Getting Too Hot ?
#1
Just thought I would share this with you all...

I have under-clocked my graphics card RAM and noticed a significant temperature reduction, with no cracking speed loss on WPA.

I went from 70 - 73 to 67 - 69 degrees on my HD5870.

Give it a try. Smile
#2
I did that for my HD 5870 too, but I noticed it was a lot slower than expected when doing dictionary attacks on MD5.

I think I heard correctly that GPU RAM:
1.) Doesn't matter for brute-force
2.) Does matter for dictionary

So adjust RAM clock accordingly.
#3
(01-02-2012, 06:51 PM)chort Wrote: I did that for my HD 5870 too, but I noticed it was a lot slower than expected when doing dictionary attacks on MD5.

My post was regarding WPA as mentioned above. Meaning hccap files. Smile

(01-02-2012, 06:51 PM)chort Wrote: I think I heard correctly that GPU RAM:
1.) Doesn't matter for brute-force
2.) Does matter for dictionary

So adjust RAM clock accordingly.

I understand it is more to do with the speed of the algorithm rather than dictionary or bruteforce. I may be wrong though.

WPA (hccap) is a slow algorithm so there is little or no reduction in performance when I reduce my RAM speed.

Almost forgot to mention, now my temperatures are down I have been able to overclock more aggressively, perhaps this may be possible for you chort when cracking MD5 ?


#4
I have both the shader clock and RAM clock OC'd to max for these MD5s Smile
#5
(01-04-2012, 10:48 PM)chort Wrote: I have both the shader clock and RAM clock OC'd to max for these MD5s Smile

Good grief !!! Big Grin Ha ha !

I only usually work on WPA and Using the HD5870 I can get an average of 95000 c/s GPU. I am not sure what the c/s real figure is but it is always slightly lower.

Do you know what c/s GPU and Real c/s are or mean ?
#6
REAL is the speed result which comes out after dividing all the finished words from zero to current by the start of the cracking phase at all.

GPU is 1. a current view, that means is just divides the time for the last 32 kernel launches by the number of finished words from the last 32 kernel launches and 2. the time without the copy overhead. that means, without the time for parsing the wordlists and then the pci-express overhead for copying the wordlist data to the gpu memory.

This might be interessting for Wiki, too


#7
Big Grin

Oh dear, I am always reminded of a funny phrase when I read your answers !! Big Grin

I may be no wiser, but I have been better informed !! Smile

However I think I know what you mean, thanks for the reply.

Can you make sure I have this right please ?

Thanks.
#8
I changed the last sentence. Please review. If its ok, then its OK. But i would not do it as a form as FAQ, just add it somewhere to the oclHashcat-plus article.
#9
OK all done Smile

I have also removed the FAQ page.
#10
ok, cool, thanks