Riddle me this...
1. Why does nVidia have about an 80% market share in computer graphics?
I think the big turning point for Nvidia was around the time of Maxwell. That's when their Tegra program forced them to focus keenly on energy-efficiency. Maxwell was the first big beneficiary of that focus, with tiled-rendering being a specific example. They doubled-down on this, in Pascal.
Then, having built such a lead, they decided they could afford to shift their focus to new technologies, in Turing. So, they added Tensor cores and Raytracing. Now, AMD had not only a performance gap to close, but they also needed to invest in competing on those new feature sets.
RDNA was AMD's response to the efficiency problem, but it was a long time in coming. Pascal launched in 2016, RDNA came 3 years later. AMD is still playing catch-up on raytracing and deep learning accelerators.
In other markets, Nvidia did a brilliant job of jumping on the deep learning bandwagon, early. They made CUDA stable and supported it on virtually all of their hardware, which made it the natural platform of choice for deep learning researchers. AMD was also suffering, financially, right when Nvidia was making these crucial investments in CUDA and deep learning.
2. Since 1998, almost every computer component has become stupid cheap in pricing, except for video cards, whose prices have tripled. Why is that?
First, graphics is an "embarrassingly parallel" problem, which means you can keep throwing more silicon at it, and still get decent performance increases. If you plot performance per Transistor of GPUs, over time, it would be
much more linear than with CPUs - that's for sure!
Also, I take issue with the assertion that PC components have really become that much cheaper. Increased integration helped reduce PC prices a lot. But, it seems like decent-performance CPUs never really got
that cheap, and prices certainly have been on the rebound since the core-count race really got going.