speeds slowing on 6990's

[Kaliecious]

i've got 2 radeon 6990 that i've been using to mine, here re4cently I have noticed 370-400mhash on 1 gpu while tthe other gpu will do about 250-280 Would this be a PSU or a heat issue? on the other card the the speeds are in the 225mhash area. all setting are the same and at one time both cards would scream 370-400 on all gpus.  any ideas would help as im a little stumped.

[frozenkai]

i've got 2 radeon 6990 that i've been using to mine, here re4cently I have noticed 370-400mhash on 1 gpu while tthe other gpu will do about 250-280 Would this be a PSU or a heat issue? on the other card the the speeds are in the 225mhash area. all setting are the same and at one time both cards would scream 370-400 on all gpus.  any ideas would help as im a little stumped.

Hello,
Welcome to the forum!
Could you clarify on your setup?

[Kaliecious]

Sure, its 2 radeon 6990 oclocked to 915 ram 625 I think it is, i would generally get 370-400mhash per gpu, then I started noticing the speeds on gpu2 would start lagging back to 250 or so, eventually 3 and 4 seemed to do the same thing with gpu1 holding that speed, This started happening on a water cooled systhem, I recently noticed I kept getting air bubbles in the system and thought it might be a heat issue do to lack of circulation. So I yanked it all out today and put them in air cool as I keep the room at a constant 70degree. The same issue seems to have lagged over to air cooled, The cards are pretty close right now in the system and the psu is a 1600watt so im pretty sure I have enough power not unless something is going shits caprits on the psu. So im to a heat factor at this time.

I have found very few articles on this, but the ones I have been lucky enough to find with my old ball shearches are from here so I will proceed to install 2.3 catalyst which I have read has helped people with simular problem otther then that there have still been some I found that have no definiative cure for my problem.

now its like they will all start out about 380-400 and they will all slowly go down in Mhash maybe low 300's then 250, then high 100's and so on untill I get a connection problem with GUIminer.

oh the system is a 6 coure 1100t with 1 or 2 gigs of ram 1600psu.

[Dum3]

Have you tried to increase your power limit to +20%?

[Kaliecious]

Have not tried that but ive always assuemd raising that would be a main contributor to a short lifed card. both the cards have the OC switch flipped on them and then speed set at 915, I went off vid I found on youtube a year ago. speeds have been good till here recently.

Rolled back to 12.3 catlyst driver set Fired it up, it ran for about 1-2 mins and whole system shut down, restarted and tried one more time and same thing.

any how, just got in from the work, time to trouble shoot it more if I can. LOL.

[Dum3]

actually no,increasing power limit means that the maximum power used by the card be higher,simple definition is that your card will throttle at higher power output,it will not increase voltage or anything.Just try it and see.

[Shadow383]

Have not tried that but ive always assuemd raising that would be a main contributor to a short lifed card. both the cards have the OC switch flipped on them and then speed set at 915, I went off vid I found on youtube a year ago. speeds have been good till here recently.

Rolled back to 12.3 catlyst driver set Fired it up, it ran for about 1-2 mins and whole system shut down, restarted and tried one more time and same thing.

any how, just got in from the work, time to trouble shoot it more if I can. LOL.

Post a screenshot from HWmonitor - mining with the full OC settings on + some heatsink dust over time could mean your cards are throttling due to high core or VRM temperatures.

[omegaaf]

My best explanation for this phenomenon is bus speed. While you may have a high end motherboard, the 6990 actually hits the limits of PCI-e 2.0. And running them in crossfire means you have to share the pipelines on the motherboard with another card. Most motherboards operate at 16x for the primary slot, but the secondary usually only runs at 8x, so you are pushing the same amount of data through half the pipelines.

[Shadow383]

My best explanation for this phenomenon is bus speed. While you may have a high end motherboard, the 6990 actually hits the limits of PCI-e 2.0. And running them in crossfire means you have to share the pipelines on the motherboard with another card. Most motherboards operate at 16x for the primary slot, but the secondary usually only runs at 8x, so you are pushing the same amount of data through half the pipelines.

Not relevant for mining. For mining purposes a 6990 will sit happily on PCI-e 1x.

