03-17-2012, 12:21 AM
I found this weird problem after getting a failed sl3 hash. The problem is as follows:
After attacking this hash with single 5870, Hashcat is reporting that the code is found on 45.9723% (459723570421760/1000000000000000)
The same hash, attacked on node with 5 x 5870, Hashcat reports that the code is found on 45.9979% (459979570421760/1000000000000000)
The same hash, attacked on a node with 8 gpus (2 x 5970 + 4 x 5870), Hashcat reports, that the code is found on 46.0027% (460027360256000/1000000000000000)
And here comes the drama: I distribute the whole sl3 sequence (100%) in a 1% chunks. With the problem described above, when I give a node with more than 5 x 5870's the chunk 45-46%, Hashcat reports me, status "Exhausted" and does NOT return a code. But it doesn't find the code in the next chunk (46-47%) either.
Why Hashcat is skippng the code, and returns "Status:... Exhausted" prior to "Status:... Cracked" ?
I tested this with 0.08 and 0.09, I get absolutely the same result.
P.S.: Actually this should be considered mainly as bug report. The percentage examples were just to better explain it. In short instead of getting status "Cracked" I receive status "Exhausted". You can try yourself. Problematic hash and imei are CF2EBC05F24A33D27B81959D5AF03707322196BC:003582750452025100.
For this hash, to observe the problem, you need greater speed than 12000 M/s. But as close to the end of the chunk the code is, (for example 45.999%) the chances to get Exhausted prior to Cracked increase.
After attacking this hash with single 5870, Hashcat is reporting that the code is found on 45.9723% (459723570421760/1000000000000000)
The same hash, attacked on node with 5 x 5870, Hashcat reports that the code is found on 45.9979% (459979570421760/1000000000000000)
The same hash, attacked on a node with 8 gpus (2 x 5970 + 4 x 5870), Hashcat reports, that the code is found on 46.0027% (460027360256000/1000000000000000)
And here comes the drama: I distribute the whole sl3 sequence (100%) in a 1% chunks. With the problem described above, when I give a node with more than 5 x 5870's the chunk 45-46%, Hashcat reports me, status "Exhausted" and does NOT return a code. But it doesn't find the code in the next chunk (46-47%) either.
Why Hashcat is skippng the code, and returns "Status:... Exhausted" prior to "Status:... Cracked" ?
I tested this with 0.08 and 0.09, I get absolutely the same result.
P.S.: Actually this should be considered mainly as bug report. The percentage examples were just to better explain it. In short instead of getting status "Cracked" I receive status "Exhausted". You can try yourself. Problematic hash and imei are CF2EBC05F24A33D27B81959D5AF03707322196BC:003582750452025100.
For this hash, to observe the problem, you need greater speed than 12000 M/s. But as close to the end of the chunk the code is, (for example 45.999%) the chances to get Exhausted prior to Cracked increase.