11-30-2023, 11:58 PM
Many attacks might have trouble fitting into 512MB, even if all 512MB could be used (with CUDA). But when using OpenCL, there's a maximum allocation of 1/4 of GPU memory (so 128M here). You could force CUDA to see if that helps with you 512MB card (but honestly, such a card is so old that it might not perform as well as a modern *CPU*, depending.
~