[Darkness Productions]

Hey everybody, I'm trying to gather MD5CRK benchmark information. If you'd like to help, please do. If not, no worries.

I've been told that G4/G5s are the best processor for this project, so I'd be especially interested in numbers for those, however, I'll take numbers for anything you have.

If you want to help, make sure you have client version 1.0.7 (http://md5crk.com?sec=download). Start the client, and when it finishes benchmarking the cores, go to http://external.dp.cx/md5crk/. Select your team name, then your username. Point to your md5_test.xml file, and upload it. Once finished, you're done.

If you're not currently participating in the project, then select any team/user you'd like, as it's really overall irrelevant.

If you have a client version prior to 1.0.7, please download 1.0.7 (using the above link). Extract it to a folder, then open a command prompt (DOS box for windows people, shell for everyone else)... Run the command "md5crk -test" (for Windows, "./md5crk -test" for *Nix). Then, visit the above url, and follow the same steps.


Thanks in advance! Please know that your information is kept safe and somewhat secure, and is only interesting to me and a handful of others.

Also, the more machines you're willing to run the benchmark for, the better my results. Unfortunately, at this time, the results are not being made public, but I do plan to make them so once I get enough data to make accurate predictions.

[reader50]

For the curious, this is another encryption cracking exercise. In this case, they are trying to 'break' the MD5 hash used for website security certificates. The project's explanation can be found here.

If someone became good at creating duplicate hashes, they could fake secure certificates for websites, and your browser would trust the faked certificates. That would not be so good, your browser would indicate it's the genuine website, so you can enter your username and passwords, etc.

In practice, MD5 is a 128-bit hash. To find a duplicate, you need to start with the secure hash of the banking site you wish to attack. Then you attempt up to 2^128 keys until you find one that yields a duplicate hash. In other words, this requires the same approach as RC5, bearing in mind that a successful crack would yield only one duplicate hash - and the bank may have been issued a new certificate in the 1000+ years since you started the cracking effort.

RC5-64 took several years. RC5-72 is in progress. If RC5-64 were to take one year with today's computers, cracking RC5-128 would require approx 10^20 years with today's computers. For reference, the universe has been around for ~1.3 x 10^10 years. If they'd started on RC5-128 right away, the universe would have to add 10 more zeros onto it's age before the problem got cracked.

The MD5CRK website claims MD5 has been considered insecure since 1998 - that would be around the time that the beige G3 desktops came out. Apparently, they tipped the scales with their superior cracking performance.

If our members are interested in this project, we will of course provide stats and support. We would have to do so for a very long time - I'll make sure my great grandkids continue our support down the road.

ps - for those who do not know, "Darkness Productions" is a Free-DC member who came from the Ars teams before Free-DC. He does some stats programming, and has not said anything bad about Macs that I know of. This is not the first time he has asked around for benchmarks - he has previously assembled benchmark reference pages for other projects.

[Welnic]

Here is the performance ranking on the last version of the client:
http://www.md5crk.com/stats/perf.php
You don't have to scroll down too far to see where my 2GHz G5 ended up.

I have been running the project with just one machine so I can experience the joy of not having to decide how many machines it gets. It gets one and one is the number that it gets.

[Darkness Productions]

Thanks for the summary reader50... didn't think about the fact that people here wouldn't have run it before.

For those that are interested, the preliminary results are available at http://external.dp.cx/md5crk/perf.php

Please take a look, and see if you can beat Welnic's G5!

[mikkyo]

It appears multiple threads/processors are not supported.

:-(

[Darkness Productions]

Not currently. Just one run copy of it, and post that benchmark. It's almost totall 1:1 if you add processors (HT, however, is not).

[reader50]

Say DP, why is your screen name truncated? If you hit the Register button early, I might be able to get it fixed for you.

[Lateralus]

My 1.33 is in second place.

[Lateralus]

And now my 1.33 @ 1.47 is in second place.

[mikkyo]

And now?

[Darkness Productions]

Originally posted by reader50:
Say DP, why is your screen name truncated? If you hit the Register button early, I might be able to get it fixed for you.

That's all the more text it would let me type in.... I wanted the rest, but it wouldn't let me. Se la vie.

[reader50]

After long hours of hard work, we managed to fix your screen name that IE truncated on the registration page. Our charge for all this labor is that you give up IE for a better browser.

This seems reasonable.

[Darkness Productions]

Thanks reader50!


Anyone else care to add benchmark information, be it for a G5, G4, G3, or any other processor type that they have a client for?

Any help is requested, so give it a shot! Can you beat the #1 G5?

ALL MY STAT ARE BELONG TO DARKNESS PRODUCTIONS!

[Scotttheking]

Originally posted by stibnite:
ALL MY STAT ARE BELONG TO DARKNESS PRODUCTIONS!

And you belong to me

I'm moving to Maryland!!!!!