A 6950/phoenix problem

[maurezen]

Greetings!

I've set up a single 6950 rig recently. It runs phoenix 1.48 as a dedicated miner, 6950 is on stock frequencies/voltage, connects to btcguild pool. It's temperature is steady 65 C.

The problem is, once every several hours, the monitor goes blank (vga signal lost) and if I connect via TeamViewer I can see gibberish like this. The mining also stops, but the computer is otherwise functional.

I've written a watchdog program which reboots the computer if my last submitted share is older than a certain amount of time, but that's just a workaround and I'd like to get to the core of the issue.

Have anyone encountered a similar problem? Have anyone got an idea of why does it happen?

Thanks.

[zipxavier]

I find that when my 6950 hovers around 65-66 the same thing happens to me. When I have it so it never goes above 64 it seems to solve the issue. I'm guessing it's some fail safe in the card to prevent damage..

[maurezen]

Hmm, sounds strange (afaik 65C isn't a high temperature by gpu standards).
Thanks anyway, I'll try to get it on 60-62C and check for stability.

[btcsquirrel]

I run 2 6950s and phoenix and I had some lockups until I changed the "aggression" flag to 7.  I had it on 11.  There was no drop in mhashes/second when I made the change and it's been running for awhile with no problems.  Post if changing that fixes the problem so I can feel good that I helped someone in the newbie forum. lol

[Zman0101]

I have three 6950's running in trip crossfire in one of my rigs.. My middle card runs at 92C and it runs 24/7. Sounds to me like you have a damaged GPU. Put your fan on 100% to keep it under 65C in ATI Overdrive and see if this keeps it stable. Rule out if its a heat issue.However, i doubt it Also make sure you are running over a 550watt PSU. Update your bios... Try a different PCI Express slot on your motherboard. Lastly try a different card. I have over 10 6950 cards total. 2 of them were defective when i got them. So it wouldn't surprise me if your shit was messed up.

[btcsquirrel]

Put your fan on 100% to keep it under 65C in ATI Overdrive and see if this keeps it stable.

Is that 65C under load or idle?

[Kermee]

I run 2 6950s and phoenix and I had some lockups until I changed the "aggression" flag to 7.  I had it on 11.  There was no drop in mhashes/second when I made the change and it's been running for awhile with no problems.  Post if changing that fixes the problem so I can feel good that I helped someone in the newbie forum. lol

This.

+1

Cheers,
Kermee

[spule]

OP, what model is your PSU and how many W?

[btcsquirrel]

Glad that did the trick!

[Salain]

You may also consider moving off of the VGA port, if possible.  I had some success reducing artifacts by moving to HDMI.

[bigcoiner]

I have two cards next to each other and one of them is running really hot.  If I overclock, that one goes over 100C.  Its fans are blocked by another card.  I have to use -f600 to keep the temperatures in the 80s, at the expense of lowering the hash numbers.

I have three 6950's running in trip crossfire in one of my rigs.. My middle card runs at 92C and it runs 24/7.

[btcsquirrel]

I have two cards next to each other and one of them is running really hot.  If I overclock, that one goes over 100C.  Its fans are blocked by another card.

The hotter of my two cards is kind of blocked by the other, too.  I experimented with air flow changes by holding a 120mm fan in various positions while watching the temp change.  I saw someone suggest that on the forums and it worked very well.

[maurezen]

Thanks for your replies!

I haven't yet had the time to try anything (I'll do something this weekend).

btcsquirrel: when I was toying with the settings, I tried aggression 7. phoenix reported performance was actually greater than with aggression 11, but btcguild stats reported much lower performance, so I didn't try that much further. Will give it a check once again.

Zman0101, btcsquirrel: it's steady 65C under load.

spule: it's Chieftec APS-700C, 700W.

Salain: thanks, will try.

[btcsquirrel]

Thanks for your replies!

I haven't yet had the time to try anything (I'll do something this weekend).

btcsquirrel: when I was toying with the settings, I tried aggression 7. phoenix reported performance was actually greater than with aggression 11, but btcguild stats reported much lower performance, so I didn't try that much further. Will give it a check once again.

Zman0101, btcsquirrel: it's steady 65C under load.

spule: it's Chieftec APS-700C, 700W.

Salain: thanks, will try.

The btcguild miner performance numbers can be way off (it says on their site that those are just estimates).  I've been using btcguild and noticed those numbers didn't match my software's numbers.  I've been on aggression 7 for a few days now and will no modification of the standard settings, each 6950 is getting a steady 310 Mh/s.  I was at about 320 Mh/s when I ran aggression 11, but I'm willing to make that small trade-off because my mining rig is my main gaming computer, too.

Adding another one of these today: http://www.newegg.com/Product/Product.aspx?Item=N82E16835185054

If I can get the hotter of my 2 6950s down to 75C under full load, I'll be a happy camper.

[sabe]

I have two cards next to each other and one of them is running really hot.  If I overclock, that one goes over 100C.  Its fans are blocked by another card.

The hotter of my two cards is kind of blocked by the other, too.  I experimented with air flow changes by holding a 120mm fan in various positions while watching the temp change.  I saw someone suggest that on the forums and it worked very well.

For me I squeeze a little piece of plastic between the 2 cards, around 1/4" thick to space out the card, that can bring down temperatures on the "blocked" card by easily 15 degrees.

I run 4x 5830 on a single board. Before this little "tricK", the hottest card was 91C+ (throttling) and the coolest was 58C. Now the hottest card is only 77C or so.

[btcsquirrel]

For me I squeeze a little piece of plastic between the 2 cards, around 1/4" thick to space out the card, that can bring down temperatures on the "blocked" card by easily 15 degrees.

I run 4x 5830 on a single board. Before this little "tricK", the hottest card was 91C+ (throttling) and the coolest was 58C. Now the hottest card is only 77C or so.

I might play around with doing something like that if my new high cfm fan doesn't do the trick.  If the new fan doesn't bring temps down a bit more, I'd say it's the radiant heat from one card effecting the other and a piece of plastic between them is exactly what I need to do.

To what is your plastic divider connected?  The case or one/both video cards?

[sabe]

i just squeeze it in between 2 cards near their power connectors. Its really small, like a thumbdrive cap or something.

[naypalm]

I'm getting about 320Mhps per card out of my HD6950's but my temp is hellacious at 85C+! I double sided sticky-taped an 80mm fan to the butt end of both HD6950's and it seemed to help keep it at 80C but only if the side is open. I'm hoping to add a few more fans in a day or so... these things crank out the heat!!!!

[Kermee]

I used the styrofoam peanuts that Newegg uses with their shipping as the spacers between cards. Works great =)

Cheers,
Kermee

[shulace]

Grab a PCIx16 flexible extender.  I currently have four moderately high overclocked cards running on my MSI 890fx-GD70.  Lol my cooling solution is to leave the rig in front of a small AC unit, temps are always under 72C.  Oh and the cards I'm using are a 6970 thats running @ 420mhash, a non reference locked shader 6950 @ 370 mhash, and 2 5850s @ 350 mhash a piece.

If a small AC unit is not a viable options, get a bunch of case fans! Keep your cards under 80C!