What kind of mining set up are you runnning?

Ah, they're the same as my 1750 timings.

Makes sense. Why reinvent the wheel, right?

I found these super tight timings on a Russian forum once. They were ok but I had to drop the speed and cbfd trying to figure out the sweet spot..
777000000000000022AA1C00EF615C41B0551016BA0D96060060060004061420EA8940AA030004C01914292E692E3B16

 

Fair enough, I only just came across them a week ago now. Wish I had them a year ago! Worth a few grand over that time to me for that duration.
What sort of clocks/power draw are you seeing?

 

See for yourself if you want. I'm pretty conservative. Plain old ~27.5mhs eth ~80w each

see attached zip

 

80w would be a software reading of the core usage. You have memory, aux controllers, fans and lights to account for too. 80w core would be like 130-150w at the wall.

 

I setup a 6x rx580 rig back in Aug. Learnt a lot setting it up with regards to hardware and software configurations. I fine tuned these cards as most efficient I could. Total system power while mining is 900w +- 50w fluctuations, measured at wall. When I started I was making nearly 2 ETH a month. Now around 0.4 ETH a month. I have mined around 9-10 ETH so far. Any questions welcome.

I was having occasional pool connection issues recently. Went to nanopool the other week to get the EU server to setup a failover and seen they have an Aussie server now. Never misses a beat now.

 

Not too sure, thats pretty high power draw. His ROM shows 700mv on the core, and 950mv on the memory.
My GPUs pull 100-110W at the wall (actual) @ 900mv core/mem, and dropping to lower core at my end (850mv) pushes them under 100W.
If they work @ 700mv, then power draw will be quite low.

Mem voltage +/- 50mv didn't show much of a difference at the wall across 6 cards for myself (850-900mv)

I'll drop my clocks/voltage on a rig tonight and test it out. If they can be stable @ 700mv, and the clocks shown in bios (1150 core) then the hash drop from 1250 would only be about 50-80h/s on CNv7, but the power drop should be noticable on hash per watt ratio...

 

What's your results?

 

Not great at the settings you've got set in your bios.
Can you confirm via watt meter what you're pulling from the wall, or via software?

My better cards are lucky to do 1150@840mv, let alone 700...Memory works fine at 850, even lower for most cards.
Did you have to do any sort of checksum fix to the rom when you set your voltages in bios?

 

Really? Hmm. Mine are XFX.

745w @ wall right now with 3x570 + 2x1070 @ 50% -200 ("100w") on old Q6700

checksum fix? That atikmdag fix thing for the drivers? Yeah, of course.

 

Nah not the driver fix, I was reading a while back that just entering the voltages direct in bios with intended values didn't always work. The defaults show like 65000 odd, but its not that high in the system. It might be a hex value or something for a specific voltage.

Does GPU software confirm core @ 700mv for you?

Even if so, without knowing how much the q6700 system pulls on its own (I assumed 80w (double what a newish Celeron system with 1 stick of ddr4 + ssd pull according to my measurements), it leaves you with an average of 130w or so per GPU (without knowing what sort of drawn the 1070s are pulling as I have no experience with them) after power supply (in)efficiencies.

I'm seeing 700W peaks at the wall (as low as 580) with 6x 570s.

 

I haven't looked tbh. About to head overseas for a week, so might not be able to answer for a while. I was only counting it during ethhash btw. If you're doing a more core smashing algorithm, well, you'd expect higher draw.

The 1070s report 100w ea. So... I guess (745-10%(PSU) -200-~130) /3 = 113w ea?

 

Its good whatever it is. Simply looking for the draw at the wall, and how many cards are in use.

 

erm, I meant 1070s for the first part

 

illdrift have you confirmed the extra power draw with a power meter ?

 

Yeah, i only go by the wall meter reading as the software numbers aren't useful for anything other than a reference on any individual card.

An older rx570 gives a 'core' reading yet the same model newer rx570 seems to give a 'core+mem' software reading and is atleast 20w+ higher (with a similar actual wattage reading from the wall)

I've moved 3 of 4 rigs over to HiveOS, and so far no downtime

 

How much was the electricity bill?

 

It uses around 900w at the wall. I had it only on a 1000w PSU for first 3 months or so but I was worried on how close to the 1000w it was spiking and didn't want to burn anything out so I added a 2nd 500w psu later. I have it setup in a commercial place where I get a 16amp power line for free. I've used it to take up some of the excess. The normal commercial rates I get are around 12c kwh off peak to around 22c peak. Unless you have a cheap power deal or have a similar situation to me it probably isn't worth the risk to run a miner in aus. We are top 3 countries in the world for power cost. Specially now Bitmain is tapping into the top 3 GPU PoW coins now.

 

So i've set a scheduled task to reboot my machine to start the mining software @5am and a scheduled task to kill it at 6:30pm. This means mining is now essentially free as my solar will cover it. Due to me changing the plan I am on I now get 20c per kWh I put back in the grid and no further discounts. As I have a large array I need to minimise my power usage at night to ensure I keep my imported kWh low

 

Wifi connected smart meters are another option, though would cost you $50 or so. You can set them to turn on and off at various schedules over the week. You can also turn them on/off anywhere with your phone, which can be useful

 

How would a smart meter shutdown a pc gracefully?