07-26-2024, 03:04 PM
This is kind of hard to answer, as there is a lot involved in masks but put simply, Hashcat distributes the first 4 bytes to as many threads on the GPU as it can (called the base loop), then each of those threads has an inner loop (called the mod loop), where the rest of the candidate is generated by the GPU itself, so if you had ?l?l?l?l?d?d?d?d, the outer loop would be ?l?l?l?l and the inner loop would be ?d?d?d?d. This is the reason why if you use a mask like "ABCD?a?a?a?a?a?a", it'll be extraordinarily slow, because there'll only be a single candidate, "ABCD" to distribute to the CPU/GPU threads, while if you do it the other way around, as "?a?a?a?a?a?aABCD", it'll be near full speed as it can be parallelised fully