UMA removes the PCIe bottleneck and replaces it with a memory controller + bandwidth bottleneck. For most high-performance GPUs, that would be a direct downgrade.

> For most high-performance GPUs, that would be a direct downgrade.

You really can't say that, it depends on what you're running. If your model fits within a dGPU's VRAM then yes, obviously, but plenty of models are larger.