true, but it's probably easier said then done...
AND it means that those tables will be fairly useless for people without a fast GPU.

i guess when we start using a new reduction function (the one Sc00bz has made), it would be a good time to move towards GPU generated rt's... It would be nice to join forces with the distributed GPU cracker that BarsWF is developing. It would be nice to have large rainbow tables for quick lookup of important hashes, and to crack less important hashes (or at least with less priority) in bulk amounts with the GPU brute force.

yes it would. BUT we do not know how to do a gpu client. if you know it and can handle CUDA then we'd be happy if you made us a client otherwise you will have to wait for things to develop like everyone else has to.

y that ? People without a compatible GPU generate tables with their CPU - like they do it now.
The only problem we get is : Bigger tables ...more traffic --> needs more time & a fast connection


Sorry i don't know a program language like c++ or java. Only know html and a lil bit php So i can't help you. But i think there are a lot of guys who know how to program a client. I just found a link yesterday on google code. Seems interesting ! :
http://code.google.com/p/dist-hash/

snooopy

because these tables are big and using them with cpu would take very long. producing the tables with gpu is only the first step.

edit : disthash: bars is planing the same but with gpu. checkout http://3.14.by/forum

yes that's right... but i think looking up is still fast enough ....and all the new computers ( even cheap one ) got graphic cards which support CUDA.
The links seems interesting .. They should cooperate

are you sure http://code.google.com/p/dist-hash/ is still a living project ? because there are many dead ones on google.

and btw CUDA is only supported by 8000 series up.
and no it would not be fast enough. if you look up multiple hashes now with rcracki thats already slow. (taking hours for 500 hashes) and those tables are not to big yet.

hm ...yes but CUDA able graphic cards will get spread soon....
and is it normal that you look up 500 hashes at once ? ^^ from where do u get all these hashes

the project is still alive ...have a look in issues ...i posted there the link
http://code.google.com/p/dist-hash/issues/detail?id=1

not the point but thanks for the link

You know if you have 500 hashes it's about where it's faster to do brute forcing than rainbow tables. I don't know if there is a program that does multiple hashes on a video card, but if it did exists and you got about 500 million on a nice video card or on multiple video cards. It would be about 2.5 hrs (key space of 2^42 the key space of the average rainbow table here). Well if there was a fast rainbow table video card program it would take a lot less time than "hours."

i recently tried some 10k md5 hashes on CUDA_nV brute forcer, did about 300-325 million hashes/s
i could even brute force loweralpha numeric 1-9 in about 4 days

Brute forcing is faster for short lengths. Run hashes through a brute forcer before throwing them at tables. Use tables for things brute forcing can't quickly do. And a good graphics card and good code should be able to run around 700M checks per second or more for a simple brute force algorithm.

Can't we just add some md5/ntlm so the project comes back to life? We're wasting a lot of time now..

i think alot of time is wasted on the current cracking queue...

Going through the whole cracking queue takes an awful bit of time. I bet most of these passwords that were submitted a short while ago could have been broken by a bruteforce / dictionary attack.

Is there an API such that those with fairly rapid hash cracking or brute force systems can interface with the queue and deal with some of the smaller passwords? It's easy enough to verify the results coming back - just hash them.