Svarychevski Michail Aleksandrovich put up a version of BarsWF for ATI BROOK not long ago, but the results were less than spectacular. I recently updated to CCC 9.5 and wanted to give it another shot. Here are my hardware specs for my Vista 64 box:
CPUZ – Stock clock speeds.
cpuz BarsWF on ATI Cuda like MD5 bruteforce
GPUZ – Also stock speeds.
gpuz BarsWF on ATI Cuda like MD5 bruteforce

And here is the BarsWF run.
barswf BarsWF on ATI Cuda like MD5 bruteforce

This particular run is a demo of ideal circumstances. I know the passwords charset and length, so I specified it in the command line:

C:\sec\md5>BarsWF_Brook_x64.exe -h d19ff6fb09cafa03d51ecac250bf71a8 -c 0A -min_len 10

Here is the same run under less than ideal circumstances. I don’t know anything about the hash, so I will just run the defaults:
barswf 2 BarsWF on ATI Cuda like MD5 bruteforce

C:\sec\md5>BarsWF_Brook_x64.exe -h d19ff6fb09cafa03d51ecac250bf71a8 -c 0aA~

So, we can see that it is indeed quite fast when running the full charset, though it should be noted that the ETA is quite off. It will likely reset a few times before finding the hash. It’s great to know that ATI owners have a GPU bruteforce option now. I’d really like to see this tech developed further to include other hash methods as well.

**UPDATE**
Here’s a screen of its memory usage after about 5 hours of running.
taskmgr BarsWF on ATI Cuda like MD5 bruteforce
Looks like it’s using around 100MB RAM per hour at this point. I’ll continue to watch it and hopefully have it finish before start paging every thing (which is a must-reboot scenario with Vista for some reason).