It depends entirely on workload. We do a lot of game development work. One of the common bottleneck loops is, run code, edit code, build code, launch and run again.
The modern editor feel better to use with the high single code peak frequencies. Same with the build step. Yes, parts of the build fan out to fill 100+ threads, but there are still parts of the build that are stuck on a few (less than 4) threads. By time on an incremental build (most common), they are the majority of the time.
Workstation is different than server because light load latency has a much bigger impact on user perception. So the extra light load boost clocks actually help a lot.
As to the extra memory channels, that’s a lot harder to pin down and will really depend on what you do with it. We run compute tasks that can take hours, I don’t think from looking at the profiles it really stresses the memory subsystem. For us full speed AVX-512 was a lot more important which is why we are only now considering switching off Intel.
I’d get the pro, less for the extra memory channels, but for the IO. The number of times we found ourselves out of PCIe lanes or slots with older machines is just an area where it’s future proofing for minimal cost.
