Combinator attack slow down

[Wolfdale]

Hi! I'm using the combinator attack with cudahashcat. I have a gtx 680. For MD5 bruteforce I get 2700 MH/s. If I try combinator attack with 2 dicts (20k words and 120M words) I get 1500MH/s (I believe it is because i'm not using any rule?)
The question is why towards the end it slows down to 1000 KH/s?



Probably I'm doing something wrong, I've got alot to learn

[coolbry95]

Combinator attack is always going to be slower. Everything is slower than a mask attack.

OclHashcat slows down on purpose towards the end. It is adjusting the workload. Also I do not know if you are doing this or not, but the left dict needs to be larger than the left. The reason for this is the way GPGPU works.

[atom]

It's also because with mask-attack (or general with brute-force) we can do some optimization because we know what comes "next". Therefore we can make use of meet-in-the-middle etc.