AMD budget GPU listed with twice memory of RTX 4060, RX 7600 XT with 16GB GDDR6

In what way is the headline misleading? The 4060 Ti is not a 4060 in the same way a Ryzen 7900 is not a Ryzen 7900X3D. The GPU you, and the article, mention is a completely separate model with a completely separate chip and drastically higher price.

> 1) Driving monitors. If I want a dozen word processor windows open at the same time, I don't need a thousand CUDA cores, but I do need enough ports

My iGPU compact desktop drives 4 monitors just fine. After that... you're not really in budget GPU territory.

> 2) Acceleration. This costs more money relative to CPUs today. Most people don't do much that requires this (although this might change with SD, LLMs, etc.)

iGPUs can do the same accelerations as dedicated GPUs these days. The difference is in whether it's part of the main SoC or a larger separate chip. As a further example the iGPUs in the MacBook even scale to being better consumer accelerators than mid-high range desktop GPUs for large models.

> 3) Offloading bus load. Things like screen refresh, for an iGPU, requires constant access to main memory. On a GPU, there is dedicated graphics memory. A server showing a lock screen has constant bus load with an iGPU, but not with a dedicated graphics card.

This I largely agree with, though as its impact on the overall measure of performance not an inherent requirement of a budget GPU.

> As a footnote, fundamentally, what I would prefer is an architecture where I have:

It's your preference so I'm not going to argue any of it is right or wrong but I'm curious what the value in having the output and accelerator as physically separate cards is. You can already do things like DisplayLink to output over arbitrary bus interfaces and you can indeed multi-head with different brands (works fine, doing it right now) so what's the value difference in the built in outputs being physically separated by some additional interconnect?

Most power in a computer is spent on a bus. Most power within CPU is spent on internal bus and memory access. This interconnect between framebuffer and accelerator would limit the speed and eat lots of power, unless the framebuffer would be replicated by GPU during rendering.

That is why GPUs normally have a small part that serves as framebuffer, but memory bus is shared with the accelerator that needs it most. Cheaper to reuse the expensive part.