Dual 7970 system producing half the expected hashrate

[bradtaco]

***MODS, could you move this to mining support? Thanks***

I just got my system put together and I'm having troubles getting CGminer configured correctly. I tried starting out with a few example configs and eventually went through the readme and built my own from scratch but am still running into similar issues.

My problem: I'm getting about half the performance I should be with both cards enabled. If I disable a GPU or give one a higher intensity than the other then the individual card starts outputting the hashrate it should be. I'm also seeing one of them hit 95C (OMFG!) even with autogpu and my target temps set ~70C.

Any and all help would be greatly appreciated!!!!

Rig:
Core2Duo
3GB DDR2
Single HD
Corsair HX1050 PSU: http://www.newegg.com/Product/Product.aspx?Item=N82E16817139034
2x XFX 7970 WITHOUT crossfire bridge: http://www.newegg.com/Product/Product.aspx?Item=N82E16814150632
Xubuntu Linux

cgminer.conf:

Code:

{
"pools" : [
        {
                "url" : "http://stratum.btcguild.com:3333",
                "user" : "user_miner",
                "pass" : "123"
        }
]
,
"intensity" : "6,6",
"vectors" : "1,1",
"worksize" : "64,64",
"kernel" : "diablo,diablo",
"lookup-gap" : "0,0",
"thread-concurrency" : "0,0",
"shaders" : "0,0",
"gpu-engine" : "0-1100,0-1100",
"gpu-fan" : "0-100,0-100",
"gpu-memclock" : "450,450",
"gpu-memdiff" : "0,0",
"gpu-powertune" : "0,0",
"gpu-vddc" : "0.000,0.000",
"temp-cutoff" : "80,80",
"temp-overheat" : "75,75",
"temp-target" : "70,70",
"api-port" : "4028",
"expiry" : "120",
"gpu-dyninterval" : "7",
"gpu-platform" : "0",
"gpu-threads" : "2",
"hotplug" : "5",
"log" : "5",
"no-pool-disable" : true,
"queue" : "1",
"scan-time" : "60",
"temp-hysteresis" : "3",
"shares" : "0",
"kernel-path" : "/usr/local/bin"
}

CGminer console outputs:
Both GPU enabled:

Code:

 cgminer version 2.11.4 - Started: [2013-04-18 17:54:38]
--------------------------------------------------------------------------------
 (5s):657.7M (avg):655.4Mh/s | A:122  R:0  HW:0  U:9.1/m  WU:9.1/m
 ST: 2  SS: 0  NB: 1  LW: 222  GF: 0  RF: 0
 Connected to stratum.btcguild.com diff 1 with stratum as user user_miner
 Block: 00b92a0ea608fd9a...  Diff:8.97M  Started: [17:54:38]  Best share: 208
--------------------------------------------------------------------------------
 [P]ool management [G]PU management [S]ettings [D]isplay options [Q]uit
 GPU 0:  96.0C 3506RPM | 329.7M/91.94Mh/s | A:23 R:0 HW:0 U: 1.72/m I: 6
 GPU 1:  58.0C 3325RPM | 329.0M/564.5Mh/s | A:99 R:0 HW:0 U: 7.40/m I: 6
--------------------------------------------------------------------------------

One GPU enabled, GPU 0 had been disabled for 10 minutes:

Code:

 cgminer version 2.11.4 - Started: [2013-04-18 17:54:38]
--------------------------------------------------------------------------------
 (5s):657.3M (avg):656.3Mh/s | A:105  R:0  HW:0  U:8.8/m  WU:8.8/m
 ST: 2  SS: 0  NB: 1  LW: 198  GF: 0  RF: 0
 Connected to stratum.btcguild.com diff 1 with stratum as user user_miner
 Block: 00b92a0ea608fd9a...  Diff:8.97M  Started: [17:54:38]  Best share: 88
--------------------------------------------------------------------------------
 [P]ool management [G]PU management [S]ettings [D]isplay options [Q]uit
 GPU 0:  96.0C 3510RPM | OFF   /85.21Mh/s | A:18 R:0 HW:0 U: 1.52/m I: 6
 GPU 1:  58.0C 3329RPM | 658.3M/574.3Mh/s | A:91 R:0 HW:0 U: 7.66/m I: 6
--------------------------------------------------------------------------------

[1715222749]

1715222749

[killingtime]

try adjusting the intensity to 9,9 (under GPU menu)

[GuiltySpark343]

I have a similar setup with two 7970s in one rig (without crossfire bridge connecting them). Running Win7 x64 and BFGMiner with just the -I 9 flag, and am getting about 650 Mh/s per card with core clocks set to 1075 MHz.

