1 of 1 people found this helpful
For more complete info on AMD Vega GPU Architecture, you should refer to the Vega Whitepaper.
Long story short: the architecture and micro-code are designed to better respond to DX12 and Vulkan instructions.
As for why AMD CPU cache is better... technically speaking, a cache is just that, there's no radical difference, except on AMD everything is interconnected via Infinity Fabric, and Ryzen integrates a lot of functions directly in the CPU via SoC, rather than relying on motherboard features, which leads to a faster communication between components (i.e. RAM, GPU, SSD, PCI and others). Also if you follow AMD progress, you'll notice they prefer maximising PCI lanes as much as possible.
Infinity fabric applies even more so and is probably more relevant on high-end desktops with Threadripper, or servers - read a bit more into EPYC in this research whitepaper.
Hope this answers some of your questions.