How to Build a Password Cracker with NVidia GTX 1080TI & GTX 1070

Kent Ickler //

The Task

  • Update our in-house password cracking/hashing capabilities
  • Purchase a new cracking machine
  • Update the old cracking system
  • Stay within budget

Buy The Things:

Total for new password cracking machine
$5110

A Few Quick Lessons

The CPU cooler doesn’t actually clear the case cover. This was OK because we don’t want to suffocate the GPU’s and I hadn’t planned on placing the cover on the unit anyway.  The case was specifically chosen as it was rack-mount with enough room for the motherboard to fit and properly support the four dual slot GPUs. A CPU cooler more fitting might be something like this: https://www.newegg.com/Product/Product.aspx?Item=9SIA1K65BT9403

The motherboard selected was chosen in part for its support on the M.2 SSD support. Unfortunately, it doesn’t support SATA M.2, which was what I had mistakenly purchased. It does support PCIE M.2 for those looking. In the build list above, you will note I have replaced the data storage with a Samsung 500GB SSD. You’ll also notice in the photos that I have temporarily used a PNY SSD that we had “laying around”.

The Build

Piecing the equipment together is pretty straight forward. The most complicated part was mounting the CPU cooler/fan. The cooler supports multiple CPU formats and consequently comes with multiple mounts that must be manipulated in such a way to work with the LGA 2011-v3 socket.

Motherboard

Installing Motherboard in case  Note M.2 SATA installed (see above)

CPU & CPU Fan fitment, note does not clear case top.

Add Power Supply and Ram

GPU’s to install

GPU Installed, ready to be cabled

GPU’s cabled and readied.  Note PNY SSD (see above)

The Software

Nvidia and Linux are happy together!

  • Ubuntu 16.04 LTS server distro with full disk encryption 
  • Nvidia drivers directly from the Nvidia website 
  • Hashcat 3.5 

This took about 15-30 minutes. The Nvidia driver’s worked great on Ubuntu 16.04 and I didn’t have any driver-headaches getting Hashcat to run.

The Older Brother

Updating an older cracking-machine

 We were under budget and used the excess funds to buy GPU’s to replace our old password cracking machine’s water-cooled AMD 290x’s. We chose to replace those 4 GPUs with Nvidia GTX 1070 Founders Edition.

The Benchmark Numbers

The result of this project was a new password cracking machine capable of over 208GH/sec NTLM and a refurbished machine capable of an other 119 GH/sec NTLM.

Combined, our password cracking/hashing capability just topped 327GH/sec for NTLM hashes. That’s 327,000,000,000 password attempts per second.
Not bad if comparing our investment to the $21,000 Brutalis that has been seeing 334 GH/sec.

The Benchmark Previews

During the design and product selection, I was a bit annoyed by the lack of public benchmarks for the GTX 1080TI and GTX 1070 cards. Our next post will be exclusively listing our benchmarks from our two most powerful cracking rigs.

Full Benchmarks on the second part of this blog post!

OpenCL Platform #1: NVIDIA Corporation
======================================
* Device #1: GeForce GTX 1080 Ti, 2793/11172 MB allocatable, 28MCU
* Device #2: GeForce GTX 1080 Ti, 2793/11172 MB allocatable, 28MCU
* Device #3: GeForce GTX 1080 Ti, 2793/11172 MB allocatable, 28MCU
* Device #4: GeForce GTX 1080 Ti, 2792/11169 MB allocatable, 28MCU

Hashtype: MD4

Speed.Dev.#1.....: 53850.2 MH/s (69.76ms)
Speed.Dev.#2.....: 54047.7 MH/s (69.51ms)
Speed.Dev.#3.....: 52955.5 MH/s (70.94ms)
Speed.Dev.#4.....: 53750.1 MH/s (69.86ms)
Speed.Dev.#*.....:   214.6 GH/s

Hashtype: MD5

Speed.Dev.#1.....: 31103.4 MH/s (60.39ms)
Speed.Dev.#2.....: 31676.5 MH/s (59.26ms)
Speed.Dev.#3.....: 30600.9 MH/s (61.33ms)
Speed.Dev.#4.....: 31198.4 MH/s (60.20ms)
Speed.Dev.#*.....:   124.6 GH/s

OpenCL Platform #1: NVIDIA Corporation
======================================
* Device #1: GeForce GTX 1070, 2028/8112 MB allocatable, 15MCU
* Device #2: GeForce GTX 1070, 2028/8114 MB allocatable, 15MCU
* Device #3: GeForce GTX 1070, 2028/8114 MB allocatable, 15MCU
* Device #4: GeForce GTX 1070, 2028/8114 MB allocatable, 15MCU

Hashtype: MD4

Speed.Dev.#1.....: 33622.2 MH/s (59.85ms)
Speed.Dev.#2.....: 32953.6 MH/s (61.07ms)
Speed.Dev.#3.....: 33108.6 MH/s (60.78ms)
Speed.Dev.#4.....: 34089.1 MH/s (59.02ms)
Speed.Dev.#*.....:   133.8 GH/s

Hashtype: MD5

Speed.Dev.#1.....: 18534.9 MH/s (54.28ms)
Speed.Dev.#2.....: 17880.8 MH/s (55.68ms)
Speed.Dev.#3.....: 18188.7 MH/s (55.32ms)
Speed.Dev.#4.....: 18401.1 MH/s (54.66ms)
Speed.Dev.#*.....: 73005.5 MH/s

Top: Nvidia 1080TIs password cracker

Bottom: Nvidia GTX 1070 password cracker

The Links

Hashcat: https://hashcat.net/

NVidia Drivers: http://www.nvidia.com/Download/index.aspx

Ubuntu 16.04: https://www.ubuntu.com/download/server

Benchmarks: Here!



Want to learn more mad skills from the person who wrote this blog?

Check out these classes from Kent and Jordan:

Applied Purple Teaming

Defending the Enterprise

Available live/virtual and on-demand!