GPU cracking is too slow
#1
Hi,
I have an Alienware 18 which has two GeForce GTX 770M GPUs

I tried to test crack an MD5 hash and the speed I was able to achieve was 2163 Khz /s even though the speed of my card is 797 Mhz I also looked at some videos on hashcat and read about it and mostly everyone is able to achieve a lot faster numbers. In fact I noticed my regular CPU is faster than both my GPUs put together !!

Any advice?
Code:
khawasli@khawasli-pc ~/programs/oclHashcat-1.01 $ ./cudaHashcat64.bin -m 0 /home/khawasli/hashes/md5Hash /home/khawasli/hashes/all.lst
cudaHashcat v1.01 starting...

Hashes: 1 total, 1 unique salts, 1 unique digests
Bitmaps: 8 bits, 256 entries, 0x000000ff mask, 1024 bytes
Rules: 1
Applicable Optimizers:
* Zero-Byte
* Precompute-Init
* Precompute-Merkle-Demgard
* Meet-In-The-Middle
* Early-Skip
* Not-Salted
* Not-Iterated
* Single-Hash
* Single-Salt
* Scalar-Mode
* Raw-Hash
Watchdog: Temperature abort trigger set to 90c
Watchdog: Temperature retain trigger set to 80c
Device #1: GeForce GTX 770M, 3071MB, 797Mhz, 5MCU
Device #1: WARNING! Kernel exec timeout is not disabled, it might cause you errors of code 702
Device #2: GeForce GTX 770M, 3071MB, 797Mhz, 5MCU
Device #1: Kernel ./kernels/4318/m0000_a0.sm_30.64.ptx
Device #1: Kernel ./kernels/4318/bzero.64.ptx
Device #2: Kernel ./kernels/4318/m0000_a0.sm_30.64.ptx
Device #2: Kernel ./kernels/4318/bzero.64.ptx

Cache-hit dictionary stats /home/khawasli/hashes/all.lst: 47308224 bytes, 3917186 words, 3917186 keyspace

c5799bd58590d877ae5ca2ad9799d992:fdsjiojoigsdgfd
                                            
Session.Name...: cudaHashcat
Status.........: Cracked
Input.Mode.....: File (/home/khawasli/hashes/all.lst)
Hash.Target....: c5799bd58590d877ae5ca2ad9799d992
Hash.Type......: MD5
Time.Started...: Fri Apr 18 07:32:50 2014 (2 secs)
Speed.GPU.#1...:  2163.1 kH/s
Speed.GPU.#2...:  2187.0 kH/s
Speed.GPU.#*...:  4350.1 kH/s
Recovered......: 1/1 (100.00%) Digests, 1/1 (100.00%) Salts
Progress.......: 3917186/3917186 (100.00%)
Rejected.......: 353/3917186 (0.01%)
HWMon.GPU.#1...: -1% Util, 56c Temp, 100% Fan
HWMon.GPU.#2...: -1% Util, 53c Temp, 100% Fan

Started: Fri Apr 18 07:32:50 2014
Stopped: Fri Apr 18 07:32:52 2014
#2
nvidia gpus are slow, mobile gpus are slow, nvidia mobile gpus are the slowest.

that said, the attack you are running is not well-suited for gpus, so you're not getting full acceleration. straight dictionary attacks with fast hashes are better suited to cpus, not gpus. to achieve full acceleration with a dictionary attack, you need to add some rules.

in summary, your gpus are slow and your attack is suboptimal.
#3
(04-18-2014, 08:25 AM)epixoip Wrote: nvidia gpus are slow, mobile gpus are slow, nvidia mobile gpus are the slowest.

that said, the attack you are running is not well-suited for gpus, so you're not getting full acceleration. straight dictionary attacks with fast hashes are better suited to cpus, not gpus. to achieve full acceleration with a dictionary attack, you need to add some rules.

in summary, your gpus are slow and your attack is suboptimal.

Thank you for your reply. Do you mind telling me which attacks are suited with GPUs ? And why are some attacks not well suited with GPUs? And which GPU card would you recommend?
#4
gpus work on massively parallel problems, so you can only achieve full acceleration on gpu if you give the gpu enough work to do. a straight dictionary attack does not give the gpu enough work. in order to achieve full acceleration, you have to run an attack with a large enough keyspace. try adding '-r rules/d3ad0ne.rule' to the end of your command.

Radeon 290X is the best card at the moment.
#5
(04-18-2014, 09:32 AM)epixoip Wrote: gpus work on massively parallel problems, so you can only achieve full acceleration on gpu if you give the gpu enough work to do. a straight dictionary attack does not give the gpu enough work. in order to achieve full acceleration, you have to run an attack with a large enough keyspace. try adding '-r rules/d3ad0ne.rule' to the end of your command.

Radeon 290X is the best card at the moment.

Thanks again for your reply. Is there a place where I can read a description on that rule? I googled it and could find much details about it.

What about regular bruteforcing (without a dictionary) do GPUs work well with that type of attack?
#6
(04-18-2014, 09:55 AM)khawasli Wrote:
(04-18-2014, 09:32 AM)epixoip Wrote: gpus work on massively parallel problems, so you can only achieve full acceleration on gpu if you give the gpu enough work to do. a straight dictionary attack does not give the gpu enough work. in order to achieve full acceleration, you have to run an attack with a large enough keyspace. try adding '-r rules/d3ad0ne.rule' to the end of your command.

Radeon 290X is the best card at the moment.

Thanks again for your reply. Is there a place where I can read a description on that rule? I googled it and could find much details about it.

What about regular bruteforcing (without a dictionary) do GPUs work well with that type of attack?

Go to the wiki to learn more about rule attacks and all other kinds.

And yes, bruteforcing with GPU will use almost full work load and achieve the fastest speed.