Author

Topic: Can not get miner to see other 5970.. help? [solved... kinda] (Read 3120 times)

hero member
Activity: 696
Merit: 500
So im getting some dropouts on a random gpu every few minutes. Just a quick blip but it drops to bottom. My card temps are never getting higher than 80 and are in the 60's on 2 of them and mid 70s on the other. My V temps are in the 80's and 90s. I don tknow why im getting these drops. do you get any?
full member
Activity: 143
Merit: 100
If it goes on, I think I will try to pinpoint the problem by slowly changing things rather than slowing everthing down and going back to what worked before which is 0.3.19 and an older poclbm.  Let us know how many blocks a day you get since we have the same setup going now.

Quoting myself!  Wonderful.  With no more bitcoins awarded and over 95% probability of a hit unrealized, I chickened out and went back to pool mining.  Blocks are accepted, who cares?  I've never been real luck in games of chance anyway.  Chalk it up to bad luck and move on.
hero member
Activity: 696
Merit: 500
I've got two 5970 cards (you have 4?! like 2.5ghps?) and have been only mining in the pools. I'm averaging 1.3 btc per hour. I havent yet tried going solo as the payout would be nearly the same I believe.  I'm having trouble figuring out the payout per hour on slush's pool right now. Maybe you will hit two blocks in a day? that would be pretty sweet.
full member
Activity: 143
Merit: 100
so stable at ~315ghps on all 4 cores for about an hour. as long as i dont use the -f switch on poclbm.

My settings are in Post 8 above.  Control of the fan is often a problem in Afterburner.  Gamers and other overclockers talk about it and the interaction with Catalyst Control Center ("CCC").  Sometimes the slider does not have control over the fan.  Other times its stuck on auto and other times you can't get auto to turn on.  Don't know precisiely what the problem is, but the problem seems to settle down with time, just like getting the cards to work together was hard at first, but probably works every time now.  I have my fans always above 70% and sometimes closer to 100% becuase cooling affects hash rate even if the temps shown in MSI doesn't change much.  315 is pretty good if they are all stable at that rate.  Often one of the 4 GPU's will bounce around a bit because that one also controls what it going on with your screen.  My max speed is about 330, but at that rate the one erratic GPU becomes more erratic and I have theory that it is better to be a little lower and stable on all four (flatline) than to have higher speeds with periodic drops down to the low 200 range.  Anyhow, sounds like you got it working just in time for the Bitcoin price crash and corresponding 30% difficulty increase!  Just kidding.  Kind of.  One question of my own . . .

I have been doing this for a couple of months.  I am familiar the the Poisson distribution, the average rates I should be getting, what the probabilities of "average", etc. mean.  Having said all of that, I feel I am no longer getting the averages I should when everything appears to be running fine.  I have gotten the equivalent of ONE block over the last four days, when I should be getting a block almost twice a day with 4 5970's.  I think if I don't have a block by tomorrow morning it approaches statistical impossibility.  In other words, the usual "just be patient" does not apply here.  If it goes on, I think I will try to pinpoint the problem by slowly changing things rather than slowing everthing down and going back to what worked before which is 0.3.19 and an older poclbm.  Let us know how many blocks a day you get since we have the same setup going now.
hero member
Activity: 696
Merit: 500
so stable at ~315ghps on all 4 cores for about an hour. as long as i dont use the -f switch on poclbm.
hero member
Activity: 696
Merit: 500
My gpu temps wer all about 84-90 overclocked.. I just found out my fans were running slow. I bumped up to 100% and things cooled down a lot. 68-70 on all cores.
hero member
Activity: 696
Merit: 500
Ronald ive been stable for about 11 hours on stock settings. Gpu says most of my vccd temps are 70-85, but on one of the 4 listed gives me temps of up to 95c. what are you running at?
full member
Activity: 143
Merit: 100

I've spent countless hours trying to figure out how to get things going in win7! Thanks so much!!

I tried this withe the f5 option and got 4 instances of 305mhps but it became unstable and crashed the system Sad

whats your btc address...

