v62 is up
#21
OK, getting 1535M/sec with b67 using cat 12.6 so will update to cat 12.8 now.

UPDATE: With cat 12.8, results: 1538.7M/sec

Tested both with and without Crossfire.

Better, but no cigar Sad
Owner of: www.md5decrypter.co.uk
#22
b68 is up!

- added support for up to 128 GPU (we will need it with vcl)
- allowed more memory allocation on amd cards (you can load more hashes @multihash)
#23
(09-02-2012, 10:42 AM)blandyuk Wrote: OK, getting 1535M/sec with b67 using cat 12.6 so will update to cat 12.8 now.

UPDATE: With cat 12.8, results: 1538.7M/sec

Tested both with and without Crossfire.

Better, but no cigar Sad

try b68, make sure you use cat 12.8, double check you removed old .kernel files in kernels/4098/ in case you overwrite stuff
#24
For b67 on AMD, I get the following error on bruteforce:

clEnqueueNDRangeKernel() -54

Everything else works fine.
#25
on all kernel or just 3200 ?
#26
I always delete the "kernels" directory before extracting so it's totally clean. Tested b68, results same as b67 Sad sry dude. I always do clean installs of the cat drivers new and I have cat 12.8 currently.

Let me know when u need anymore testing.
Owner of: www.md5decrypter.co.uk
#27
(09-02-2012, 04:12 PM)atom Wrote: on all kernel or just 3200 ?
Yes, sorry. My bad posting. You are right, only on -m 3200.
#28
(09-02-2012, 07:06 PM)blandyuk Wrote: I always delete the "kernels" directory before extracting so it's totally clean. Tested b68, results same as b67 Sad sry dude. I always do clean installs of the cat drivers new and I have cat 12.8 currently.

Let me know when u need anymore testing.

I totally can not understand this. I tried on 5xxx, 6xxx and 7xxx and in all cases its much faster than 0.081. I tested on windows and linux.
#29
Guys, beside the 3200 issue, is there anything left for fix? I am preparing for release otherwise. Speak now! Smile
#30
System: Windows7 64-bit, nVidia, 304,79 beta, oclHashcat-plus-0.09b68.

oclHashcat-plus-0.08
-m0:
hc64p -n320 --gpu-loops=512 -a3 -1 ?l?u?d?s --runtime=120 ..\M\1mfakemd5.hash ?1?1?1?1?1?1?1?1
Speed........: 164.9M c/s Real, 164.8M c/s GPU
-m2811:
hc64p -n320 -m2811 -a3 -1 ?l?u?d?s --runtime=60 5kfakeipb.txt ?1?1?1?1?1?1?1?1
Speed........: 52526.1k c/s Real, 52576.8k c/s GPU

oclHashcat-plus-0.09b68
-m0:
hc64p -n320 --gpu-loops=512 -a3 --bf-min=8 --runtime=60 ..\M\1mfakemd5.hash ?a?a?a?a?a?a?a?a
Speed........: 104.4M c/s Real, 104.6M c/s GPU
-m2811:
hc64p -n320 -m2811 -a3 --bf-min=8 --runtime=60 5kfakeipb.txt ?a?a?a?a?a?a?a?a
Speed........: 62915.0k c/s Real, 63064.1k c/s GPU

IPB's speed is very nice compared to the previous version. That leaves MD5 questionable.

These algos might not be very important, but in case you want this to be perfect:
-m3100: When I use -a3, all is being cracked but if len(pass) > 8 it truncate the passes to 8.
-m3000: LM is behaving much better! BUT still:
-a0: does NOT crack any 6-character-long passwords!
-a6: Sometimes adds a char in the end of a cracked password. (Very similar to: https://hashcat.net/forum/thread-1314.html )

Going to re-test everything. Hopefully today.

All tested @ Windows7 64-bit, nVidia, 304,79 beta, oclHashcat-plus-0.09b68
Everything works flawlessly EXCEPT the two algos mentioned above (-m3000, -m3100).

Will try to do some speed\performance tests later.