Ever wondered that after having only OpenGL around for many years on graphical processing units mainly for games and lot of other GPU related standards such as Open CL, EGL, SPIR, VULKAN, SYCL, NNEF, etc. are poring up? Well we are confronted that a lot of areas such as chemistry, engineering, defence, image analysis, AI Learning, climate, quants and many more processor intensive calculations are pressing into IT with full automation. In the past it was usual to restrict those ares to small sample sizes, but today many of these compute intensive ares are being automated.
As such we get into into processor hungry scenarios that often scale exponential to the natural logarithm e. AS such the architecture of the GPU with the many cores and shared memory is of superior use with large savings.
However this of course also means change beyond the hardware in architecture, such as that enterprise and business architects are expected to be good in math instead of just processes, for solution and software architects to design the best and bests callable processor design. As such I would recommend that at least all architects look at the basics as with the basics you can already do a great deal as most others do not even know the basics.