Performace changes

[Novinir]

Hi there

I have experienced some strange things in hashcat 3.40.
I tried to crack wpa2 hashes but my speed changed all the time when i checked the status of the process. 
At the first minute it was stable 30 KH/s than it jumped to 75-141 KH/s which is insane and cant be true.

I'm using this code:

hashcat64 -m 2500 -w 3 filename.hccapx -a 3 ?u?u?u?u?u?u?u?u

and my rig is:
nvidia 940mx with the latest drivers and i3 6100u processor.
I know this build is not good for cracking but I was curious about the performance.

I also tested and older version Cudahashcat 2.1 and the performance was stable 34 Kh/s which is good
I used the same code which is:
cudahashcat64 -m 2500 -w 3 filename.hccap -a 3 ?u?u?u?u?u?u?u?u

One more thing what I dont understand is
At the progress line I see how many possibilities are there.

Progress.........: 191578112/626481193728 (0.03%)
It means there are 626481193728 different passwords and it already tried 191578112 right?

Now in cudahashcat it looks differnt for the same code:
Progress.......: 15826944/208827064576 (0.01%)
It says there are  208827064576 possibilities but it is nowhere near to that number what hashcat 3.40 gave me.
What am I doing wrong or what am I thinking wrong.

I wanted to try 8 character long, only uppercase letters from the English abc.


I've tested the same things on a different laptop which had an nvidia 950m and an i5 proc. The problems were the same.

[rico]

208827064576 * 3 = 626481193728

There are three handshakes in your .hccapx file (that's OK). I think this is also what's throwing your hashrate off.

I brought this up before too.

[hulley]

Try the latest beta, it don't seem to have that problem, and wpa/wpa2 is more optimize than v3.40 release. The speed variables deference is relatively stable https://hashcat.net/beta/