Wordpress Hashes and Temperate Limit
#1
Hi,

I often face an issue while running an attack on Wordpress hashes and other slow hashes like MD5(Unix).

I run a dictionary attack against around 10 to 20 wordpress hashes and after a while, it reaches the temperature limit.

I understand that these hash types cause oclhashcat-plus to increase the GPU usage and also the memory usage, so is there a work around?

Code:
oclHashcat-plus v0.08 by atom starting...

Hashes: 20
Unique salts: 18
Unique digests: 18
Bitmaps: 8 bits, 256 entries, 0x000000ff mask, 1024 bytes
Rules: 1
GPU-Loops: 64
GPU-Accel: 400
Password lengths range: 1 - 15
Platform: AMD compatible platform found
Watchdog: Temperature limit set to 90c
Device #1: Cypress, 1024MB, 0Mhz, 20MCU
Device #1: Allocating 1203MB host-memory
Device #1: Kernel ./kernels/4098/m0400.Cypress.32.kernel (371944 bytes)

[s]tatus [p]ause [r]esume [q]uit => ERROR:
Temperature limit on GPU 1 reached, aborting...

What settings do I need to change?

My GPU: ATI Radeon HD 5870
Catalyst Version: 12.4
OS: Windows 7 Ultimate
#2
Whats different about them is they do many iterations of their result digest. You can control how many of them are calculated by specifying --gpu-loops. Try with --gpu-loops 1 first, see if it helps and then increase the value. Notice this will slower your performance!
#3
@atom: Thanks. I tried lowering the value of -n from 400 to 200 and it allowed my cracking session to run a little longer. I am now able to run a dictionary attack successfully with a dictionary of size ~500 Mb.

However, with a larger dictionary, it still fails.

Yes, these hashes are highly iterated, I understand. I would like to understand how the --gpu-loops parameter is related to the number of iterations used in the hashing algorithm.

For instance, MD5(Unix) has a fixed number of 1000 iterations. How would this relate to a --gpu-loops value of, let's say, 200.

Will the GPU compute 200 iterations in one go?

Also, is the parameter, --gpu-loops only related to number of iterations computed by the GPU? Because, I also use it with unsalted hashes such as MD5 and SHA-1.

Until now, I was using the -n option.

Now, I have changed it to:

--gpu-accel=80 --gpu-loop=8

let's see, how far the attack goes.

Test 1:

Dictionary size ~1Gb
Progress before temperature limit reached: 10 %
option used, -n 200

Test 2:

Same dictionary:

--gpu-accel=80 --gpu-loops=8

temperature limit reached within 5 seconds from the start of attack :O

Test 3:

Same dictionary.

--gpu-accel=80 --gpu-loops=1

The cracking session is running, lets see how far it goes.

**** Update ****

The attack is running good with --gpu-loops=1 but yes the speed has come down a lot.

From, 430k c/s to 58k c/s Sad

I hope, slow and steady wins the race Big Grin
#4
Quote:Now, I have changed it to:

--gpu-accel=80 --gpu-loop=8

let's see, how far the attack goes.

Test 1:

Dictionary size ~1Gb
Progress before temperature limit reached: 10 %
option used, -n 200
Just to make sure that it was just an error when you made in your post. You first set --gpu-accel=80 an use -n 200. I don't know what happens if you put both but they are the same switch. Use either --gpu-accel or -n.
#5
I know that its stupid question to ask people who does hash cracking things but - did you set manually your fan speed? Did you tried to leave case open while cracking? 5 second of run indicates bad gpu contact with heat spreader
#6
@mastercracker:

Please disregard "Test 1" in my post. In that Test, I did not use the fine tuning options (--gpu-accel) and (--gpu-loop).

Yes, we need to use either,

--gpu-accel and --gpu-loop together or specify -n value.

With --gpu-loop=1, attack runs good and for a longer duration without any problem. However, since the speed came down at least 8 times, I will try increasing this option and see what is the best combination.

It is still comparatively faster than what speed I would get on CPU ^^