Pipe is slow
#1
Dear atom, in https://hashcat.net/trac/ticket/604#comment:2 you recomend use pipe, but this mode is slow.

Example:

..\hashcat-0.49\hashcat-cliAVX2.exe -r bf.rule "D:\AIR\wordlist\concentrate" --stdout | cudaHashcat64.exe -a 0 -m 2500 --session=s6 -w 2 --gpu-temp-abort=90 --gpu-temp-retain=90 -o "D:\keys\hash_found.txt" --outfile-format=3 D:\in.hccap

Quote:cudaHashcat v1.33 starting...

Device #1: GeForce GTX 980, 4096MB, 1278Mhz, 16MCU
Device #2: GeForce GTX 970, 4096MB, 1253Mhz, 13MCU
Device #3: GeForce GTX 970, 4096MB, 1304Mhz, 13MCU

Hashes: 1 hashes; 1 unique digests, 1 unique salts
Bitmaps: 8 bits, 256 entries, 0x000000ff mask, 1024 bytes
Rules: 1
Applicable Optimizers:
* Zero-Byte
* Single-Hash
* Single-Salt
Watchdog: Temperature abort trigger set to 90c
Watchdog: Temperature retain trigger set to 90c
Device #1: Kernel ./kernels/4318/m02500.sm_52.64.ptx
Device #1: Kernel ./kernels/4318/amp_a0_v1.64.ptx
Device #2: Kernel ./kernels/4318/m02500.sm_52.64.ptx
Device #2: Kernel ./kernels/4318/amp_a0_v1.64.ptx
Device #3: Kernel ./kernels/4318/m02500.sm_52.64.ptx
Device #3: Kernel ./kernels/4318/amp_a0_v1.64.ptx

Starting attack in stdin mode...

Session.Name...: s6
Status.........: Running
Input.Mode.....: Pipe
Hash.Target....: ********* (********* <-> *********)
Hash.Type......: WPA/WPA2
Time.Started...: Sat Mar 14 18:28:04 2015 (5 mins, 22 secs)
Speed.GPU.#1...: 110.6 kH/s
Speed.GPU.#2...: 89863 H/s
Speed.GPU.#3...: 89887 H/s
Speed.GPU.#*...: 290.4 kH/s
Recovered......: 0/1 (0.00%) Digests, 0/1 (0.00%) Salts
Progress.......: 92037170
Skipped........: 0
Rejected.......: 2221200
HWMon.GPU.#1...: 99% Util, 76c Temp, N/A Fan
HWMon.GPU.#2...: 86% Util, 65c Temp, N/A Fan
HWMon.GPU.#3...: 51% Util, 76c Temp, N/A Fan

.png   1.png (Size: 72.29 KB / Downloads: 4)

Classic wordlist example:

cudaHashcat64.exe -a 0 -m 2500 --session=s4 -w 2 --gpu-temp-abort=90 --gpu-temp-retain=90 -o "D:\keys\hash_found.txt" --outfile-format=3 D:\in.hccap "D:\AIR\wordlist\BIG"

Quote:cudaHashcat v1.33 starting...

Device #1: GeForce GTX 980, 4096MB, 1278Mhz, 16MCU
Device #2: GeForce GTX 970, 4096MB, 1253Mhz, 13MCU
Device #3: GeForce GTX 970, 4096MB, 1304Mhz, 13MCU

Hashes: 1 hashes; 1 unique digests, 1 unique salts
Bitmaps: 8 bits, 256 entries, 0x000000ff mask, 1024 bytes
Rules: 1
Applicable Optimizers:
* Zero-Byte
* Single-Hash
* Single-Salt
Watchdog: Temperature abort trigger set to 90c
Watchdog: Temperature retain trigger set to 90c
Device #1: Kernel ./kernels/4318/m02500.sm_52.64.ptx
Device #1: Kernel ./kernels/4318/amp_a0_v1.64.ptx
Device #2: Kernel ./kernels/4318/m02500.sm_52.64.ptx
Device #2: Kernel ./kernels/4318/amp_a0_v1.64.ptx
Device #3: Kernel ./kernels/4318/m02500.sm_52.64.ptx
Device #3: Kernel ./kernels/4318/amp_a0_v1.64.ptx

Cache-hit dictionary stats D:\wordlist\BIG/08: 4497719949 bytes, 499746661 w
ords, 499746661 keyspace

