of these purchasing decisions are a bit time sensitive as I have an NVIDIA employee discount for about 2 more weeks. (I am an intern)

I currently have a MacBook Pro 2017 15” with maxed out specs, including the GPU. That means it has a Radeon Pro 560.

I don’t know specifically what demands the class will have but I am suspecting we will be doing something cross platform probably without a real graphics api.

Do I need to get an NVIDIA RTX GPU

It is unclear to me if Direct X or Vulcans raytracing APIs will work with my GPU or if I need a GPU with accelerators for those tasks. Is this a requirement? Furthermore should I expect raytracing to work on a Bootcamped Windows machine?

Should I get a remote windows virtual machine, if so from where

I use digital ocean for my Linux boxes (they make it so ridiculously easy to get a cheap box going compared to AWS and Azure) however they don’t do Windows machines, I’m not sure I can add a GPU, and I am not sure how good their service is for compute in general.

What cloud service might make it the easiest for me to do rendering in the cloud by sending them my program.

I remember people talking about services where you could just rent hardware for a few seconds and run your workload on them. This always seemed a bit magical to me because I know whenever a remote machine is opened I’m going to have to spend a good 20 minutes installing dependencies and getting it all set up for running the task before doing it. So do I really want to pay for gpu time while doing that?

What kinds of services are out there for this these days? Has anyone abstracted it to the point where I can just send roughly a program and request it be run on x hardware?

What should I be looking at here?

Easiest way to be able to do some Cuda programming

Another factor that might make me want to get an NVIDIA gpu is that i want to dive into CUDA programming. The thing is as a laptop user that is VERY mobile the idea of carrying around a housing and a gpu all the time just doesn’t sound pleasant. Is there another way?

Furthermore I have heard that Apple has really stalemated the efforts to get NVIDIA gpus working within Mojave and I’m not even sure if it’s possible anymore?

Again this is where boot camp might come in clutch but will that really work with egpu and everything?

Is there some cloud service where I can get a box capable of doing some Cuda programming? Nothing too intense I just need to be able to run it.

So I guess ** in summary** if you have any recommendations on whether I should get an egpu or if there is a cloud service I should look into please let me know. I love doing my work on cloud boxes because they don’t heat up in front of you, slow down your usual tasks, or fail when you close the lid of your laptop because you have to move onto your next task for the day. BUT I’m not a fan of using cloud services when they are difficult to setup and maintain. I have a tolerance for a few hours and that’s it.

Any recommendations, thoughts, and advice appreciated. Especially if you know whether or not Vulcan or Direct X require RTX gpus.