I'm building a workstation right now for photogrammetry, 3d modeling and video editing. A threadripper or epyc processor would be superior, but since we're doing it on a "budget" we were gonna go for 9950x. Unfortunately, during my research I found that none of the 9xxx processors officially support 128gb - 192gb of ram (which is something I need). I'm told I can probably get it to run at lower, DDR5 default, clock speeds... but at that point, not only am I paying more for a little bit of efficiency, but I'm also losing a lot of performance that I feel I should be getting.
Initially I was gonna build the 9950x with just a good air cooler. But we're going to switch to a 14900k which saves me like $200, throw $50 more into the cooling solution and get an AIO, for a total of $150 in savings. It's not a personal PC, and it's gonna be running in a noisy office with lots of different equipment, so the extra noise and extra power consumption is really irrelevant to me.
I know my photogrammetry software performs better with more L2+L3 cache, which is why I wanted the 9950X. But AMD really should put more effort into their memory controller. As well Intel has bigger L2 cache sizes which is probably part of the reason why they continue to perform very well in multithreaded situations. So for a user like me who is building a machine for video editing / 3d modeling / highend, lots of ram workstations that are step below server gear... there's a reason why I will continue to opt for Intel over AMD.
And it's funny too, because AMDs threadripper and Epic processors are king for what I wanna do. So it's not like AMD doesn't possess the technology to completely wipe the floor with Intel. They just chose to cheap out and offer an AMD sidegrade version to the 14900k instead of releasing a 14900k killer. They should have released a killer.