08-05-2022, 06:25 AM
Hi I had an combined (mode 1) session running with 8GPUs and an estimated time of 7 days and a few hours.
After ~5 days I accidentally killed the session.
Problem: Like always I instantaneously got the message:
Approaching final keyspace - workload adjusted.
So --restore will just start at the beginning and will not find any password for the next 5 days.
If I try to divide the keyspace by 70% (little less than 5/7) and add -s 96308073321 I will similarly get:
Restore value is greater than keyspace.
So I like to try to just split the dictionary into used and unused passwords.
As I understand that passwords of the first eighth of the leftside dictionary will always be cracked on GPU 7 these of the second eighth always on GPU 5 these of the third on GPU
6...
So I can just use split -l $(wc -l leftdict) to show the work of one gpu. But I really do not understand the order within a GPU. Just removing the upper 70% percent of every eighth the dictionary seems to be wrong.
Btw: Are there any workload balancing options? That you often have to make use of combined mode to get -S mode because mask mode isn't able to split an homogeneous workload of several days on 8 GPUs seems a similar problem like no restore points because the dictionary is to small.
After ~5 days I accidentally killed the session.
Problem: Like always I instantaneously got the message:
Approaching final keyspace - workload adjusted.
So --restore will just start at the beginning and will not find any password for the next 5 days.
If I try to divide the keyspace by 70% (little less than 5/7) and add -s 96308073321 I will similarly get:
Restore value is greater than keyspace.
So I like to try to just split the dictionary into used and unused passwords.
As I understand that passwords of the first eighth of the leftside dictionary will always be cracked on GPU 7 these of the second eighth always on GPU 5 these of the third on GPU
6...
So I can just use split -l $(wc -l leftdict) to show the work of one gpu. But I really do not understand the order within a GPU. Just removing the upper 70% percent of every eighth the dictionary seems to be wrong.
Btw: Are there any workload balancing options? That you often have to make use of combined mode to get -S mode because mask mode isn't able to split an homogeneous workload of several days on 8 GPUs seems a similar problem like no restore points because the dictionary is to small.