Although I've grown exceeding rusty with Assembly from lack of use (and maybe because I was never much better than "See the dog. The dog can run fast"
) there was a time when I played extensively with hacking firmware and intimately aware of the steps in POST. I swapped CMOS chips
while the system was running to try out modded images, reprogram chips, enable and disable features and add some new ones. As long as the I/O chips are similar even different manufacturer and model mobo bios images could be used, sometimes to substantial effect for what settings and how much control was possible.
As long as early PCs could retain ISA peripheral slots only that simplicity allowed early BIOS chips to have sufficient memory space to handle any hardware that could be optionally plugged in but soon the progression began and the number of Interrupts and size of memory had to be nearly doubled. It was inevitable that continued progress could not be handled by simply adding more BIOS hardware. UEFI wasn't just a lark. It is needed at the hardware level.
So maybe it can be seen that I'm not quite a Luddite and respect and concur with the move to something like UEFI. I do take some small issue with what has been done with it however, which has been largely cosmetics, convenience and a few agenda-ridden (largely from Wintel) "features" that have little to do with what is needed before handing off to Int 19h. The time required for POST is quite variable based on what and how many settings are employed in Setup and since UEFI has massively more horsepower and resources it can be faster than BIOS but since it also has more settings and more hardware variables possible, it can also be slower. That's why I focused on Boot as being the time after POST since once the image is loaded into RAM, and even though that image is constantly "consulted", it is all up to the OpSys and it's init system.
From my POV UEFI could, and likely will, become far more important and useful in the future, but as of 2019 it's largely just a bauble and has but little gain in boot time or any other function beyond hardware compatibility. I sincerely hope Legacy support will continue for at least another decade. It is worthy of note that although Moore's Law is still advancing, so is software bloat and complexity. Just consider how much was accomplished with so few resources with the Apollo Moon Project computers.
--- Software Apocalypse --- 
Be that as it may garpu, it is not because of UEFI. If you look at the boot process it still reaches a point where everything is handed off to the kernel of any OpSys. This is one of the reasons that while it is possible to access drives like CDROM/DVD/USB from the CMOS and even run apps and boot from them, once an OpSys begins to load a driver is required to do the very same thing. FWIW my mobo supports UEFI and I've tried it, then opted for Legacy mode though I do have a 4TB HDD that is GPT. I can no longer boot from it but that isn't a problem since I have 2 other smaller drives in this box. It is fine as a large storage drive. The point is that not only is the theory against any increased boot speed due to UEFI (other than for the Windows Fast Boot option) but also is the practice. Boot speed is entirely about the read speed of the drive, RAM, and the complexity of the kernel and init system. BIOS and UEFI are simply a recipe and set of rules for how hardware "talks" to each other. It is somewhat analogous to our human autonomic system like heartbeat and breathing that can be taken over by a conscious imperative.
