Bitcoin Forum

Hello,

So i just bought 2 new MSI R9 280x's and tried to run them together using cgminer but it seems like one card is getting alot hotter than the other. And other times it just turns to SICK.
Anyone knows how to fix this?

The difference is sometimes almost 20 degrees.

Thnx in advance

do you have the cards in a case? or open stand?

In a case atm.

And i'm using this batch file:

setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECTS 1
C:\cgminer-3.7.2-windows\cgminer.exe --scrypt -o --------------- -u ----------- -p ---- --thread-concurrency 8192 -I 12 -g 2 -w 256


EDIT:
Difference in temperature is smaller now, however after a few minutes my pc will feeze..

I have the same Graphic cards, 2x 280x from MSI in CrossFire. In gaming the temps are ok, gpu0 65gr. and gpu1 62gr but the cards are to close and when i am mining the top card is heating to 90 gr :(    therefore i use I12 for gpu0 and it stays @ 75gr

May i ask the rest of your settings? Since my pc keeps crashing..

Are the cards spread out evenly enough so there's a few inches above the fan intake for each one? Also have you tried swapping the cards out so they trade PCI slots and tested to see if you observe the same temperature difference?

here is my cgminer.conf:

{
"pools" : [
   {
      "url" : "stratum+tcp://xxx.com:000",
      "user" : "yyy",
      "pass" : "zzz"
   }
]
,
"intensity" : "12,13",
"vectors" : "1,1",
"auto-fan" : true,
"worksize" : "256,256",
"kernel" : "scrypt,scrypt",
"no-submit-stale" : true,
"lookup-gap" : "2,2",
"thread-concurrency" : "8192,8192",
"shaders" : "2048,2048",
"gpu-engine" : "0-0,0-0",
"gpu-fan" : "0-95,0-95",
"gpu-memclock" : "0",
"gpu-memdiff" : "0",
"gpu-powertune" : "-20,-20",
"gpu-vddc" : "0.000,0.000",
"temp-cutoff" : "95,95",
"temp-overheat" : "85,85",
"temp-target" : "75",
"api-port" : "4028",
"expiry" : "60",
"gpu-dyninterval" : "7",
"gpu-platform" : "0",
"gpu-threads" : "2",
"hotplug" : "5",
"log" : "5",
"no-pool-disable" : true,
"queue" : "1",
"scan-time" : "10",
"scrypt" : true,
"temp-hysteresis" : "2",
"shares" : "0",
"kernel-path" : "/usr/local/bin"
}

Thnx i'll try this. Just had in running for about 10 min at 400kh/s each (yeah low) and then gpu 1 would turn SICK again

That's weird.

What are your temps when it gets SICK?

Also what psu are you using to power them?

Are you using them in windows or linux?

and what version of the driver and sdk you got installed?

Try lowering your thread concurrency if your PC keeps freezing. Reduce it by x64 steps.
To reduce your temp the best way is to undervolt them and also point additional fans at them.
Risers can also help.

Do they get enough power from your PSU?

Got an 850 watt psu and it tends to get sick around 80 degrees already.
Just got it out of store so it should have the latest drivers i guess..

850w should be fine, each card draws about 260w, a cheap processor draws about 65w, and you got a bit extra from hdd and ram, but not much. That only if it is 850w from a good brand...

Does the psu have several 12v lines or just one? Brand and model would be usefull...

Also if you are using linux and the sdk version you are using...

My shop said 700 should suffice so i figured i should get a little better one.
It's a ms-tech 850w ATX PSU. MS-N850-VAL

I'm using windows 8.

The temperature problem seems fixed, my pc will still freeze/crash after running cgminer fout about 1 min tho

Well that might be the issue. According to this web:

http://uk.hardware.info/productinfo/130941/ms-tech-ms-n850-val-850w#tab:specifications

That psu has two 27a 12v lines. Each line should provide enough power for 1 card. You sure you have each card on a separate 12v line?

Also I'm not quite sure that the psu can deliver what it says. Haven't found any stress tests in my quick search and nothing about its efficiency, voltage drop or if it has an active pfc. If I would have to bet the problem relies on the psu.

How hot does the psu case gets at the moment of the crash?

Can you measure the voltage while mining? grab a molex and put one end of the tester in the yellow and the other in the black...

Also to see how limited you are running the psu we would need to know everything that is connected to the rig. Processor, ram, periphericals, and other cards if you got any installed.

The other option would be that you test the gpus individually, I'm positive that it can handle one card if you convine both 12v rails. According to specs it should be able to handle 2, but I only trust a few psu brands to provide what they say.

Have you tried lowering the voltage bit by bit till its unstable, I have ASUS cards that come with 1.2VDDC lowering that to the 1.087VDDC to 1.1VDDC range doesn't affect my stability and lowers my temps by a decent amount, 6c for my hottest 15c for my coolest.

Well atm my whole system is unstable as soon as i put the intensity above 11 so i think undervolting would make it too unstable?
And when it's at 11 the cards are only getting 260kh/'s each. It should be alot higher shouldn't it?

thnx alot for all the answers moneymorpheus but i really wouldn't know if they're on seperate lines since my local computer store put all the pieces together for me..

Ok, it was defo the psu, it just broke lol..

ahww, hmm probably the psu is faulty, im running 3 280x underclocked with a 1100 gold psu, and 2 on 800 silver psu. probably you have a faulty psu, since i reckon 850 is enough to suffice 2 280x + system.

Ok I was right, a shame it broke though. For the next psu, try to get one that has only 1 big 12v line. Also for every psu I buy I google the model and review. I look for a good stress test and read how it went, the thing you gotta look for is how it handles a full load for a long period of time. You have to think that you will use the psu 24/7, a cheap one will fail on you.

pd: I'm running 4 280x under a 1200w gold coolermaster and it runs just fine. To run just two cards a good 850w psu will do.

I am planning costs for my rigs, and I am kinda confused about Your post. If 4 cards = 1200W, 2 cards = 600W. Why You said 850W??

2 cards = 600W, .............+ MB + CPU + HDD, etc = 800W

so then 4 cards = 1000W + + MB + CPU + HDD, etc = 1200W
250W per card, so

2cards = 500W + rest 700W :P

257w each card, a cheap processor uses 65w, a high end one 120-160w, each ram module 10-15, hdd 10, a bit more due to risers resistance. You can use a 1200w for four cards or a 1200w may fall short for 4 cards, it all depends on the rest of the rig.

4x Asus 280x using 990W at the wall, hashing @ 3Mh/s