[bradtaco]

try adjusting the intensity to 9,9 (under GPU menu)

I have a similar setup with two 7970s in one rig (without crossfire bridge connecting them). Running Win7 x64 and BFGMiner with just the -I 9 flag, and am getting about 650 Mh/s per card with core clocks set to 1075 MHz.

Set intensity to 9 with the same results. Temps are lower because I've left it off:

Code:

 cgminer version 2.11.4 - Started: [2013-04-18 19:00:02]
--------------------------------------------------------------------------------
 (5s):660.8M (avg):652.2Mh/s | A:9  R:0  HW:0  U:8.1/m  WU:8.1/m
 ST: 2  SS: 0  NB: 1  LW: 24  GF: 0  RF: 0
 Connected to stratum.btcguild.com diff 1 with stratum as user user_miner
 Block: 01a5b95f64f506aa...  Diff:8.97M  Started: [19:00:02]  Best share: 81
--------------------------------------------------------------------------------
 [P]ool management [G]PU management [S]ettings [D]isplay options [Q]uit
 GPU 0:  74.0C 3637RPM | 328.9M/365.4Mh/s | A:4 R:0 HW:0 U: 3.62/m I: 9
 GPU 1:  48.0C 3425RPM | 328.8M/324.8Mh/s | A:5 R:0 HW:0 U: 4.52/m I: 9
--------------------------------------------------------------------------------

Any more thoughts?

[bradtaco]

I'm also starting to wonder if only one of the cards is actually being used since only one of them is getting hot. Here's an updated console output a few minutes later:

Code:

 cgminer version 2.11.4 - Started: [2013-04-18 19:00:02]
--------------------------------------------------------------------------------
 (5s):657.9M (avg):656.6Mh/s | A:48  R:0  HW:0  U:9.1/m  WU:9.1/m
 ST: 2  SS: 0  NB: 1  LW: 104  GF: 0  RF: 0
 Connected to stratum.btcguild.com diff 1 with stratum as user user_miner
 Block: 01a5b95f64f506aa...  Diff:8.97M  Started: [19:00:02]  Best share: 81
--------------------------------------------------------------------------------
 [P]ool management [G]PU management [S]ettings [D]isplay options [Q]uit
 GPU 0:  87.0C 3551RPM | 327.2M/333.3Mh/s | A:21 R:0 HW:0 U: 3.96/m I: 9
 GPU 1:  50.0C 3371RPM | 328.8M/323.8Mh/s | A:28 R:0 HW:0 U: 5.29/m I: 9
--------------------------------------------------------------------------------

[Slix313]

Does you mobo support pcie3.0? if it does, try going into bios and changing the pci ports to run 2.0 instead

[GuiltySpark343]

A couple of thoughts:

1) which catalyst drivers are you using (I'm using latest)

2) do you have your monitor connected to just one card? if so, try shutting down, adding the cross-fire bridge, rebooting, and re-running your mining software ... see if that changes things

[bradtaco]

I installed the drivers via the reps:

Code:

sudo apt-get install fglrx-updates fglrx-amdcccle-updates fglrx-updates-dev

The system is headless at the moment so there aren't any monitors connected. I've read everywhere that the crossfire bridge only causes problems.

Mods, can you move this topic to the appropriate section? Thanks

[bradtaco]

Also check this out, the system has been sitting for a few minutes like this. GPU 0 is overheating and has the fan running yet it is in "REST" mode while GPU 1 is a nice idle 59C yet it's doing all the work. Make sense of that?

Code:

 cgminer version 2.11.4 - Started: [2013-04-18 23:48:12]
--------------------------------------------------------------------------------
 (5s):599.2M (avg):611.0Mh/s | A:99  R:1  HW:0  U:9.2/m  WU:9.3/m
 ST: 2  SS: 0  NB: 2  LW: 181  GF: 0  RF: 0
 Connected to stratum.btcguild.com diff 1 with stratum as user user_miner
 Block: 016b9b768b281a11...  Diff:8.97M  Started: [23:56:12]  Best share: 356
--------------------------------------------------------------------------------
 [P]ool management [G]PU management [S]ettings [D]isplay options [Q]uit
 GPU 0:  90.0C 3474RPM | REST  /69.25Mh/s | A:11 R:0 HW:0 U: 1.02/m I: 6
 GPU 1:  59.0C  664RPM | 599.1M/545.9Mh/s | A:88 R:1 HW:0 U: 8.17/m I: 6
--------------------------------------------------------------------------------

[Signus]

Turn down your memclock to save the fans, I'd try about 300-350MHz.

