GPUSieveSize:
Improved performance can be obtained by setting GPUSieveSize to 2047, however this will not work on older GPUs (all GTX 1xxx and older, and any with less than 8GB RAM). If you have problems at 2047, set it to 128 in mfaktc.ini and try increasing it (256, 512, 1024). Higher values give slightly better performance at risk of lag (or crash) if the GPU has insufficient RAM.
Note that the default mfaktc.ini setting is 256, this should be increased as much as supported by your GPU (2047 for RTX 20x0 and newer).

Why so many builds?

Different CUDA versions support different hardware (compute capabilites), and different driver versions support different CUDA versions. As such, we need multiple builds to support all the various hardware/software combinations.

Use mfaktc with AutoPrimeNet.

Which build do I use?

The one that matches (or is the closest less-than-or-equal-to) the CUDA version supported by your driver.
e.g. Driver supports CUDA 12.1, use the build that targets CUDA 12.0.

To check your driver's CUDA version using the NVIDIA Control Panel, open the "Help" menu, "System Information", "Components" tab,
look for filename NVCUDA64.DLL (or NVCUDA32.DLL on 32-bit) in the left column, then in the right column ("Product Name") you
should see something like "NVIDIA CUDA 12.7.33 driver": the "12.7" is your CUDA version.
Another way to check is to run nvidia-smi (from a command prompt), which reports the supported CUDA version on the first line
(on driver 410.72 and newer¹).

The below table shows the CUDA versions and recommended builds for some of the more popular driver versions: