Requesting more help with cgminer

[timbck2]

I'm a newbie, so I can't post to the proper forum, I'm hoping someone will take pity on me and help with a cgminer problem.

I'm running cgminer 2.3.1 (latest version) on Windows 7 with a pair of Radeon HD 6670 cards. One card has dual monitors attached to it, the other is just there for mining. At night when I'm not using the PC, I run with a configuration that runs two threads per GPU - no problems there. During the workday, I have cgminer configured to use only the headless GPU for mining - it's effectively disabled on the GPU that has my displays attached. However, it's still severely affecting my display performance - screen refreshes are painfully slow. I didn't have this problem before installing the second GPU, so I'm thinking it's likely something I have configured incorrectly.

Here's my daytime cgminer config:

Code:

{
"pools" : [
	{
		"url" : "*****",
		"user" : "*****",
		"pass" : "*****"
	}
],

"device" : "1",
"gpu-threads" : "0,2",
"gpu-engine" : "800,850",
"gpu-memclock" : "667,800",
"auto-fan" : true,
"temp-target" : "75",
"temp-overheat" : "80",
"temp-cutoff" : "85",
"temp-hysteresis" : "3",

"kernel-path" : ".",
"intensity" : "d,7",
"expiry" : "120",
"failover-only" : true,
"log" : "5",
"queue" : "1",
"retry-pause" : "5",
"scan-time" : "60",

"donation" : "0.1"
}

Shouldn't the line "device" : "1" effectively disable device 0 as far as mining is concerned? The cgminer output makes it look like it's disabled, but it doesn't perform like it's disabled. Here's the cgminer console screen header, clearly showing that GPU 0 (the display card) is "OFF":

Code:

 cgminer version 2.3.1 - Started: [2012-03-06 12:49:21]
--------------------------------------------------------------------------------
 (5s):63.3 (avg):79.9 Mh/s | Q:3  A:0  R:0  HW:0  E:0%  U:0.00/m
 TQ: 1  ST: 1  SS: 0  DW: 0  NB: 1  LW: 0  GF: 0  RF: 0
 Connected to ***** with LP as user *****
 Block: 0000043571d84726ffaaa1064d78b8ca...  Started: [12:49:21]
--------------------------------------------------------------------------------
 [P]ool management [G]PU management [S]ettings [D]isplay options [Q]uit
 GPU 0:  35.0C  80%    | OFF  /  0.0Mh/s | A:0 R:0 HW:0 U:0.00/m I: 0
 GPU 1:  55.5C  84%    | 112.2/118.5Mh/s | A:0 R:0 HW:0 U:0.00/m I: 7
--------------------------------------------------------------------------------

I had better display performance before I added the second dedicated Radeon HD 6670, and was using one card (with just one mining thread) for both mining and driving my two monitors.

Any help will be greatly appreciated!

[1713857459]

1713857459

[timbck2]

I'm still experiencing the sluggish display problem described in my previous post. Anyone have any ideas?

[DeathAndTaxes]

Are you sure you are disabling the right GPU?  They aren't necessarily in any particular order (i.e. it could be GPU #1 that your display is connected to not GPU #0).  This is a windows = stupid issue not a cgminer issue.

Even simpler solution have you tried just using dynamic intensity.  It will react to non-mining workload and lower intensity when display is used.

[Xmufa23X]

I haven't used cgminer in awhile, but maybe I can remember a few things...
Set the Intensity of the primary card to a lower setting instead of "d". Something like 4-7 should be good. Then something like 11-14 for the secondary card. "d" just auto-balances the load to give a likable FPS.

I do prefer GUIminer (poclbm) so I can adjust the FPS rather then intensity. The fact is, the human eye can only detect up to 72fps. Assuming it's a constant 72fps for that whole second. I usually set my flag up like this "-v -w128 -f80" to keep an enjoyable desktop. But that's just my opinion.

[timbck2]

Is there a way to find out which GPU is which in Windows? I just assumed that the one in the first PCI-E slot was #0, the second slot #1, and so forth.

I think I already tried running it with intensity set to dynamic, and my desktop performance was just as bad, but I've tried so many things I'm not entirely sure. I'll give that a try.

[DeathAndTaxes]

Is there a way to find out which GPU is which in Windows? I just assumed that the one in the first PCI-E slot was #0, the second slot #1, and so forth.

I think I already tried running it with intensity set to dynamic, and my desktop performance was just as bad, but I've tried so many things I'm not entirely sure. I'll give that a try.

Not easily but :
1) Close cgminer.  download GPU-Z.
2) use the GPU for somethings desktop related (like playing hardware accelerated hd video or flash).
3) look in GPU-Z which GPU shows load
4) record that #. 
5) turn on cgminer, enable mining on all GPUs (GPU-Z should show load of 99% on all GPUs)
6) disable mining on each GPU one by one until the GPU# you recorded in step #4 goes to 0.  That GPU in cgminer is the one linked to your display.

Yeah it is a lot, windows it totally asinine when it comes to inconsistent device ordering.

Given you only have 2 GPUs I would just try disabling each one.  Start cgminer w/ both GPU enabled with dynamic intensity.  Hit G to bring up GPU menu and D to disable one GPU.  See if it helps.  If not enable it and then disable the other one.

TL/DR there is no specific order.  It is totally random and the order for you will vary from everyone else.  Even someone who thinks they know the order likely doesn't not across all systems and all card configurations.  Just experiment.

[timbck2]

Thank you so much! They were in the opposite order I thought they were. I've been running with my display GPU doing all the mining <*facepalm*>!

[DeathAndTaxes]

Thank you so much! They were in the opposite order I thought they were. I've been running with my display GPU doing all the mining <*facepalm*>!

Don't feel bad it is windows.  They may be in the #0 = display GPU on the next computer you build.  Windows = no rhyme or reason to the ordering of GPUs.