Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Mon, 30 Jun 2014 21:53:55 -0500
From: Richard Miles <>
Subject: Re: Question about GPU and CPU.

Hi magnum,

Thanks for your comments.

On Fri, Jun 27, 2014 at 7:47 PM, magnum <> wrote:

> It is the "only" way, Solar just didn't know about it. Latest Hashcat can
> do it on its own, but the previous version needed od6config. We are not
> currently planning to add any such things to Jumbo, just keep using
> od6config.
Great, thanks for follow-up.

>  Magnum, do you also use it, correct? If yes, do you use --autofix? Or do
>> you use any other config? Can you share?
> I would use --autofix. It is the way to go if you want maximum performance
> without overclocking. Do not bother overclocking until you master
> everything else.

Nice. Just for curious, do you currently overclock? How much and what the
gain you have pay off?

>  Also is the fan very loud on your case too?
> "My case" is a laptop. I do not own a 290X but yes, it sounds like a
> chinese vacuum cleaner if you set it to 100%. You can use a lower fan
> setting but it may lead to worse performance (card will throttle clocks
> instead, to keep temps down).

Interesting. Is there any chance for you show me a benchmark of your R9

I tried just with hashcat until the moment, because with JtR it's not
detecting Opencl. I checked my bechmark and I think it's not that great:

Device #1: Hawaii, 3072MB, 1000Mhz, 44MCU

Hashtype: MD4
Workload: 1024 loops, 256 accel

Speed.GPU.#1.: 23188.8 MH/s

Hashtype: MD5
Workload: 1024 loops, 256 accel

Speed.GPU.#1.: 11552.2 MH/s

Hashtype: SHA1
Workload: 1024 loops, 256 accel

Speed.GPU.#1.:  3698.8 MH/s

Hashtype: SHA256
Workload: 512 loops, 256 accel

Speed.GPU.#1.:  1531.1 MH/s

Hashtype: SHA512
Workload: 256 loops, 256 accel

Speed.GPU.#1.:   241.7 MH/s

It's likely below the expected performance.


   - PC1: Windows 7, 64 bit
   - Catalyst 14.4
   - 1x AMD hd7970
   - stock core clock

   - PC2: Windows 7, 32 bit
   - ForceWare 331.67
   - 1x NVidia gtx580
   - stock core clock

   - PC3: Ubuntu 14.04, 64 bit
   - Catalyst 14.4-rev2
   - 1x AMD hd6990
   - stock core clock

   - PC4: Ubuntu 14.04, 64 bit
   - ForceWare 331.67
   - 1x NVidia gtx750Ti
   - stock core clock

   - PC5: Ubuntu 14.04, 64 bit
   - Catalyst 14.4-rev2
   - 8x AMD R9 290X
   - stock core clock

 Hash Type PC1 PC2 PC3 PC4 PC5  MD4 15445M c/s 4245M c/s 19868M c/s 5718M
c/s 183232M c/s  MD5 7893M c/s 2802M c/s 10436M c/s 3178M c/s 93800M c/s
SHA1 2495M c/s 879M c/s 3833M c/s 1103M c/s 29528M c/s  SHA256 1036M c/s 337M
c/s 1413M c/s 406M c/s 12328M c/s  SHA512 179M c/s 103M c/s 383M c/s
90M c/s 1952M
c/s  SHA-3(Keccak) 157M c/s 91M c/s 277M c/s 111M c/s 2005M c/s
RipeMD160 1690M
c/s 615M c/s 2255M c/s 677M c/s 20008M c/s  Whirlpool 41281k c/s
94752k c/s 91128k
c/s 54138k c/s 507744k c/s  LM 1271M c/s 412M c/s 967M c/s 398M c/s 15176M
c/s  NTLM 14789M c/s 4059M c/s 19011M c/s 5708M c/s 174152M c/s
NetNTLMv1 7912M
c/s 1844M c/s 8605M c/s 1355M c/s 91984M c/s  NetNTLMv2 545M c/s 205M c/s 491M
c/s 135M c/s 6456M c/s  WPA/WPA2 130k c/s 47k c/s 181k c/s 54k c/s 1454k c/s
Any advise?

One thing that called my attention is that hashcat do not make my GPU
cooler get very noise as with od6config. A bit weird, not?

Please, don't take me wrong, I know this list is for JtR, I just talked
about hashcat because I was unable to build jTr with opencl support and the
benchmark that I have now (hashcat) looks a bit smaller than expected.

I installed everything as described here:

64-bit Operating System Found..

