1
$\begingroup$

I have a Macbook Air and Asus ROG 702 with a built in GeForce 1070 GPU. I'm considering getting an NVIDIA Volta for AI applications (tensorflow / pytorch) to run on either the mac or windows. Would they even be compatible? And if so, what box would I need to house the GPU? Would it be system agnostic? (I was thinking Aikito Node).

Thank you! All info is appreciated by this novice :D

$\endgroup$
1
$\begingroup$

The NVidia Volta architecture is capable of delivering an astounding 110 teraflops of computing power. This coupled with the NVidia CUDA Neural Network library (cuDNN). Which can be used to speed up DNN's on learning frameworks such as Caffe 2, Microsoft Cognitive Toolkit, TensorFlow, Theano, and PyTorch. Makes GPU's based on the Volta architecture such as the Titan V specifically suited to accelerate AI workloads.

Regarding your MacBook Air. Unfortunately, it comes with an integrated GPU that can't be replaced because it is soldered to the logic board. So, unfortunately, we have to rule out upgrading it because there isn't any other port to connect a GPU. You can read more about this on Apple's official discussion forum https://discussions.apple.com/thread/5206086 .

Having said that, you possibly can upgrade your GPU within the same series which is the 10 series. Other powerful GPU's within the 10 series include The Titan X, Titan Xp, GeForce GTX 1070 Ti and the GeForce GTX 1080 Ti. Unfortunately, the Geforce 10 series is based on the Pascal architecture meaning your laptop might not by default accept a Volta based GPU such as the Tesla V100 or the Tesla Titan V. Consequently it could involve BIOS modification, chassis modification or heatsink assembly modification to make it work. This kind of upgrade will be diving into uncharted territory since it will void your warranty.

My recommendation is that a Volta based GPU might be unsuitable for your Asus laptop. I'd recommend a GPU that uses an architecture that is similar to your current GPU. Meaning a Pascal-based GPU such as the Titan X or GeForce GTX 1080 would be better suited to satisfy your demands.

| improve this answer | |
$\endgroup$

Not the answer you're looking for? Browse other questions tagged or ask your own question.