How is it possible to enable your machines GPU to audit passwords? For example dictionary or bruteforce using password crackers like john etc.

I don't think you should look here for an answer to that.

1 members found this post helpful.

Actually that isn't the case. My taxonomy says this is not about Linux Security (so I'll move it to the Software forum RSN), it's not about cracking (he's not asking for help with that) but about software in general but then the OpenCL / ATI Streams / CUDA / KGPU-enabled variety. If unsure feel free to ask a moderator before turning your suspicions into posts that do not help the OP and do not contribute to the thread (LQR#4).

@OP: AFAIK there's very little CUDA-enabled SW at the moment and I haven't come across a version of JTR that is.

I cannot immediately see how the mathematical capabilities of a GPU would be useful, nor can I recall any situation where it has been applied.

I would, of course, be quite interested to see URL's about such a novel idea.

thanks unspawn, and even if it was able to enable it. What I hear is it burns out your GPU anyways so its dangerous to try it. From what I read it is possible to speed up a password auditing session quite significantly.

Nonsense. If your GPU burns because some software is actually using it then the cooling system is poorly designed and you should change your card to one from a manufacturer with higher quality.

I would think the main reason of this is because GPUs are designed as massive parallel systems that can try many passwords at the same time.

By the way, John the Ripper can make use of CUDA and OpenCL: http://openwall.info/wiki/john/GPU

improper code can also burn your GPU every GPU has its limits even the good ones. When you go over that thresh hold then... bye bye GPU. The best thing is to know what you have and know what you are working with... thanks for the link.

You can't use you GPUs capabilities more than 100%, no matter how poor your code is. If your cooling system can't handle the amount of heat generated when using your CPU 100% it is poorly designed and needs to be replaced. Simple as that.

yes but 100% trying to crack for a few days, weeks or even months? You need to have a monster cooling sys for that...
Have you yourself tried it for that long? Ive heard CPU's crash due to dust running on 60%... I mean logically speaking think about it. Months? 100%? I think good coder should at least use intervals of slowing down to keep things cooled and stable. 100% is a lot of work for CPU... and for GPU use something its not designed for? I feel sorry for that Processor man lol
I mean I am not making claims that it can't be done... I'm just telling you from an experience of Non GPU usage and how hot things get. I guess it could work with a super heat sink + fan.

Yes of course you can use a GPU as a compute engine with the right software. Password hashing calculations have been done on GPU's before.

As for ehat from sustained use; those who put their systems, both CPU and GPU, under extremely heavy loads for extended periods of time tend to use water cooling, or even liquid nitrogen, though for occasional activities a good air cooled system will be fine.

All you need is a high quality card.
The cooling system's only purpose is to cool down the system to a maximum allowed chip temperature, which is defined by the card manufacturer (the chip manufacturer only defines the temperature that should never be exceeded). Since a more effective cooling system (lower chip temperatures) is more expensive you will have higher chip temperatures which will shorten the lifetime of the card if you choose a cheaper card of the same class. You will most likely also get other components on the card with lower quality which may also shorten the lifetime. Those cheap cards are those that I call poorly designed, I have seen cards with a cooling system that was so undersized that the chip had to activate its self-protection after two minutes burn-in test (Furmark). Those cards should be used for nothing more than simple desktop work, but neither for games nor computing.
This does not happen with (more expensive) quality cards from well known brands, like ASUS, Gigabyte, Sapphire, EVGA, ... .

Current GPUs are GPGPUs (General Purpose Graphics Processing Units) that are designed to be used for general purpose computing via CUDA/STREAM/OpenCL. The professional compute accelerator cards used in super computers use the same chips.

May be you should also have a look at things like SETI@Home, Folding@Home or BOINC, many users let their systems work at 100% (CPU+GPU) 24/7 without damaging them. Of course you must have a good cooling concept for this, but a "monster cooling sys" is not needed.

1 members found this post helpful.

The idea that certain devices cannot run under 100% of load for sustained periods of time is a result of some manufacturers cutting corners on the cooling system. I don't just mean CPUs or GPUs, but also electric motors, car engines, amplifiers, etc. For ages those devices has been produced with cooling systems which use the thermal mass of the device to limit the temperature rise. If you use it too long, the device has warmed up all the mass and the temperature continues to rise. They hope that in normal use that point is never reached. Consumers now believe that this is normal.

Thermal design is very simple. If you device dissipates an certain amount of heat, you must transfer that heat to the environment. However you cannot transfer any heat before you have a temperature difference. So the device's temperature increases, heat flows to the environment and when the heat flow path has sufficiently low thermal resistance after some time a thermal equilibrum is reached. For a given environmental temperature the device temperature won't rise anymore.

There are cars which should not be driven at maximum power for hours and hours. That is consumer design. An industrial diesel engine can be operated at 100% power continuously for months. It is just a matter what you define as 100% and whether the cooling system is able to handle the generated heat.

Same with chips. If you define one 100% load as the situation where the heat generation is a maximum with everything running at the highest clock, no idle cycles or whatever, and you design a proper cooling system, the chip won't overheat and will perform until its expected age, even at 100% load.

jlinkels

1 members found this post helpful.