Topic: Dual 5970 - Unstable Hash rate - Was stable for 27 hours (Read 2627 times)
I bought one at home depot, turned the power off at the breaker, and used a screwdriver and a pair of pliers.
I know this is an old thread, so you probably don't have that problem anymore, but I had a similar problem, which I fixed by making sure the computer with fluctuating/varying hashrate had enough power from the outlet. I had two 1250w psu 4gpu miners plugged into one surge protector off a 15 amp rated outlet, and the hashrate went up and down on one of the miners.
Then the power strip tripped, so I replaced the 15 amp outlet with a 20 amp outlet (20 amps from the breaker) and plugged each computer into its own power strip. problem solved, hashrate stable.
Then the power strip tripped, so I replaced the 15 amp outlet with a 20 amp outlet (20 amps from the breaker) and plugged each computer into its own power strip. problem solved, hashrate stable.
Although I stopped mining nearly 1month ago due to a 5970 quite litteraly going up in smoke, I can verify that this solution does work: I had placed a dedicated 3kw UPS onto my mining rig (It's good to be friends with people who own data centers
) and the cleaner power stabilised my hashing rate by a substantial amount. 
How did you replace the 15 amp with a 20 amp?
I know this is an old thread, so you probably don't have that problem anymore, but I had a similar problem, which I fixed by making sure the computer with fluctuating/varying hashrate had enough power from the outlet. I had two 1250w psu 4gpu miners plugged into one surge protector off a 15 amp rated outlet, and the hashrate went up and down on one of the miners.
Then the power strip tripped, so I replaced the 15 amp outlet with a 20 amp outlet (20 amps from the breaker) and plugged each computer into its own power strip. problem solved, hashrate stable.
Then the power strip tripped, so I replaced the 15 amp outlet with a 20 amp outlet (20 amps from the breaker) and plugged each computer into its own power strip. problem solved, hashrate stable.
Strange update on this problem.
My system has been stable all night without any hickups, but performance was still bouncing between 1.1 and 1.4Ghash/s. Now I just went to my morning job of looking after puppies (HAHA, it's money is it not
) and I remote desktop'd into my machine, and as soon as I connected the performance levelled right out, power draw became stable (Was going from 65Amps to 9Amps and then back up every second or two.
All I can think is that there is something on my system using some GPU power, that closes/stops when I remotely connect: You probably just thought the same thing I did, Windows 7 aero and it's partners. Well my connection to the machine is fast enough that I can get real time access and don't disable anything like aero on connection.
It will be interesting when I get back home to see if the system has lost 100% stability again.
My system has been stable all night without any hickups, but performance was still bouncing between 1.1 and 1.4Ghash/s. Now I just went to my morning job of looking after puppies (HAHA, it's money is it not
) and I remote desktop'd into my machine, and as soon as I connected the performance levelled right out, power draw became stable (Was going from 65Amps to 9Amps and then back up every second or two.All I can think is that there is something on my system using some GPU power, that closes/stops when I remotely connect: You probably just thought the same thing I did, Windows 7 aero and it's partners. Well my connection to the machine is fast enough that I can get real time access and don't disable anything like aero on connection.
It will be interesting when I get back home to see if the system has lost 100% stability again.
Try reinstalling with the last version of CCC and not the current package. Use DriverCleaner / DriverSweeper to clean up the registry after you uninstall, before installing a different driver set.
Got my core at 900Mhz with a Vcore of 1.137v. Memory is 600Mhz. GPU temp is 72c, and the VDDC regulators are at a toasty 105c but stable.
I tested going back to stock settings but it made no difference at all, I think something has changed in CCC that is making the ATI driver try to balance the load on the GPU's. I tried setting Cat AI to standard, advanced and off but it makes no difference at all.
I tested going back to stock settings but it made no difference at all, I think something has changed in CCC that is making the ATI driver try to balance the load on the GPU's. I tried setting Cat AI to standard, advanced and off but it makes no difference at all.
Still a good thing to check. Especially if you're attemping more than marginal overclocking.
I know that my temps between cores are not similar at all and so I have each core configured to suit it's individual tolerances.
I know that my temps between cores are not similar at all and so I have each core configured to suit it's individual tolerances.
(Make sure your cores are not linked/synced in MSI Afterburner as well. You'll need to set individual core voltages and core speeds for each. You'll have lockups if they're linked. This is under MSI Afterburner settings and it is where you'll switch between core profiles.)
Mine are all linked and I do not have any issues.
Sorry the pics are misleading. I was running 750Mhz when I overheated, which was down to me forgetting to put manual fan control back on.
Hate when that happens.
Hm, well, I'd reset back to stock 725mhz and stock voltages, with 100% fan, and see if it still occurs?
Other than that, try reinstalling drivers, though that's a long shot.
I've had that issue with the display driver 'resetting' and I only have ever seen it when I was trying to squeeze too much out of too little voltage on either of the cores.
(Make sure your cores are not linked/synced in MSI Afterburner as well. You'll need to set individual core voltages and core speeds for each. You'll have lockups if they're linked. This is under MSI Afterburner settings and it is where you'll switch between core profiles.)
Sorry the pics are misleading. I was running 750Mhz when I overheated, which was down to me forgetting to put manual fan control back on.
Still getting the same results with 850mhz at 1.1v (1100mV).
Still getting the same results with 850mhz at 1.1v (1100mV).
Your voltage is also likely why your system initially crashed.
You're running 850mhz with no increase in voltage
You're running 850mhz with no increase in voltage
http://i66.photobucket.com/albums/h271/KoFsnedie/5.jpg
http://i66.photobucket.com/albums/h271/KoFsnedie/4-1.jpg
http://i66.photobucket.com/albums/h271/KoFsnedie/3-1.jpg
http://i66.photobucket.com/albums/h271/KoFsnedie/2.gif
http://i66.photobucket.com/albums/h271/KoFsnedie/1.gif
Increased Vcore to 1100Mv. No difference
http://i66.photobucket.com/albums/h271/KoFsnedie/4-1.jpg
http://i66.photobucket.com/albums/h271/KoFsnedie/3-1.jpg
http://i66.photobucket.com/albums/h271/KoFsnedie/2.gif
http://i66.photobucket.com/albums/h271/KoFsnedie/1.gif
Increased Vcore to 1100Mv. No difference
You're undervolted for the amount of core clock mhz you're running. You either need to increase voltage or reduce speed.
So my system overheated this afternoon whilst I was out and shut itself down. Now when I start mining, the display driver will crash and then recover but my miner will happily chug along. However, the stability has now decreased and the amount of load on the GPU's is varying quite a bit.
Now I know what your thinking, it's probably still overheating. Well it's not any more, I have a desk fan drawing air from an open window directly into the side of my case and onto the cards. I have quite litterally shaved off 20c from my previous temps.
I can't find a reason for this behaviour, I've attached some screen shots for you to look at. Maybe one of you will spot something.
Thanks
Now I know what your thinking, it's probably still overheating. Well it's not any more, I have a desk fan drawing air from an open window directly into the side of my case and onto the cards. I have quite litterally shaved off 20c from my previous temps.
I can't find a reason for this behaviour, I've attached some screen shots for you to look at. Maybe one of you will spot something.
Thanks
Jump to: