You also have the benefit that, since it’s a single memory pool, it can be used for other things, and allocated dynamically.
Where in a conventional machine, VRAM is exclusively for the GPU to use, and RAM is generally for the system. The capacities are usually fixed (I think it’s possible to offload GPU memory into system memory for some uses, but that’s about it, and your performance usually falls off).
Downside is that because it’s a shared memory pool, everything uses it, so you lose some of the actual capacity for things like your OS.
You also have the benefit that, since it’s a single memory pool, it can be used for other things, and allocated dynamically.
Where in a conventional machine, VRAM is exclusively for the GPU to use, and RAM is generally for the system. The capacities are usually fixed (I think it’s possible to offload GPU memory into system memory for some uses, but that’s about it, and your performance usually falls off).
Downside is that because it’s a shared memory pool, everything uses it, so you lose some of the actual capacity for things like your OS.