help with "debugging" dropping hashrates
just playing around a litte bit with the gitversion and mask-attack noticed some strange hashcat behavior and trying to investigate the "why?", maybe someone can give me a hint where i can look further

hashcat mask-attack/bruteforcing with fixed pw-length, the mask/pw-length was set up to be long and complex enough to get near or almost to the maximum hashrate for the specific hash (benchmarking) and running for some weeks, i'm working on GPU only

after some time (days?), the actual hashrate reported by hashcat is dropping from 1-2 thousand mega to a low number of 100-300 kilohashes, GPU temperature is constant, fan 60% utilisation 95%, RAM and CPU are more or less empty and idling (no other processes except for plain windows 10)

the strange behavior
i'm connecting via rdp to this machine, hashcat is reporting the values every 30 seconds (--status-timer 30) so i get a relativ good overview about the drop, when i'm activly pressing S for status, the hashrate begins to rise again to the maximum of 1-2 megahashes per second and stays constant (dont know how long, maybe i can pipe the status to a file to get a better overview)

#what i first thought:
okay the machine seems to be putting itself to a kind of low power consumption mode due to the fact that there is no "user input", so i checked all settings like (BIOS/OS) energy saving mode, windows 10 power modes and so on, but all of them are set up to maximum power and what i can see so far, the GPU utilisation is also constant at 95%

i found this thread but it seems not to reflect my problem, because there the plain md5 hashrates seems to be constant the whole time

anyone an idea what or where i should take a look now the get the "why does this happen?"