Okay so I'm pretty much all set up. Here are my thoughts and some data. The build and setup was super smooth, no issues, and not one lock up freeze or other anomoly during the entire process of loading tons of applications, audio and video plugins, activating licenses. Just worked.
Overall performance is of course much better than my 14900K in every application I use. All I did in the BIOS after my initial CB R23 run was enable EXPO for my memory.
CB R23 MT score was initially about 41500 but that dropped to 40600 once I loaded everything up and strangely enabled EXPO memory, which I thought would have bumped the score a bit. Looks like there is a 200W package power limit in the BiOS somewhere because it bumps that in HWinfo. Average clock during the run was 4.7GHz at 88C MT, and 5.64GHz ST.
I have a 4 RAW photo test I do using PureRaw4. Using my ARC750, the 14900K could process a photo in 11 seconds, 9950X in 9.6 seconds. Keep in mind that the GPU is doing the heavy lifting here. CPU only the 9950X could process a photo in 37.5 seconds. Unfortunately I don't have this data for the 14900K but I do know it was 75 seconds per photo with PureRaw3. I'm sure v4 is faster then v3 but not twice as fast. Zen 5 is more performant in PureRaw both assisting the GPU and on it's own.
Our forum Handbrake test took 100.5 seconds. I was bumping against 200W there as well. My effective clock was much lower than Detox but my efficiency per frame higher. Perhaps lower latency memory is better than frequency in Handbrake for this test? I made sure to test using Handbrake 1.3.3. I will update those results in that thread now...
So overall I'm quite happy. First, there have been absolutely no incompatibility or instability problems. Second, performance is better across the board at lower temps and power. While Zen 5 is easier to cool than Raptor Lake for similar performance, as is always the case, if you really want to push there is a lot more performance on the table with Zen 5. I'm happy with the 200W limit. The IPC of Zen 5 really is quite extraordinary in my applications.
Oh yeah, quick "seat of the pants" test in Presonus Studio One rendering out a multitrack file. 14900K would generally do 9 times faster than realtime, 9950X is doing 12 times real time. That's a big one for me because I'm waiting on the mixdowns quite a bit.