Nearly bricked my 6990.

[ruski]

A cautionary tale for OCers..

So, I'd been running it for a week straight. OCd with AB, 940MHz core oc/820MHz mem uc. Stable at 80C per core. Using CCC to boost the power to each core by 4 and 5 pct, 0 and 1 respectively.

Anyway, I decided to take the memory down further. Unlocked AB and took it to 700. No worries. So then I tried 600. Only I didnt type 600, I typed 300.

Artifacts and malignment everywhere. Then it hung. So I turned it all off and went back to BIOS 2, restarted. Soon as windows loads  - fan at 100%, frozen. Restart. Uninstall CCC, start again. Now it dies during install. Im tearing my hair out.

I figure CCC has the power settings somewhere that didnt uninstall. So instead of doing it again, I manually delete every ATi and AB file and registry. After which, run uninstall again, and reinstall. With baited breath I waited....

And it worked! Stable, CCC runs OK. (I didnt mention - the only way to start outside safe mode was to open taskmgr and kill MOM.exe before it could do anything.)

So, I start mining again, running AB and CCC. I restore the power settings, leave the memory at a safe 820MHz uc. D'oh. I didnt disconnect the DVD drive I connected to reinstall drivers. Power down, take it out, reboot.

DOA.

I now have deny all user access to both instances of MOM, and CCC. I'm fairly sure the card memory is fried now, or at least damaged. Its mining at a happy 810Mhash/s right now, I'm scared to game with it though.

Moral of the story? WATCH YOUR GODDAMN FAT FINGERS. 

[1713512112]

1713512112

[1713512112]

1713512112

[1713512112]

1713512112

[1.21gigawatts]

You can't brick your card from underclocking the memory too much.
DOA means Dead On Arrival or Defective On Arrival.  Can't be DOA if you got it out of the box brand new and it's working

[Jack of Diamonds]

You can't brick your card from underclocking the memory too much.
DOA means Dead On Arrival or Defective On Arrival.  Can't be DOA if you got it out of the box brand new and it's working

lol. I was thinking the exact same things.

At worst a driver could get corrupted as the entire system freezes when you downclock the memory too much. The card can't die from any amount of downclocking, hell, even excessive undervolting will just freeze the system.

Screen artifacting after changing clock frequencies doesn't indicate bricked memory or a broken card, that happens every time when components are run at specifications not suited for operation, which means you have to tweak them

& A DVD drive can't be dead on arrival if you have been using it...

[fitty]

A cautionary tale for OCers..

So, I'd been running it for a week straight. OCd with AB, 940MHz core oc/820MHz mem uc. Stable at 80C per core. Using CCC to boost the power to each core by 4 and 5 pct, 0 and 1 respectively.

Anyway, I decided to take the memory down further. Unlocked AB and took it to 700. No worries. So then I tried 600. Only I didnt type 600, I typed 300.

Artifacts and malignment everywhere. Then it hung. So I turned it all off and went back to BIOS 2, restarted. Soon as windows loads  - fan at 100%, frozen. Restart. Uninstall CCC, start again. Now it dies during install. Im tearing my hair out.

I figure CCC has the power settings somewhere that didnt uninstall. So instead of doing it again, I manually delete every ATi and AB file and registry. After which, run uninstall again, and reinstall. With baited breath I waited....

And it worked! Stable, CCC runs OK. (I didnt mention - the only way to start outside safe mode was to open taskmgr and kill MOM.exe before it could do anything.)

So, I start mining again, running AB and CCC. I restore the power settings, leave the memory at a safe 820MHz uc. D'oh. I didnt disconnect the DVD drive I connected to reinstall drivers. Power down, take it out, reboot.

DOA.

I now have deny all user access to both instances of MOM, and CCC. I'm fairly sure the card memory is fried now, or at least damaged. Its mining at a happy 810Mhash/s right now, I'm scared to game with it though.

Moral of the story? WATCH YOUR GODDAMN FAT FINGERS. 

You didn't almost brick it. You could put in -10 for memory and it's fine. You could put in 50 for memory and it's fine. You can't brick it by setting the core/mem with CCC. Worst case you uninstall/reinstall. =)

[ruski]

No no no..

By DOA, I meant on arrival to the system, ie. startup. Poetic license here, people. And there was nothing wrong with the DVD drive, it's just awkward to connect with my setup.

Since I have done the whole uninstall/reinstall and I'm now back to where I started, are you sure I couldnt've bricked it? Could the overvoltage + the lowered memory clock have damaged the memory? Everywhere I look, massive distortions + freezing is said to be caused by faulty memory.

Why did it previously run stable at the same power, memory & core settings I had before, but with the same settings now freezes, if the memory isn't damaged?

[mrblu]

I get artifacts/crashes when I go below 400mhz mem but 405mhz on the other hand works fine.

OC 944mhz core/405mhz mem stable at ~67degrees. Note that the current ambient temperature is about 11C which is about 58F

[Jack of Diamonds]

Why did it previously run stable at the same power, memory & core settings I had before, but with the same settings now freezes, if the memory isn't damaged?

Trying a shotgun diagnosis here.. Did you lower the fan speed at any time?

Ages ago I just had a single 5850 which I OC'ed & mined with that.
A fan speed below 50% would make it freeze as the VRM's were not getting enough cool air. At 75%, no freezes whatsoever.

[haploid23]

i don't think a memory underclock will kill your card. it's gotta be something else

[ruski]

Nope, the fan has been at 81% for 0-79°C and 87% for 80°C+.

Anyway, it's running stably now at 945 MHz, without the overvolting (as far as I know, AB won't show voltage, it may be stuck at the higher power regardless). Guess I can do without CCC. Thanks though.