NVLinked RTX8000 and large matrix inversion

If I NVLink two RTX8000 GPUs each with 48GB VRAM, can I allocate and invert a large matrix as if I have a single GPU with 96GB VRAM?

Answer 1

[Alison Eele]

I'm afraid you won't be able to treat the two cards as a single card in this situation. The underlying system will still see the GPUs as two separate pieces of hardware and MATLAB does not currently support distributed GPU arrays.

The NVLink will allow quick communication between the two cards which can give benefits as part of a deep learning workflow.

[Jhinhwan Lee]

Sounds like a clear and reasonable answer to me, exactly the kind what I wanted. Thanks!