Starting Installation of AMD APPSDK v2.9 ....
 SDK package name is :AMD-APP-SDK-v2.9-RC-lnx64.tgz
Checking Latest Version Info.....
Continuing in background, pid 20704.
 Continuing Installation...
Current directory path is  : /tmp
Untar command executed succesfully, The SDK package available
Untar command executed succesfully, The ICD package available
Copying files to /opt/AMDAPP/ ...
SDK files copied successfully at /opt/AMDAPP/
AMD Catalyst OpenCL Runtime is available hence skipping OpenCL CPU Runtime
Updating Environment variables...
32-bit path is :/opt/AMDAPP/lib/x86
64-bit path is :/opt/AMDAPP/lib/x86_64
Environment variables updated successfully
AMD APPSDK v2.9 installation  Completed
>> Reboot required to reflect the changes
*****Please refer 'AMD_APPSDK_v2.9.log' in the same directory*****
*****Refer 'README.txt' for FAQ/help in the same directory********

However, my output of ./configure say that opencl was not found

AES-NI support .............................. run-time detection
Target OS ................................... linux-gnu
Cross compiling ............................. no
Legacy arch header .......................... x86-64.h
OpenMPI support (default disabled) .......... no
Fork support ................................ yes
OpenMP support .............................. yes
OpenCL support .............................. no
CUDA support ................................ no
Generic crypt(3) mode ....................... yes

Optional libraries found:
Rexgen (extra cracking mode) ................ yes
GMP (performance) ........................... yes
NSS/NSPR (Mozilla format) ................... yes
Kerberos5 (krb5-18/23 formats) .............. yes (MIT)
PCAP (vncpcap2john and SIPdump) ............. yes
BZ2 (gpg2john extra decompression logic) .... yes

Install missing libraries to get any needed features that were omitted.

Configure finished.  Do 'make -s' to compile.

Any idea of what may be wrong?

>  Magnum you found what was the issue, I was using --wordlist PathToWordlist
>> without the equal sign and since it have not failed I was thinking it
>> worked. Maybe you guys could create a check for it or accept this way to
>> provide parameters as well? You mean, just to avoid dumb users like me to
>> think that something is broken :)
>> What do you think?
> We can't. How are we supposed to know whether you mean "default wordlist,
> and hashfiles file1 and file2", or "wordlist file1, and hashfile file2"? It
> is just ambigous.

Make sense. I was not assuming it, since in my mind just one wordlist was
accepted per time.

>  Just for curious, even with wrong syntax, what JtR was reading to see 118
>> hashes?
> It read /test.txt and hashes.txt as two hash files. It must have found 108
> valid hashes in /test.txt. This indicates you have a wordlist of poor
> quality but that is another issue.

Interesting. Is there an option on JtR to display the hashes found and
loaded? Since the wordlist is big it may help to identify what is being
detected as hashes on my poor wordist.

>  What driver version are you using? Or better, what do you
>> recommend for better performance?
> Go with Hashcat's recommendations (currently 14.4 which is also the latest
> release). That way you can use both JtR and Hashcat.
I tried, but on AMD website, once we try download 14.4, we are
automatically replaced with version 14.10. Maybe is it my performance issue?

I played a bit more with my CPU using the --fork option and I'm a bit
unsure if it's working properly. I tried for example some unix MD5 hashes
on 8 cores and I get this stats:

0g 0:05:28:22 15.17% (ETA: 2014-07-01 23:11) 0g/s 6729p/s 20189c/s 20189C/s
0g 0:05:28:22 18.76% (ETA: 16:17:04) 0g/s 6755p/s 20266c/s 20266C/s
0g 0:05:28:22 2.32% (ETA: 2014-07-10 07:00) 0g/s 6837p/s 20511c/s 20511C/s
0g 0:05:28:22 20.52% (ETA: 13:46:56) 0g/s 6750p/s 20252c/s 20252C/s
0g 0:05:28:22 32.68% (ETA: 03:51:30) 0g/s 6666p/s 20000c/s 20000C/s
0g 0:05:28:22 0.04% 0g/s 7370p/s 22112c/s 22112C/s *)%^!&#&..*)%^!&%&
0g 0:05:28:22 4.15% (ETA: 2014-07-05 22:59) 0g/s 6817p/s 20451c/s 20451C/s
0g 0:05:28:22 16.10% (ETA: 2014-07-01 21:06) 0g/s 6862p/s 20588c/s 20588C/s

Well, I assume that each ETA is per virtual core, correct? Else it could
complete more than 100% together.

The weird thing is, all my virtual cores are exactly the same, why they
have too much discrepancy from one to another?

I don't have other applications or programs running, however, we see for
example some cores that executed only 0.04% while other executed 32.68%.

Very weird, I was expecting them all to run pretty close one to the other.
Is there a way to fix it?

I tried even a bigger word-list and the discrepancy is so big that even
exceed 1 week.

0g 0:08:18:10 2.40% (ETA: 2014-07-14 21:03) 0g/s 6819p/s 20459c/s 20459C/s
0g 0:08:18:10 13.72% (ETA: 2014-07-02 23:37) 0g/s 6782p/s 20347c/s 20347C/s
0g 0:08:18:10 0.07% (ETA: 2015-11-06 15:13) 0g/s 7333p/s 22000c/s 22000C/s
0g 0:08:18:10 19.10% (ETA: 2014-07-02 06:34) 0g/s 6755p/s 20266c/s 20266C/s
0g 0:08:18:10 8.57% (ETA: 2014-07-04 11:59) 0g/s 6752p/s 20256c/s 20256C/s
0g 0:08:18:10 33.03% (ETA: 12:14:56) 0g/s 6693p/s 20079c/s 20079C/s
0g 0:08:18:10 10.33% (ETA: 2014-07-03 19:29) 0g/s 6754p/s 20262c/s 20262C/s
0g 0:08:18:10 15.49% (ETA: 2014-07-02 16:42) 0g/s 6729p/s 20187c/s 20187C/s


Powered by blists - more mailing lists

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.