Session.Name...: s4
Status.........: Running
Input.Mode.....: File (D:\wordlist\BIG/08)
Hash.Target....: ********* (********* <-> *********)
Hash.Type......: WPA/WPA2
Time.Started...: Sat Mar 14 16:51:12 2015 (23 secs)
Time.Estimated.: Sat Mar 14 17:32:52 2015 (41 mins, 15 secs)
Speed.GPU.#1...: 156.6 kH/s
Speed.GPU.#2...: 131.5 kH/s
Speed.GPU.#3...: 137.1 kH/s
Speed.GPU.#*...: 425.2 kH/s
Recovered......: 0/1 (0.00%) Digests, 0/1 (0.00%) Salts
Progress.......: 4665344/499746661 (0.93%)
Skipped........: 0/4665344 (0.00%)
Rejected.......: 0/4665344 (0.00%)
Restore.Point..: 4605952/499746661 (0.92%)
HWMon.GPU.#1...: 96% Util, 70c Temp, N/A Fan
HWMon.GPU.#2...: 93% Util, 73c Temp, N/A Fan
HWMon.GPU.#3...: 95% Util, 72c Temp, N/A Fan

I use CPU: Core i7-4820k, 32GB quad channel DDR3.
All videocard install in PCI-E x16 Gen3
It not to be slow.

How i can get full speed with pipe?
#2
You are using two different methods and two wordlists in your examples. AS "BIG" maybe much bigger than "concentrate"? Clearly in your first there is not enough work to do for the GPUs. But if the rule from your first example is too small, GPUs are starving for load.
#3
Now "concentrate"+rules generate 2 316 340 000 password for hashcat via pipe. I think is sufficient load for Multi-GPU. (few hours to compute)

This 2 different method is contains absolutely same words. In application 2 i generate wordlist by this rule manually.

I want use much bigger rules. Since the resulting dictionary may take 200-300 GB. Сreation, sorting and regularly update this, will require huge resources, it does not make sense.
Therefore I want to use rules. But pipe speed is slow. Sad
#4
use the rule set directly on the gpu then, no need for pipe with this attack mode
#5
The Mechanic - please read first line on my first message. (Atom recommend "use pipe")
#6
i did, you were using mode a6 in that, both examples here are of mode a0, piping is going to be slower than generating on gpu
#7
If i use -a 6, hashcat utilize only one GPU. About this i write in "trac". Is to much slow vs pipe. (Speed.GPU.#*...: 150.0 kH/s)
#8
is there a reason your rules file wont work on the gpu ?
#9
Dear Mechanic, unfortunately rules file + wordlist do not work on multi-gpu Smile

Example:
cudaHashcat64.exe -a 0 -m 2500 --session=s6 -r bf.rule -w 2 --gpu-temp-abort=90 --gpu-temp-retain=90 -o "D:\keys\hash_found.txt" --outfile-format=3 D:\keys\step6\out80.hccap "D:\wordlist\concentrate"
Quote:Session.Name...: s6
Status.........: Running
Rules.Type.....: File (bf.rule)
Input.Mode.....: File (D:\wordlist\concentrate/06)
Hash.Target....: ******* (********* <-> *********)
Hash.Type......: WPA/WPA2
Time.Started...: Sat Mar 14 21:49:16 2015 (9 secs)
Time.Estimated.: Sat Mar 14 22:06:12 2015 (16 mins, 42 secs)
Speed.GPU.#1...: 0 H/s
Speed.GPU.#2...: 115.8 kH/s
Speed.GPU.#3...: 0 H/s
Speed.GPU.#*...: 115.8 kH/s
Recovered......: 0/1 (0.00%) Digests, 0/1 (0.00%) Salts
Progress.......: 1142790/115421790 (0.99%)
Skipped........: 0/1142790 (0.00%)
Rejected.......: 0/1142790 (0.00%)
Restore.Point..: 0/10389 (0.00%)
HWMon.GPU.#1...: 0% Util, 60c Temp, N/A Fan
HWMon.GPU.#2...: 0% Util, 55c Temp, N/A Fan
HWMon.GPU.#3...: 95% Util, 72c Temp, N/A Fan
#10
it works on my setup, have you tried changing the workload to w3 or to just the basics cudaHashcat64.exe -a 0 -m 2500 --session=s6 -r bf.rule -o "D:\keys\hash_found.txt" --outfile-format=3 D:\keys\step6\out80.hccap "D:\wordlist\concentrate"