Quote from: Ratler00 on March 10, 2018, 12:56:52 AM

Quote from: Piskeante on March 09, 2018, 11:55:05 PM

Quote from: BennyT on March 09, 2018, 07:56:01 PM

Quote from: Ratler00 on March 09, 2018, 06:32:07 PM

Hi, I have two Founder Edition 1060 6gb cards with Samsung memory. Following some other guides I have seen I have mine set in Afterburner as:
Power limit - 65
Temp limit - 87
Core Clock - -150
Memory Clock - +700
Fan - 65

If I increase the memory clock Claymore is showing GPU errors and today I am finding what I thought were these stable settings are causing one of the GPU's to throttle. My electricity cost of $0.18 so I have to be a bit careful. Has anyone got any ideas how I can get more out of these as the general consensus seems to be that with Samsung ram the memory clock should be able to go much higher.
Thank you

All my samsung memory cards go to +900 and I run them at +850 on 62-64% power. Try it with the core clock at 0. I never adjust my core clock despite what others say because I find there is no gain by doing so. I also don't set the fan at 65 as I prefer the AfterBurner curve but that shouldn't affect anything.

There is a chance your specific card may require more power. As a general rule I keep my power at 75 when testing to make sure it can handle +850. When I know there are no errors, I slowly decrease the power until I find the error point and then run it at 1 or 2 points higher than the minimum. My Founders edition run at 64% and my Zotacs can handle 62%.

afaik power limit in Nvidia cards only refers to how much voltage is given to the core, not the ram. In fact, when you increase that power limit, the core clock goes up aswell. If that was the case that more power limit gives more ram "room" to hit higher frecuencies, ram speed would also run faster as you move the slider, but that does not occur. Moreover, i can garantee that even increasing to 116% the power limiit, will make no difference in how much can i push the ram.

Thanks for this. You were right when I reduced the core to 0 and increased the memory clock the hash rate increased. Interestingly though one card will take +850 and the other +700 at most. This seems odd.

Yeah, I tried forcing constant voltage on my 1060s, changed the default volgate curve in AB also. It doesn't affect your memory stability in any way. It only affects the core clock in a smart way, you better don't play with voltages on nVidia GPUs, power limit does the job for you.

Let's say your GPU runs at 1500 Mhz core clock at stock (+0 core) and reduced power limit to 60%, that's at about 0.8mv. By increasing your core to +100 you don't force the GPU to draw more power, you only force it to run at 100 Mhz higher - 1600 at the same voltage and power draw. If your GPU runs 1400 stock then +100 will make it run at 1500 etc. You basically set some extra Mhz at the same voltage and power levels. Simply said you get the full potential of your GPU kind of free.

Some GPUs can handle +200 even 250 others can't take even +100. You should carefully play with your settings and see whats the maximum stable speed.

If you are concerned about electricity then work with the power limit as well and see how low you can get it without losing too much performance.

My 1060s run stable at 60% power, around 0.800mv, some of them run at 0.762 even. Losing about 1% hash rate compared to 100% power limit.

Martobtc

Posts

Merits