dual r9 290, computer crashes after a few seconds

[barisdad]

Hi,
I am trying to mine on a machine with 2 r9 290 cards. I run cgminer and it starts without any errors, but after a few seconds of work my computer crashes and I have to restart it using the power button on the back.
I tried ubuntu 12.04 server 64 bit with xdm, then installed xubuntu on the server and also tried running batm from a usb stick. Same results.
I am using catalyst 13.12 linux drivers.
I am currently running the pre-built cgminer 3.7.2. I also tried building it from source but it segfaulted when I tried to start it.
I also installed app sdk 2.9 and adl sdk 6.0 when I tried to build cgminer.

I tried these cgminer configurations:
1. --scrypt -o stratum+tcp://gigahash.wemineltc.com:3333 -u <username>.1 -p <password> -g 1 -w 512 --lookup-gap 2 --thread-concurrency 24500 -I 20 --gpu-fan 70 --gpu-engine 947 --gpu-memclock 1250 --gpu-powertune 20

2.
{
"pools" : [
{
"url" : "stratum+tcp://gigahash.wemineltc.com:3333",
"user" : "<myuser>.1",
"pass" : "<mypassword>"
},
{
"url" : "stratum+tcp://usa2.wemineltc.com:3334",
"user" : "<myuser>",
"pass" : "<mypassword>"
}
],
"api-allow" : "W:127.0.0.1",
"api-listen" : true,
"scrypt" : true,
"gpu-threads" : "1",
"gpu-engine" : "1000",
"gpu-fan" : "75-100",
"gpu-memclock" : "1500",
"gpu-powertune" : "20",
"intensity" : "20",
"vectors" : "1",
"worksize" : "512",
"lookup-gap" : "2",
"thread-concurrency" : "24550"
}

3.
{
"pools" : [
{
"url" : "stratum+tcp://gigahash.wemineltc.com:3333",
"user" : "<myuser>.1",
"pass" : "<mypassword>"
},
{
"url" : "stratum+tcp://usa2.wemineltc.com:3334",
"user" : "<myuser>",
"pass" : "<mypassword>"
}
],

"intensity" : "20",
"vectors" : "1",
"worksize" : "256",
"lookup-gap" : "2",
"thread-concurrency" : "32765",
"gpu-engine" : "1000",
"gpu-fan" : "40-100",
"gpu-memclock" : "1350",
"gpu-powertune" : "20",
"temp-cutoff" : "99",
"temp-overheat" : "95",
"temp-target" : "90",

"api-port" : "4028",
"expiry" : "120",
"failover-only" : true,
"gpu-threads" : "1",

"log" : "5",
"queue" : "1",
"scan-time" : "60",
"temp-hysteresis" : "3",

"scrypt" : true,
"kernel" : "scrypt",
"kernel-path" : "/usr/local/bin"
}


I also tried with a 947 clock instead of 1000. same results.

This is my hardware configuration:

2 X Radeon R9 290 GPU - 4GB 512 bit DDR5 mem 21227-00-40G -Sapphire
GA-870A-UD3 Rev 2.0 - AMD 870 Chipset - GA-870A-UD3 -Gigabyte
AMD Athlon II X2 270 - 3200 MHz - 45nm -ADX270OCK23GM -AMD    
Gamma 50 Intel LGA775 and AMD socket AM2 Gamma-50 -DEEPCOOL    
4GB DDR3-1600MHz, PC3-12800, CL11-11-11- F3-1600C11S-4GNS -G-Skill    
500gb - 3.5 inch - 7200 RPM - SATA3 6Gb/    DT01ACA050 -Toshiba    
Cougar Solution Middle Tower Case - Dime    67M3-NPS -Cougar    
120mm Low Noise Sleeve bearing fans - 4-   R4-S2S-124K-GP -Coolermaster    
OCZ850GXSSLI - 850W PSU - OCZ

Thanks.

[Lazerballz]

Did you check https://litecoin.info/Mining_hardware_comparison?

In the list I do not see any 290 running on  --gpu-engine 947 or "gpu-engine" : "1000", which you stated below (unless you flashed the bios)

What does it do when you try 830 for GPU engine and 1500 for mem?

