Please note, this is a STATIC archive of website hashcat.net from 08 Oct 2020, cach3.com does not collect or store any user information, there is no "phishing" involved.

hashcat Forum

Full Version: Haswell bench
You're currently viewing a stripped down version of our content. View the full version with proper formatting.
First bench in bruteforce -a 3 -m 0 gives about 96-100M on my 4770k @4,4GHz.
I do more benchmarks tomorrow.
You should wait until atom releases AVX2 binaries Wink
(06-25-2013, 03:45 AM)kartan Wrote: [ -> ]First bench in bruteforce -a 3 -m 0 gives about 96-100M on my 4770k @4,4GHz.
I do more benchmarks tomorrow.
Hey can you run this and tell me what the speeds are: https://www.tobtu.com/files/bcryptavx2.zip

I know the non-AVX2 code is a little slow. I was too lazy to interlace it. Oh I wonder if interlacing the AVX2 code would make it faster. I should've made a few versions. Anyway if you have numbers from JtR (--format=bf) or Hashcat (-m 3200). It's doing $2y$05$...
On Windows 8 x64 with the 4770k at 4.1 GHz

AVX2: 8*128 took 1.1789 sec (868.6 h/s) 0
Normal: 512 took 1.1072 sec (462.4 h/s) 0
Windows 8 x64 4770k @4.1GHz

hashcat-0.46>hashcat-cli64.exe -m 3200 -a 3 -n 1 m3200.txt -1 ?l?u?d?s ?1?1?1?1?1?1?1

Input.Mode: Mask (?1?1?1?1?1?1)
Index.....: 0/1 (segment), 735091890625 (words), 0 (bytes)
Recovered.: 0/1 hashes, 0/1 salts
Speed/sec.: - plains, 1.17k words
Progress..: 23071746956/735091890625 (3.14%)
Running...: 00:00:00:07
Estimated.: 133:19:08:08

hashcat-0.46>hashcat-cli64.exe -m 3200 -a 3 m3200.txt -1 ?l?u?d?s ?1?1?1?1?1?1?1

Input.Mode: Mask (?1?1?1?1?1?1)
Index.....: 0/1 (segment), 735091890625 (words), 0 (bytes)
Recovered.: 0/1 hashes, 0/1 salts
Speed/sec.: - plains, 7.28k words
Progress..: 23073276304/735091890625 (3.14%)
Running...: 00:00:03:31
Estimated.: 35:20:15:21
Man and I was hoping for a nice speed increase. If you compare the speed to Hashcat it'll probably be on par (hashcat -m 3200 -n 1 -a 3 bcrypt.txt ?a?a?a?a?a). Hashcat might be faster since I just realized that mine is just barely too big to fit in L1 cache.

I'm going to guess that the scatter/gather instructions are really there just so you don't need to move between general and vector registers. Which is where all the benefit of those instructions are coming from. But this doesn't help if that's all you are doing.

Damn you beat me. Also you should try 32 bit Hashcat it might be faster.
32 bit is slower
-n 1 : 1.13k
without -n: 7.07k