CudaHashcat 1.37 0H/s with Bcrypt
#1
I noticed that starting in version 1.37 I am getting 0H/s when attempting to crack Bcrypt hashes (not the case with any other algorithm) with Cuda Hashcat on a Nvidia 660 Ti. Bellow is a sample output and the args passed to Hashcat 

./cudaHashcat64.bin -m 3200 -a 0 --rules ./rules/generated2.rule -o am-out  /path/to/hashest.txt /path/to/wordlists/wordlist.txt

[s]tatus [p]ause [r]esume [b]ypass [q]uit => s

Session.Name...: cudaHashcat
Status.........: Running
Rules.Type.....: File (./rules/generated2.rule)
Input.Mode.....: File ( /path/to/hashest.txt )
Hash.Target....: File ( /path/to/hashest.txt )
Hash.Type......: bcrypt, Blowfish(OpenBSD)
Time.Started...: Mon Aug 24 14:37:58 2015 (17 secs)
Time.Estimated.: 0 secs
Speed.GPU.#1...:        0 H/s
Recovered......: 0/1995 (0.00%) Digests, 0/1995 (0.00%) Salts
Progress.......: 2/130744320 (0.00%)
Rejected.......: 0/2 (0.00%)
Restore.Point..: 0/1 (0.00%)
HWMon.GPU.#1...: -1% Util, 50c Temp, 33% Fan

[s]tatus [p]ause [r]esume [b]ypass [q]uit => s
#2
Looks like you've tested 2 candidates in 17 seconds, so your speed is ~ 0.11 H/s.

This is totally expected.

1. bcrypt is very slow, and depending on the cost value (uknown from above output), it can be horrifically slow.
2. The more unique salts you have, the slower it is (you have 1995 salts, that's a lot.)
3. Your GPU is slow.
4. For most CPUs and GPUs, bcrypt is faster on CPU than GPU. Given you do not have a high-end GPU, your CPU will likely be faster.
#3
Ok, thanks, did not realize how much slower Bcrypt is.
#4
There is a reason we tell people to use bcrypt. And, well, you're looking at that reason.