third connected vga fails in cgminer, any pair of 2 works (dual psu problem?)

[neet-o]

Hi,

I've just bought a second PSU - RM850 - with a thought of expanding my current config (Sempron 145, Asrock 970 Extreme4, 4gb ram, Corsair HX850, Asus R9 280X, Gigabyte R9 290), bridged 4th and 5th pins like Corsair's manual said (http://www.corsair.com/en/blog/testing-your-corsair-power-supply/) and connected it to my newly bought R9 290X (pci-e connector + 16x-1x powered riser).
After that when I try to launch cgminer it fails with "-4 clEnqueueWriteBuffer failed". If I disconnect R9 280X, for example, 290+290X both work fine. I've tried some options like connecting my cards in different ways like 280X+290 to secondary PSU, system+290X to primary and it fails the same way.
What am I missing here? My thoughts are that maybe PSUs are not working well together and somehow only one is being able to provide current at one period of time. In that case there's just not enough power for 3 cards hence the error.

[ineedit]

Hi,

I've just bought a second PSU - RM850 - with a thought of expanding my current config (Sempron 145, Asrock 970 Extreme4, 4gb ram, Corsair HX850, Asus R9 280X, Gigabyte R9 290), bridged 4th and 5th pins like Corsair's manual said (http://www.corsair.com/en/blog/testing-your-corsair-power-supply/) and connected it to my newly bought R9 290X (pci-e connector + 16x-1x powered riser).
After that when I try to launch cgminer it fails with "-4 clEnqueueWriteBuffer failed". If I disconnect R9 280X, for example, 290+290X both work fine. I've tried some options like connecting my cards in different ways like 280X+290 to secondary PSU, system+290X to primary and it fails the same way.
What am I missing here? My thoughts are that maybe PSUs are not working well together and somehow only one is being able to provide current at one period of time. In that case there's just not enough power for 3 cards hence the error.

what is your cgminer command line and with all three cards connected what is the response to cgminer -n ?

[neet-o]

-n:
 [2014-01-18 13:25:14] CL Platform 0 vendor: Advanced Micro Devices, Inc.

 [2014-01-18 13:25:14] CL Platform 0 name: AMD Accelerated Parallel Processing

 [2014-01-18 13:25:14] CL Platform 0 version: OpenCL 1.2 AMD-APP (1348.5)

 [2014-01-18 13:25:14] Platform 0 devices: 3
 [2014-01-18 13:25:14]  0       Tahiti
 [2014-01-18 13:25:14]  1       Hawaii
 [2014-01-18 13:25:14]  2       Hawaii
 [2014-01-18 13:25:14] GPU 0 AMD Radeon R9 200 Series hardware monitoring enable
d
 [2014-01-18 13:25:14] GPU 1 AMD Radeon R9 200 Series hardware monitoring enable
d
 [2014-01-18 13:25:14] GPU 2 AMD Radeon R9 200 Series hardware monitoring enable
d
 [2014-01-18 13:25:14] 3 GPU devices max detected

config:
"intensity" : "13,20,20",
"worksize" : "256,512,512",
"lookup-gap" : "2,2,2",
"thread-concurrency" : "8192,24550,24550",
"gpu-engine" : "1100-1100,1000-1000,1000-1000",
"gpu-fan" : "55-55,63-63,63-63",
"gpu-memclock" : "1500,1250,1250",
"gpu-powertune" : "-20,20,20",
"failover-only" : true,
"gpu-dyninterval" : "7",
"gpu-threads" : "2",
"scrypt" : true

[ineedit]

I think you have a tuning parameter clash try the following command to check if a default set of commands works okay

cgminer -o 'pool' -u 'username' -p 'password' --scrypt --auto-fan -I 8 --thread-concurrency 8192

[neet-o]

I think you have a tuning parameter clash try the following command to check if a default set of commands works okay

cgminer -o 'pool' -u 'username' -p 'password' --scrypt --auto-fan -I 8 --thread-concurrency 8192

It works, with very low hashrate though (14KH per card).
Could it be that with those settings cards do not pull enough from PSU and this is what allows them to work simultaneously?

[ineedit]

I think you have a tuning parameter clash try the following command to check if a default set of commands works okay

cgminer -o 'pool' -u 'username' -p 'password' --scrypt --auto-fan -I 8 --thread-concurrency 8192

It works, with very low hashrate though (14KH per card).
Could it be that with those settings cards do not pull enough from PSU and this is what allows them to work simultaneously?

You can test that by ramping up the intensity, try lifting them all to 10 then 12, 14, etc

Then each card individually with thread-concurrency, then engine clocks, then memory clock, when you get a failure then you know what will have failed.

[neet-o]

You can test that by ramping up the intensity, try lifting them all to 10 then 12, 14, etc

Then each card individually with thread-concurrency, then engine clocks, then memory clock, when you get a failure then you know what will have failed.

But those settings are already stable for each card, it's just that they fail when all 3 cards are together. Why is that that's what I want to know.

[ineedit]

You can test that by ramping up the intensity, try lifting them all to 10 then 12, 14, etc

Then each card individually with thread-concurrency, then engine clocks, then memory clock, when you get a failure then you know what will have failed.

But those settings are already stable for each card, it's just that they fail when all 3 cards are together. Why is that that's what I want to know.

There is no set answer, every combination of video card, motherboard, PSU, etc will give a different answer. You have changed your rig so you now need to re-tune.

[neet-o]

ok, got it. will try to retune.

[neet-o]

it really does work just fine after retuning (had to reconfig from scratch). good thing it weren't the hardware issues. thank you for your help!