Bitcoin Forum

Hi all,

my two mining rigs were running flawless for two weeks but lately it occurs more and more frequent that two workers for always the same GPUs just stop hashing. The console kind of freezes with the last Mh/s rate it was working with, but the cursor at the end of the line is still flashing.
This is how it looks like:
http://img14.imageshack.us/img14/923/phoenixnu.th.jpg (http://imageshack.us/photo/my-images/14/phoenixnu.jpg/)

What makes me wonder is why the GPU usage of the other cards goes down as well when the third card fails. Afterburner updating interval is set to 10 seconds and I can see this usage drop on the other cards in the graph every time, so it must take at least some seconds.

I'm using win7/x86, ATI 11.6 + APP 2.4, phoenix miner with phatk kernel, settings: -k phatk DEVICE=2 VECTORS BFI_INT WORKSIZE=256 AGGRESSION=12 FASTLOOP
I also altered the phatk kernel with the formula given in this forum to improve the hash rate. I had some issues with this on our first machine, one miner chrashed the same way every two minutes with this change, but it also happened once just before I made the change so the issue is not cause by this but it makes it more likely to occur.

Anyone knows what is causing this behaviour?

Just FYI, it's the driver that is crashing and recovering. I still don't know what is causing this though...

I see something like this on Linux too, with one of my two 5850s (but not the other). Unfortunately when it happens on Linux, restarting the computer is required. I suspect a less-than-stable overclock, but I'm still messing with it.

When my driver crashes and recovers, it's from an overclock that was too high.  Back it down 1-2 MHz and try again

BUMP I am having this same problem!

I have three identical rigs with 3x5870s.

1 is perfect, never crashes. Can go for a month without a reboot.
1 I have to restart every 3 to 4 days,
1 I have to restart everyday.

After trouble shooting for weeks, these are the things I found help, but still not solved completely.

1. If the temp on any of the cards is over 80C the chance of the driver/card crashing is much higher.

2. Make sure the display card (the one your monitor hooks up to) has with an aggression level of 9 with fastloop on, then Max out the other cards with an aggression level of 12ish. Don't max out your display  card, give some space for the OS.

3. If your Mhash reading jumps around all the time the system WILL crash. Sometimes I reboot the machine and Mh/s are stable. Sometimes they jump around by 30 to 50 Mh.  This may have to do with the power supply.

4. You might have a screwy card. I replaced one card that was jumping around in terms of Mhash performance and now that rig is fine. Just because your system boots and the card can start mining doesn't mean the VRMS are not burnt out.

i`m using clocktweak to monitor the gpu load every 5 min (via a batch script) and if load is under 60% (in my case) i`m killing/restarting that specific miner