11-20-2012, 10:55 AM
11-20-2012, 04:33 PM
(11-18-2012, 11:33 AM)atom Wrote: [ -> ]Here's a representative example of how it is (and has always been in the previous versions too). I took the snapshot at the very start of the attack but it's the same pattern as when it's in the middle of the run. What you don't see is that the GPUs were loaded in this order:(11-18-2012, 04:18 AM)mastercracker Wrote: [ -> ](11-17-2012, 10:16 AM)atom Wrote: [ -> ]Yeah its what to expect. If there is no workload left to share some GPUs will idle.My point was just that, like before, the GPU number for load and speed are not matching.
I see, you meant this?
Quote:This is fine but the speed is shown for GPU1 and the HWmon.GPU shows the %util for GPU2.
Thats strange. Do you have Afterburner? It would be nice if you can take a screenshot of a running Afterburner plus oclhashcat-plus showning the wrong values
GPU1: GTX470
GPU2: GTX460
GPU3: GTX460
You can see that the speed is shown for GPU1 but the %util is shown on GPU2. It's not on the picture but GPU-Z does confirm that it is the GTX470 that got the load. Another proof of the problem is that my GTX470 and one of my GTX460 are watercooled and can't show fan speed. If you look at the fan speed of the GPU1, there is a value. This means that it's not the GTX470 like it's supposed to be. Finally, here's the actual device order that I can establish:
-d 1: GTX470
-d 2: GTX460 watercooled
-d 3: GTX460 stock fan
So for some reasons, the cracking speed is accurate but for the hardware monitoring -d 3 becomes -d 1.
11-23-2012, 04:55 PM
Damn that damn Catalyst drivers!
I have posted another rc1 (same host code) but with kernels compiled for catalyst 12.8.
Please guys, downgrade to catalyst 12.8 and try with the new version again.
On all my test systems this feels much better.
NOTE!!! It will tell you your driver is outdated. Just add --force to your cmdline to make it work.
I have posted another rc1 (same host code) but with kernels compiled for catalyst 12.8.
Please guys, downgrade to catalyst 12.8 and try with the new version again.
On all my test systems this feels much better.
NOTE!!! It will tell you your driver is outdated. Just add --force to your cmdline to make it work.
11-23-2012, 09:53 PM
I'm seeing some good speed increases with the 12.8 build from 0.09. Catalyst 12.9+ isn't working with VCL so I cannot test.
I would really like to stick with 12.8 for a while. Yes it's "old," but it seems to be the best driver for the time being.
oclHashcat-plus-0.09
oclHashcat-plus-0.10-12.8
I would really like to stick with 12.8 for a while. Yes it's "old," but it seems to be the best driver for the time being.
oclHashcat-plus-0.09
Code:
./oclHashcat-plus64.bin -n 160 --gpu-loops 1024 example0.hash -a 3 ?a?a?a?a?a?a?a?a --force
Speed........: 31918.5M c/s Real, 37662.5M c/s GPU
LD_LIBRARY_PATH=/usr/lib/vcl vclrun ./vclHashcat-plus64.bin -n 160 --gpu-loops 1024 example0.hash -a 3 ?a?a?a?a?a?a?a?a --force
Speed........: 27628.3M c/s Real, 36797.0M c/s GPUoclHashcat-plus-0.10-12.8
Code:
./oclHashcat-plus64.bin -n 160 --gpu-loops 1024 example0.hash -a 3 ?a?a?a?a?a?a?a?a --force
Speed.GPU.#*...: 43841.8M/s
LD_LIBRARY_PATH=/usr/lib/vcl vclrun ./vclHashcat-plus64.bin -n 160 --gpu-loops 1024 example0.hash -a 3 ?a?a?a?a?a?a?a?a --force
Speed.GPU.#*...: 42385.5M/s11-26-2012, 11:27 AM
Yeah I noticed the same, 12.8 still the best driver so far
11-27-2012, 02:25 PM
RC-2 is up and forces catalyst 12.8 to be installed.
11-27-2012, 11:20 PM
i'm so glad you went back to catalyst 12.8. i've had so much trouble with the new drivers.
11-28-2012, 10:57 AM
Yeah same here. I've added a very hard driver version check on statup to inform the users about this situation and they have to manually add --force to ignore it and use other drivers. I think this will help 