A user on the hashcat channel confirmed that 4 6990 dual GPU cards can now utilize all their powers with the new catalyst 11.40 under linux.
Here is a log of the user:
gpu-01 x86_64 # ./oclHashcat-lite64.bin -1 ?l?d?s?u --pw-min 8 --runtime 60 -n 800 -m 0 ffffffffffffffffffffffffffffffff ?1?1?1?1?1?1?1?1?1
oclHashcat-lite v0.3 starting...
Platform: AMD compatible platform found
Watchdog: Temperature limit set to 90c
Monitor1..: 99% GPU, 82c Temp
Monitor2..: 99% GPU, 81c Temp
Monitor3..: 99% GPU, 81c Temp
Monitor4..: 99% GPU, 80c Temp
Monitor5..: 99% GPU, 73c Temp
Monitor6..: 99% GPU, 78c Temp
Monitor7..: 98% GPU, 81c Temp
Monitor8..: 98% GPU, 75c Temp
Credit go to user "login" for his beautiful rig: https://www.facebook.com/pages/TMTOdotORG/162988683720444?sk=wall
Yes, it's true, the new version of oclhashcat-lite is ready for release!
There are a few great improvements in the 0.04, to mention a few:
- improved SL3 kernel
- Oracle 11G
As you see, atom again put a great amount of work into the new version. The support of the above mentioned algorythm lift oclhashcat-lite on a new level. For instance the new SL3 kernel brings a significant performance improve and should make a lot of the most new users happy. And of course the support for DES(Unix) of which atom was bothered since the beginning is a huge milestone.
These are just the big new features, there are plenty of little features and bugfixes that would blow this post, please take a look at the changes, it's worth it!
There is a new feature in the v0.04 that looks tiny and easily could unintentionally be skipped by most of the users but it is really powerfull and my brain imediatly started thinking as i heared about it. The feature i am talking about is the outfile-watchdog. This tiny feature forces hashcat to check the outfile every x seconds (60 by default) for found hashes and aborts the attack if the hash is found.
So, this doesn't sound very powerful at first but let me show you my thoughts. In the past you were able to distribute hashcat over several nodes by splitting the masks or by splitting the hashlist. If you split the mask you came across the problem that if one of your nodes found the hash, the other nodes would still continue it's attack until the hit the end of their mask.
There are a few projects out there that have a central webservice distributing the hashes and partial masks, with that you are now able to build your own fully automated cluster if you add a daemon that continuously fetches the found hashes in addition to the hashlist and masks. Imagine a fully automated and flexible cloud that you can control through a webinterface.
Just with a click you can create as many nodes as you want, add a hashlist and start an attack within minutes. This would be mindblowing on amazon if they would have AMD GPUs instead of Teslas. In a few words, this feature makes distributed environments so much more attractive!
Cheers to the users that build their own low-budget rigs. It makes us and especially atom really proud to the those pictures of homegrown gpu farmes!
Beware of the cat!
As quoted in my last hashcat trailer "hashcat is a bitch that never sleeps" i'm glad to announce a atom is working on the oclhashcat-lite-0.04b!
Despite the release of the long expected version 0.02 lasts now only 5 days and still SL3 guys are flooding the channel and the forum with their attendance atom has decided to pull the trigger on a feature that oclhashcat users are awaiting for quite longer. DES(unix). It's the world first implementation of this algorithm on amd/ati GPU and indeed the world fasted!
Let's face the facts, these days for example the most popular cracker JTR is able to do about 3.5M/sec in DES(unix) on a slightly overclocked (2.66 GHz to 3.6 GHz) CPU. You can imagine that state of the art multi CPU machines with special compiled version of JTR can do much more and are much more expensive (one XEON MP X7550 will cost you about 2500 Eur, not to mention about the mainboard and ram) as you can see at JTR's wiki http://openwall.info/wiki/john/benchmarks.
Atom on the other side today showed us his FIRST example on erebus, you remember the box with 8x6970. His demonstration showed amazing results. Keep in mind that this is atom's first try without optimizations at all! I'm pretty courious how much he can optimize it.
Especially if you have multilple GPUs under linux you came across the issue to adjust the fan speed.
The trick is to change the DISPLAY to the device you want to alter.
Example for two devices:
aticonfig --pplib-cmd "set fanspeed 0 60"
aticonfig --pplib-cmd "set fanspeed 0 60"
Thanks to d3ad0ne!