Relevant sidenote: Drake's multithreaded perf likely exceeds its single thread perf by a significant margin. So some CPU bound games might come over and some might not. The whole reason a console is able to punch above it's weight is that it's an elegantly designed ecosystem, and not just a GPU.
Is this based on Geekbench results? If so, I have reason to be somewhat skeptical of Geekbench's multicore results, specifically on Geekbench 6.
Recently I upgraded my desktop from a Ryzen 3700X (8 core Zen 2) to a 5900X (12 core Zen 3), and just out of curiosity I did a handful of before and after benchmarks. The Geekbench 6 scores in particular were a bit surprising. Single core scores went from 1647 to 2194 (+33%), and multi-core scores went from 8061 to 10623 (+32%). Single core scores went up by slightly more than multi core, despite core count going up 50%, which is a bit weird. For comparison, on Cinebench I was getting a 23% increase in single core results and a 70% increase in multi core, which is more along the lines of what I was expecting. Of course I don't expect everything to scale as well as Cinebench, and moving from 1 to 2 CPU chiplets would have an impact on some multithreaded workloads, but seemingly not getting any benefit to having 50% more cores in the multi core results was puzzling.
Looking through the results of the sub-tasks, though, I think the reason is that some of the underlying tasks in Geekbench 6 just aren't that well multithreaded. I don't have the original results handy, but I found comparable results for the
3700X and
5900X on Geekbench's results browser, which should be representative enough. The one to look at here is "Text Processing", where the 3700X hits 1379 in single core mode, but just 1730 in multi core, so it's only getting a 25% speedup from running on 8 cores instead of 1. At the other end of the spectrum the Ray Tracer test jumps from 1666 in single core to 17541 in multi core, about a 10.5x increase.
Having some tests be better multithreaded than others is probably a fair reflection of real-world code, but in this case it seems to be skewing the results somewhat. Text processing is one of the sub-tasks which the 5900X improves the most in single core mode, with a 55% better score than the 3700X. Conversely, it sees much less improvement in ray tracing, with only a 22% better score on a single core. However, ray tracing counts for a whole lot more than text processing in the multi core results, because it's actually using those extra cores, so the multi core results are a lot less impressive than the single core results and it seems like the 5900X doesn't have good multicore performance, which isn't necessarily the case.
It seems like, in part Geekbench 6's multi core results are measuring the true multi-threaded capabilities of the CPU, but in part they're also measuring how good the CPUs are at the specific tasks which happen to be well multithreaded in Geekbench 6.
) 
