Drop performance md5 salted

[marked_d]

Hi there,

Configuration:
Ubuntu 10.10 64 bits
Forceware 280.x
Nvidia GTX 580 (3x)
ocl

I have a drop of performance when I try salted md5 hashes (-m 2) compared to md5 (-m 0):

Hash.Type....: MD5
Mode.Left....: Mask '?1?1?1' (238328)
Mode.Right...: Mask '?1?1?1' (238328)
Speed.GPU*...: 2581.4M/s

vs

Hash.Type....: md5($salt.$pass)
Mode.Left....: Mask '?1?1?1' (238328)
Mode.Right...: Mask '?1?1?1' (238328)
Speed.GPU*...: 35.9M/s

(./cudaHashcat64.bin -m 2 -d 1,2,3 -o /tmp/tmp --gpu-loops=1024 hashes -1 ?l?u?d ?1?1?1 ?1?1?1) on a list of 64 hashes.

Strange also that the % indicate:
HW.Monitor.#1: 0% GPU, 71c Temp
HW.Monitor.#2: 0% GPU, 70c Temp
HW.Monitor.#3: 0% GPU, 48c Temp

(for both type of hashes)

Thanks in advance !
M.

[atom]

thats the purpose of salting hashes, it makes multihash impossible. 64 * 39.6 = 2297,6 and you have to subtract some performance for the salt handling.. so everything works perfect

[marked_d]

thats the purpose of salting hashes, it makes multihash impossible. 64 * 39.6 = 2297,6 and you have to subtract some performance for the salt handling.. so everything works perfect

Hi Atom,

Thanks for you reply, but I think salt is to defeat precomputing attack not bruteforce ?
Also there is not so much difference on your performance table (on http://hashcat.net/oclhashcat/) less than 50%.

Anyway, it was my mistake I think I mixed the type of hashes I had (not md5 salted but Unix md5).

Cheers.

[atom]

a salt is much more efficient to defeat multihash than to defeat precomputing. the performance table is correct. they all show single hash cracks. the loss you see is about the salt handling. if you want unix md5 its called md5crypt -m 500 (plus version has it).