I don't have the game but the 3D V cache CPUs tend to do well in stuff which is traditionally single thread heavy - there are some exceptions though where the 14900 is king in some older single thread heavy games. (Depends if the main game thread is repeatedly accessing the same information which can be cached the 3D cache CPUs do well, otherwise MHz is key).
I don't think those cases are Mhz is key. That is memory.
Look at factorio. If the data set fits within the L3 v-cache the X3D parts are miles faster. If you keep growing though the 14900K can catch up and overtake due to lower memory latency.
We see similar with Stellaris benches. Initial testing at the start of the game shows the X3D parts being miles faster than anything else but by the time you get towards end game that difference is a lot smaller as the data set exceeds the v-cache amount (although the 9800X3D does still do the best in that title by a margin so there is more to it than just pure memory latency or bandwidth).
That is why these titles are really interesting to test because where is the cross over point? Is the X3D part actually the best in the heaviest scenarios in those games? We don't know because they are rarely tested.
The other thing is those games work on potato GPUs so they are also use cases where someone who primarily plays those games (and then dabbles in AAA games that strike their fancy) will be better off spending more on CPU than on GPU.
We see this in ACC where even at 4K the X3D parts absolutely dominate everything else. I think similar happens in iRacing but that is less frequently tested. The interesting question in iRacing or ACC though is does that apply when you are in a race with 20 other people and you have a lot of networking going on as well as the game.
Who really cares about 10 sites testing different areas of starfield when nobody really plays it when we have actual questions that could really do with answers in games that are a lot lot more popular and that are not just fleeting popular but are popular for years and years (Civ 5 has more players on steam than Hogwarts, Hitman, Jedi Survivor, Star Wars Outcasts and many other games that you frequently see in gaming suites and that is a 14 year title that is competing with Civ 6).
Current game suits in CPU tests are simply not broad enough to cover a decent chunk of the actual gaming landscape. It would be like a site doing a productivity test suite but instead of having 3D renderers, encoding, adobe suite, compression / decompression, office suite tests they just test 10 3d Renderers and call that a productivity suite, it is clearly a nonsense.