Have you ever been able to run longer than an hour?

[barisdad]

I based it on this reddit thread and another one which I can't find.
http://www.reddit.com/r/litecoinmining/comments/1rurj5/what_are_your_r9_290_and_290x_settings/

running with 830 and 1500 doesn't crash the computer, but my x session freezes so I can't tell the hash rate. When i remote login via ssh the cgminer process seems to be running. Is there any way to tell the hashrate via ssh?
Am I supposed to run cgminer from within an x session?

These are my first tests with this rig, I never ran it for more than a minute.

[Lazerballz]

mmmm.

Well the way you should be able to see the hashrate via ssh is either via a monitor tool or run cgminer from ssh. In BAMT (Debian Linux distro, special made for mining) there is an application called GPUMON which allows me to see the current hashrate.

This is how it looks

https://images.weserv.nl/?url=i45.tinypic.com/ww18gm.jpg

About the settings. Basically it's taking one of the midrange settings and go from there. I've seen people with same cards and different settings.

[Lazerballz]

Also, with 2 290's running at 100% you might need to check your PSU set up.

Getting too much power from the same +12v lane would exhaust the psu. (normally your system should reboot then)
Getting too much temp should reboot the system also

Like I said in my previous post, use a mid range setting (like you have now) and work from there untill you have a stable good working mining config.

Next to that.

Make an .sh file with the following:

export DISPLAY=:0
export GPU_MAX_ALLOC_PERCENT=100
export GPU_USE_SYNC_OBJECT=1
cd <location of cgminer>
./cgminer <settings or use config file>

[barisdad]

I can't run cgminer via ssh. When I try to run it says "no device detected". I tried adding export DISPLAY=:0 before running and it didn't help.

[Lazerballz]

Mmmm 
 

Have you tried running BAMT (latest version because of the drivers) from USB directly or with 1 card only?

[barisdad]

going to try this for ssh:
http://wiki.tiker.net/OpenCLHowTo#Running_OpenCL_code_remotely_on_AMD_GPUs_.28via_SSH.29

I am not sure I understand your question. I ran BAMT from USB, is there another way? I didn't try with one card only.

[barisdad]

I tried booting with bamt 1.2 on a USB stick, but the computer shuts down after a minute before I have a chance to do anything. I didn't even start the miner.

[infinitybo]

@Barisdad Yes it's sure you can fix the config quickly furthermore you should use the "mining" section.

[Lazerballz]

BAMT starts the miner right away. go to Start menu button -> SMOS Control -> Stop mining sessions.

[barisdad]

https://www.dropbox.com/s/xspwzrxnnaqnk2t/13885927386020.jpg
The same power cable seems to be connected to both cards. Is this what you meant?

[goldenarmZ]

Have you tried just running a very basic cgminer command with none of the gpu settings? It's best to eliminate all variables when testing a problem then bring them back one at a time.

I would remove all config files and just run cgminer with nothing but -o -u and -p specified. Then once you have it stable, try configuring from there.

[barisdad]

I did. It ran without any problems at about 20k per card.

[nick_a]

I am using these config for my 7970 in desktop shortcut.
C:\cgminer372\cgminer.exe --scrypt -o stratum+tcp://xxxx -u xxxx -p 123 --failover-only -o stratum+tcp://xxxx -u xxx -p 123 -w 256 --thread-concurrency 8192 --auto-fan -I 13 -g 2

[truthanddelusion]

I am almost curious if your power supply is not giving the system enough power. Are you using any logging software for power usage and heat to see if they correlate with the shut down?

You could also try

-I 20
--thread-concurrency 24550
--gpu-engine 1000
--gpu-memclock 1500

Either way I am curious. Let me know if you get them working.

[goldenarmZ]

I did. It ran without any problems at about 20k per card.

That probably suggests some issues with your overclocking settings imo. Try adding back things like shader number and thread concurrency but the leave the overclocking/fan/power control out for now.

[barisdad]

Can you see the picture i put above? It says "invalid picture" in my phone.

[goldenarmZ]

Can you see the picture i put above? It says "invalid picture" in my phone.

No, it doesn't show up.

[barisdad]

http://imageshack.com/a/img20/8417/qp4h.jpg