You don't really need to send me a bitcoin.  Just jokingly making the point to all in the TechSupport forum that we all compete with eachother.  As to crashing and/or freezing, I don't believe it has anything to do with the fact that you don't have the monitor plugged  into the second card.  First off, you can verify in Afterburner which card is crashing (or just see the one blowing cold air) and start over and plug the monitor into that card when everything is running.  That makes it active, but it doesn't matter.  Windows does not need for it to be designated active for the program to continue to run.  I have run for months on such a setup so I would consider the following:

You are overclocking to get 305mhps.  Try stock until you are satisfied.  Then play with the settings, which I will give you for one more bitcoin :-)  Nevermind, the settings are 1100, 850, 1025.  Start with the clock no more than about 800 for starters.  I leave the core voltage lower, as Dingus says, because it can run 800-850 and gives a little cooler VRM(?) temps.  These are temps you don't see in Afterburner and need to look at Sensors tab in GPU-Z. 

I have found that what happens is, even if your temps are okay and about 90 degrees, the card will begin to falter with the program as you increase heat (either just warmer case, or heat increase by voltage increase).  There begins to be dropouts when it looks like a heartbeat monitor in Aterburner.  You want a flat line, even if it is a lower flat line, becase each time there is a dropout (like a heartbeat) it costs you dearly in hashes for the duration (on average you get LESS than by going a little slower/cooler).  The dropouts happen at higher heat.
hero member
Activity: 696
Merit: 500
Ok well this isnt quite sop flawless. The second card stops working somwhere between 15 - 60 minutes for some reason.
hero member
Activity: 696
Merit: 500
I've tried this before. I've tried connecting additional monitors as well.

Have you been able to do this with two ati 5970's in win7?

OK I screwed around for the better part of 20 min and figure it out.

I have BOTH 5790 cards working in windows7 now.

I found the way to do it every time.

Start the an instance of your miner on each thread. Then unplug your cable and plug it into one of the slot on the second card. Let the monitor come back on then plug it into the other slot on the second card. Let the monitor come on then plug it back to the original slot. (this doesn't seem to make a difference)then start your miners again using the same thread value like you said!!

You might have to do this 10 times before it works unfortunately. Closeing everything then then restarting the process.


I've spent countless hours trying to figure out how to get things going in win7! Thanks so much!!

I tried this withe the f5 option and got 4 instances of 305mhps but it became unstable and crashed the system Sad

whats your btc address...
full member
Activity: 143
Merit: 100
You likely have an earlier motherboard that does not support the crossfire, but not to worry.  You do not need the crossfire in your situation.  I have an earlier ASUS A8N32-SLI with this problem.  Plug the monitor into one card, either port..  That enables the CARD in CCC.  Run the program.  Promptly switch the monitor cable to the other card.  That enables the other card on the SAME device numbers but does not stop the program from running.  Now re-run the program.  If the windows just sit there blank after a few seconds, close ALL windows and restart the program.  Now that will run the program on the current card and you can see in Afterburner that such is the case.  Switch the cable back to the other card and try again.  Keep repeating until you have all four windows running.  If this doesn't work after five or six times switching, restart the computer and try again.  If that doesn't work, reinstall Stream SDK.  Have patience.  It will work.  Now send me a bitcoin from the first block generated on the additional card since I have just advised you on how to compete with me.
full member
Activity: 126
Merit: 100
I have the crossfire cable on my 5970 on windows 7 64bit, and it works with poclbm.
hero member
Activity: 696
Merit: 500
neg device 2 appears to be cpu. hmm How quickly could I get linux up and running? I already run two installs oif windows on this pc.
hero member
Activity: 840
Merit: 1000
You are correct in not using the crossfire cable, you should also make sure crossfire is disabled in the drivers.  However, I don't know if it is possible to get windows to detect multiple 5970s for opencl.  I wasn't able to get it to work when I tried (this was back in november, SDK 2.1, Catalyst 10.10).  You probably want to use linux, as it supports up to 8 GPUs for opencl and is what most people around here use for mining.

Device 0 is probably your CPU.
hero member
Activity: 696
Merit: 500
Hey guys got two 5970's. I do not have the crossfire cable on themn do I need to? I tried it with it on with no luck.

when I tel miner to use device 1 and 2 they each use a core, but if I tell it to use 3 it cant find it and 0 uses hlaf of one of the other cores..

Any ideas?

win7 64

Jump to: