Are there any good resources for understanding how CPU performance requirements scale given plot compression levels (either using some standardized benchmark measure for each CPU or with a bunch of individual CPU examples)?
There isn’t yet, it’s still very new. You can run a simulation with your own hardware to maximize its potential.
On the official Discord you’ll find two polls that collect that type of data, the results will be shared there too.
There’s a bunch of results here:
Most is GPU but also some CPU results.
you can download the latest bladebit alpha (Index) and run the bladebit simulate on your CPU:
./bladebit_cuda -t <threads> simulate -p <parallel context> -n 1000 --no-cuda <plotfile>
Threads x parallel context = total system threads
So if you have a 2c4t CPU, you can try -t and -p 1:4, 2:2 and 4:1
./bladebit_cuda -t 2 simulate -p 2 -n 1000 --no-cuda C:\K32-C4-abc123.plot
In general CPU performance scales pretty ok with passmark multicore results
You can try Gigahorse tools to estimate max plot size.