md5 wordlist and md5 wordlist+mask 1000 times difference

[blaster666]

Could someone explein me how is this and why happening

When i am running wordlist attack
-m 0 -a 0 1.hash 2.pass

I am getting 15 000 KH\s Speed


When running 
-m 0 -a 7 ?a?a?a?a i am getting 15 000 MH\s SPEED! Which is 1000 times higher


Is this because of hashcat is easier to generate and bruteforce passwords from arguments like ?a?a?a?a, than brruteforcing straught txt ?

Do I understand correctly , help me with advice please

If I am willing to bruteforce some words, and add all possible 4 symbols in the end of the password, i MUST use attack mode 7 , and add ?a?a?a?a, and I MUST NOT generate wordlist by myself, where will be passwords of all possible combinatoins with 4 symbols, becauase for hashcat generating these symbols is easiser and perfomance faster, comparing with common wordlist bruteforce ?

[blaster666]

But as for bcrypt method, th e situation is another

When i am trying bruteforce $2a$10 bcrypt i have 2000 h\s on common -a 0 wordlist attack

But when I am using smaller wordlist + mask for combinations, I am not getting the same effect as on MD5 , it remains 1800-1900H\s

I would really appreciate anyone who could advise optimal way for bcrpyt 10 round attack

[royce]

One of them supplies the GPUs with more-efficient "work" to do.

Shipping the wordlist to the GPUs has latency.

Executing mask attacks happens directly on GPU, and is not affected by that latency.

[blaster666]

One of them supplies the GPUs with more-efficient "work" to do.

Shipping the wordlist to the GPUs has latency.

Executing mask attacks happens directly on GPU, and is not affected by that latency.

do you know if I could somehow increase amount of Vec:1 , orr it makes no sence?
While trying to use -vector(i do not remember full) command I am getting error that it is outdated

[blaster666]

One of them supplies the GPUs with more-efficient "work" to do.

Shipping the wordlist to the GPUs has latency.

Executing mask attacks happens directly on GPU, and is not affected by that latency.

And this is how it works for md5, but for bcrypt, which is actually slow method itself, it does not help, do I understand it correctly sir?