AMD Llano A8 3850
#1
Is it possible to use AMD Llano A8 3850 gpu with oclhashcat-plus ?
#2
(11-22-2011, 02:35 AM)mr008 Wrote: Is it possible to use AMD Llano A8 3850 gpu with oclhashcat-plus ?

Although this question is a little off topic ... Smile

This wasn't as easy to find out as I first thought, however I think you can.
It has the AMD Radeon HD 6550D which apparently has 400 stream processors.

So I am guessing it will, but you should make better checks before you buy it.

If your computer you have now has a PCIe 2 slot you might want to think about buying a separate card if your interests are solely on cracking.
#3
yes the docs say it is supported by opencl. so in theory, it should work. however, no one ever tested it Smile
#4
If someone test AMD Llano with oclhashcat, post result here.
#5
I modified the oclExample.sh script to take arguments passed on the command line. Here is the performance with -n 800

chort@hydra:~/code/oclHashcat-lite-0.06$ ./oclExample.sh -n 800 -d 1
oclHashcat-lite v0.6 by atom starting...

Platform: AMD compatible platform found
Watchdog: Temperature limit set to 90c
Device #1: BeaverCreek, 256MB, 0Mhz, 5MCU
Device #2: skipped by user
[s]tatus [p]ause [r]esume [q]uit => s
Status.......: Running
Hash.Type....: MD5
Time.Running.: 4 secs
Time.Left....: 2 mins, 9 secs
Plain.Text...: ***etrf
Plain.Length.: 7
Speed........: 1006.5M/s
Progress.....: 4194304000/134960504832 (3.11%)
HW.Monitor.#1: 0% GPU, 27c Temp


I found that the APU paired with a 6950 (Cayman, 830Mhz clock) increased the h/s slightly. Unfortunately, the APU paired with a 5870 (Cypress) decreases the h/s. I'm eagerly anticipating the next release of oclHashcat-lite that will allow running on mis-matched GPUs.

PS this is what hashcat gets running on the CPU (same hash from the oclExample.sh):
chort@hydra:~/code/hashcat-0.37$ ./h64 -a 3 examples/oclHashcat-lite.md5
Initializing hashcat v0.37 by atom with 8 threads and 32mb segment-size...

NOTE: press enter for status-screen

Added hashes from file examples/oclHashcat-lite.md5: 1 (1 salts)
Activating quick-digest mode for single-hash
<snip>
Charset...: abcdefghijklmnopqrstuvwxyz
Length....: 7
Index.....: 0/1 (segment), 8031810176 (words), 0 (bytes)
Recovered.: 0/1 hashes, 0/1 salts
Speed/sec.: - plains, 31.22M words
Progress..: 2302302676/8031810176 (28.66%)
Running...: 00:00:01:13
Estimated.: 00:00:03:03

PPS The APU is a little better than 1/6th the speed of a 5870 OC'd to 900Mhz:

chort@hydra:~/code/oclHashcat-lite-0.06$ ./oclExample.sh -n 800 -d 2
oclHashcat-lite v0.6 by atom starting...

Platform: AMD compatible platform found
Watchdog: Temperature limit set to 90c
Device #1: skipped by user
Device #2: Cypress, 512MB, 0Mhz, 20MCU
[s]tatus [p]ause [r]esume [q]uit => s
Status.......: Running
Hash.Type....: MD5
Time.Running.: 4 secs
Time.Left....: 18 secs
Plain.Text...: ***qe8v
Plain.Length.: 7
Speed........: 5913.5M/s
Progress.....: 28311552000/134960504832 (20.98%)
HW.Monitor.#1: 0% GPU, 40c Temp
#6
Do you try with oclHashcat-plus and wpa-wpa2 ?

But AMD Llano only allow CrossFire with HD 6450, 6570 or 6670
#7
I've used it with -plus, not with WPA because I don't have any .PCAPs to use.

It doesn't need to be crossfired to work, just go into the BIOS settings and force the APU to be "always on." Then (on Linux at least) put a dummy plug in any card that you don't have hooked up to a monitor (on mine the APU is connected to monitor). Normally the BIOS will disable the APU if you have another graphics card installed and a monitor connected to the expansion card.
#8
Do you tray d 1 + d 2 ?
#9
Yes I've tried with both GPUs currently (the default), but it slows down the 5870 too much. The 6550D runs full speed the entire time, but the 5870 stalls, goes to full speed, stalls, over and over. The overall speed is about 4/5 the speed of the 5870 by itself. I just tried on oclHashcat-lite-0.08 and the same is true Sad

It can still be helpful if you split a hash file into 1/6th and give 1/6 to the 6550D and 5/6 to the other GPU (or with -lite, do 5/6 of the brute-force space on one GPU, 1/6 on the on-die GPU, etc).
#10
if you are on windows this might be because of the 100% cpu killer bug