[omegaaf]

My best explanation for this phenomenon is bus speed. While you may have a high end motherboard, the 6990 actually hits the limits of PCI-e 2.0. And running them in crossfire means you have to share the pipelines on the motherboard with another card. Most motherboards operate at 16x for the primary slot, but the secondary usually only runs at 8x, so you are pushing the same amount of data through half the pipelines.

Not relevant for mining. For mining purposes a 6990 will sit happily on PCI-e 1x.

Indeed, however a computer only works as fast as its slowest part, and in 90% of computers these days its the bus.

[Kaliecious]

ok So I came home, I took out the third card that was not running in the system and splitt the other 2 apart to give more airflow, reformatted cause one time when I was checking out amd catalyst I noticed that the speed was dropping from 915 to 500 and would sometimes drop to either 150 or 250.

after the format I put in cataylst 12.8 driver sdk 2.4 cause I've read that is the best to use. strangly its all working again im pussing 1.5-1.6 gh for those 2 cards air cooled. its been running about 20 minutes and no drop in speeds on any of the gpu's. So im figuring this was either a heat issue or a corrupt driver issue, due to the fact I noticed after uninstalling and reinstalling drivers 2-3 times after every install the 915 setting stayed instead of it going to default factory settinsgs.

Which ever it was i am not sure but its a tootling along now. I would be interested in finding out if there is ways to push the cards to faster speeds or if I am better off leaving them where they are at the 1.5-1.6 I have been running gui miner and at one time was running pheonix 1.5 does the new pheonix 2.0 help to get more out of 6990's or should I keep with the gui.

Thanks for the imput peoples

[Fiyasko]

ok So I came home, I took out the third card that was not running in the system and splitt the other 2 apart to give more airflow, reformatted cause one time when I was checking out amd catalyst I noticed that the speed was dropping from 915 to 500 and would sometimes drop to either 150 or 250.

after the format I put in cataylst 12.8 driver sdk 2.4 cause I've read that is the best to use. strangly its all working again im pussing 1.5-1.6 gh for those 2 cards air cooled. its been running about 20 minutes and no drop in speeds on any of the gpu's. So im figuring this was either a heat issue or a corrupt driver issue, due to the fact I noticed after uninstalling and reinstalling drivers 2-3 times after every install the 915 setting stayed instead of it going to default factory settinsgs.

Which ever it was i am not sure but its a tootling along now. I would be interested in finding out if there is ways to push the cards to faster speeds or if I am better off leaving them where they are at the 1.5-1.6 I have been running gui miner and at one time was running pheonix 1.5 does the new pheonix 2.0 help to get more out of 6990's or should I keep with the gui.

Thanks for the imput peoples

Go into your registry, And REPEATEDLY search for ALL entrys of "enableULPS" set it to 0,
ULPS is Ultra Low Power State.
Perhaps at somepoint the drivers got "touched" or the comp was restarted, ANNYWAYS
ULPS will change and Mess with your clocks, And in Majority of cases crash the entire computer
However, Touching the ULPS settting is only neccesary for people with 2cards or more

[Kaliecious]

I think its deffinantly a heat issue.. just saw your post on the enableULPS I will do that here shortly.
after seeing this and reading another forum about someone getting 745 and seeing 69c i sorta figured my shit might be getting to hot. lol

[Dum3]

Well,they are hot indeed,tweak the voltage and lower it at minimum

[Fiyasko]

Well,they are hot indeed,tweak the voltage and lower it at minimum

you call 69° hot? I run duo 6990 system at 86°C fans at 54%, My 6990's are rated for 110°C

[Kaliecious]

no sir not calling 69 hot, Im calling 100c hot. which as you can see from the pic two of my gpus are doing or right at it. I called msi they said they have tested up to 100c but don't know much after that.

[Shadow383]

no sir not calling 69 hot, Im calling 100c hot. which as you can see from the pic two of my gpus are doing or right at it. I called msi they said they have tested up to 100c but don't know much after that.

Get MSI afterburner, undervolt a bit and set up a custom fan curve to try to keep all GPUs below 80.

[peepee]

submersion cooling ftw

[MadHZ]

no sir not calling 69 hot, Im calling 100c hot. which as you can see from the pic two of my gpus are doing or right at it. I called msi they said they have tested up to 100c but don't know much after that.

i think 6990 is never going 100 - it will be in trottling process after 95C.. 69-80 is OK..