Secondly, try first setting your Intensity to 8. If you're using Linux some drivers can't support higher than 8. I suggest starting with a clean slate and letting cgminer auto configure everything, so just move the cgminer.conf to a location to back it up, and run without any arguments. It will then autoconfigure your settings. If you start blank and change a setting at a time, we can see where a change affects your cards the most.

Can you show us a picture of your setup? To be sure your cards are being suffocated somehow.

[bradtaco]

My configs are from scratch and the system is sitting in an open hallway with plenty of airflow. One of the cards in the case is towards the top but it has enough space to suck in air.

I think this is some weird config issue or maybe a bug in cgminer.

Here's my latest running config:

Code:

{
"pools" : [
        {
                "url" : "http://stratum.btcguild.com:3333",
                "user" : "user_miner",
                "pass" : "123"
        }
]
,
"intensity" : "6,6",
"vectors" : "1,1",
"worksize" : "64,64",
"kernel" : "poclbm,poclbm",
"lookup-gap" : "0,0",
"thread-concurrency" : "0,0",
"shaders" : "0,0",
"gpu-engine" : "0-0,0-0",
"gpu-fan" : "0-85,0-85",
"gpu-memclock" : "0,0",
"gpu-memdiff" : "0,0",
"gpu-powertune" : "0,0",
"gpu-vddc" : "0.000,0.000",
"temp-cutoff" : "95,95",
"temp-overheat" : "85,85",
"temp-target" : "75,75",
"api-port" : "4028",
"expiry" : "120",
"gpu-dyninterval" : "7",
"gpu-platform" : "0",
"gpu-threads" : "2",
"hotplug" : "5",
"log" : "5",
"no-pool-disable" : true,
"queue" : "1",
"scan-time" : "60",
"temp-hysteresis" : "3",
"shares" : "0",
"kernel-path" : "/usr/local/bin"
}

[-ck]

You are mining only on the first GPU with two openCL instances pointed at the same card. You have driver fuckage. Use an uninstall application (not the uninstall with catalyst control centre) and try a different driver.

[Signus]

Are you using Linux?

I've seen a number of threads where people and 7950's are having low hashrates or problems with cgminer working entirely.

[bradtaco]

I'm running Xubuntu (linux) and as mentioned before I installed the drivers via the public repos. If I'm only mining on one GPU with two OpenCL instances then wouldn't that be a configuration issue in cgminer and not my driver?

I'll see if I can find another linux driver but honestly I don't think that's the issue. I'll also try with Windows 7 just to make sure one of my cards isn't defective or something.

Any other thoughts folks? I was under the impression a lot more people were using linux on their miners but it seems like most of the folks who replied to this thread are using windows.

[Signus]

Try each card independently of eachother first.

If they work fine independently, I would check that each of your PCI-E slots is configured correctly (if you're not using risers) and they they are both getting the appropriate power. So try each card in both of your PCI-E slots.

[-ck]

I'm running Xubuntu (linux) and as mentioned before I installed the drivers via the public repos. If I'm only mining on one GPU with two OpenCL instances then wouldn't that be a configuration issue in cgminer and not my driver?

Fine, don't believe the person who wrote the application 

[Signus]

I'm running Xubuntu (linux) and as mentioned before I installed the drivers via the public repos. If I'm only mining on one GPU with two OpenCL instances then wouldn't that be a configuration issue in cgminer and not my driver?

Fine, don't believe the person who wrote the application 

Aha, well like I mentioned check each card on it's own to see that cgminer does read them each correctly in both PCI-E slots, and then you can confirm whether it is cgminer or not.

[bradtaco]

I'm running Xubuntu (linux) and as mentioned before I installed the drivers via the public repos. If I'm only mining on one GPU with two OpenCL instances then wouldn't that be a configuration issue in cgminer and not my driver?

Fine, don't believe the person who wrote the application 

Ya got me and you're most likely right. Eating my hat now....... (I was doubtful because I was installing the driver per the instructions of every tutorial I've seen so far.)

I ended up trying Windows out since that seems to be what most folks on this forum are using (by my guess) and now I have both cards achieving their expected hashrates.

My new problem is that with --auto-fan and --auto-gpu set along with the default target temps my cards now heat up until they hit my cutoff temp of 90C, disable, cool down, then re-enable, but DON'T take on any jobs after that and just sit at an idle 0 hashrate.

Any thoughts?

[Signus]

Set your memclock to ~300MHz-350MHz and set your fans as high as you can if you have any OC tools to use.