Bitcoin Forum

So recently, I've been seeing a lot of people talking about how high of clocks they are mining at, and how much voltage they are pushing into them to hit those clocks.

I decided to do a test on my own system with 3, reference 6970's in it. The results are rather surprising.

All the wattage's are measured with my UPS and are from the wall while running guiminer on 3 gpus.

System IDLE wattage is about 350watts from the wall.

Default Voltages(1175MV), 880 is stock core clock.
880/1375: 905 Watts
880/685: 835 Watts
880/340: 782 Watts
~385MH/s per GPU

Default Voltages(1175MV) (Highest overclock without changing voltage)
940/340: 815 Watts
~420MH/s

1235MV
970/340: 915 Watts
~430MH/s

1300MV
1000/340: 1050Watts
~450MH/s

As you can see from the above results increasing the voltage greatly increases power usage. So much so that it will cause a negative impact on your BTC earned, even though you are creating more MH/s. For me with 6970's it seems that the sweet spot is the highest overclock without raising core voltage. This should apply to any other GPU also.

Now, one more thing is that my video cards are water cooled, which means they will run much cooler than air, and cooler power components use much less power than hot ones. So if you are on air cooling your power usage from overclocking/over volting could be much higher.

System Specs:
Intel Xeon W3520 @ 4ghz
Asus P6t7
3, 6970's
12GB 2000mhz ram
15 Case fans
Corsair AX1200
2, 120GB OCZ Agility SSD
3, WD 1TB blacks
Asus Xonar STX
https://i.imgur.com/mpKQIm.jpg (http://imgur.com/mpKQI)

I was expecting some pics of GPU's all melted or something :(

Great info though! :)

Good info.  I need to set mine back!

Also goes to show why you should underclock the memory too - 123 watts saved that way, not to mention ~430 BTU/h of heat :)

Was expecting to see blown up cards lol.
Iv been measuring mine with my kill-a-watt Idle power has been subtracted which was 129W, Full load with just CPU 203W.

6850 (775mhz stock)
all stock 153W 202Mhash/s
960mhz core 155W 266Mhash/s
1.22V 1000mhz core 158W 272Mhash/s
1.22V 1000mhzh core. 350mhz mem 155W 278Mhash/s
Temp change wasnt much.
Stock 67C
OC 70C
Voltmod 73C
Voltmod with mem uc 72-73C
Using CGMiner all the other ones use 100% cpu on 1 core but they get 4-5Mhash more but its not worth the heat and power the cpu uses when CGMiner uses 0-1% cpu.

Still going to push for a higher OC but stability testing this clock and voltage at the moment.

Interesting results.

Even aware of this, it's a difficult decision, since if you could not mine more coin this week, you will have to pay more electricity to mine the same amount of coin next week, as difficulty rises

I dont know if this is entirely true...I thought components use the same amount of power no matter what temperature they are at...

How do you set your memory to 350?  I use MSI's Afterburner and I can only go down to 685.

Heh.. I'm undervolting a 6850... So far runs stable at 1062mV, core 933MHz/memory 300MHz.

Power draw does theoretically depend on factors such as temperature (electronic components aren't ideal and their parameters can fluctuate) but the differences would definitely be much too small to notice.

Set it to 685, close, open, set it lower.

Or use Sapphire Trixx and you don't have to mess with Afterburner anymore...AND it can overvolt multiple cards.

can you only use trixx with sapphire?

No it works on other cards.

Thanks for job

Yeah same here. glad it wasnt though

You might be surprised by how large the difference can actually get...

Take a look at this. Power usage vs temps on a 5870.
http://www.silentpcreview.com/article1103-page5.html

Scroll down to the graph...

Basically, stock cooler @ 87c used 415watts from the wall. With an aftermarket cooler keeping the card at 66c it dropped power draw to 386watts from the wall, a 29watt difference. That's also only with a 21c drop in temp. I went from 98c to 45c, a 53c drop, and I did see a large difference in power draw from the wall of about 100+watts IIRC. We can also rule out the different fans for the power draw usage since most fans are only a couple watts. Even if you compare the temperatures on the same cooler, you can see a 16watt drop with only a 16c decrease in temperatures.

ohhh ok interesting