Author

Topic: Cointerra power stepping vs wattage (vs hashrate) (Read 2386 times)

member
Activity: 117
Merit: 10
I did a short experiment on one side of one unit. Apparently it was too short an experiment, since it can take a long time for things to stabilize. However, here are some numbers that I had enough time to gather.

Except for the 9 setting which was running for days, each power stepping was left to stabilize for approximately 5-10 minutes. Even then both power and hashrate would go from idle (~230W/0GH/s) to "full" load quite frequently.

Power settingHash Rate (GH/s)Temperature (°C)Power (W)
980861.91020
876053.4885
753564.0789
655056.0712

Yes, these numbers don't make sense - they're not linear, or even decreasing (the temperature stumped me). A better test would involve letting the units run 24 hours at each setting. I'm not really willing to do that, so better numbers from me will have to wait until it isn't as costly to test.
legendary
Activity: 1330
Merit: 1026
Mining since 2010 & Hosting since 2012
Maxwell,

Next time I am at our DC, ill make a video with the unit hooked up to power meters and change the settings and collect some data.   Cheers.


-D
hero member
Activity: 702
Merit: 500
I doubt that anyone has any reliable data of this kind.

The key component of a miner is an un-binned CMOS chip where expected manufacturing tolerances are probably +/-25% or even +/-50%. I don't think that any of the vendor does any bin-sorting of their chips, anything that hashes and doesn't smoke is getting shipped to the customers.

Instead of investing into more true power meters I would suggest splicing sub-1-ohm resistors into the neutral wire of the extension cords. Just measure the voltage drop on them with any half-decent AC multimeter.

If you have reasonably US-electrical-code-conformant wiring this would be quite safe from accidents, especially after all I've seen here about GPU mines. But please don't follow my advice only if you are reasonably sure that children can't walk into your mine.

The child-safe option is to get a clamp-on magnetic AC current probe for your multimeter and just cut the outer layer of the power cord isolation to be able to clamp on a single wire out of three.

Im pretty sure the in the current model.. the cointerra asics are binned, and that the systems are burned in, at 1.6 TH...  and some get to 1.7 TH.

as for the power stepping, you probably want to try it on maximum first (setting 9), and then step down from there if there are problems.  The firmware doesnt let you overdrive the hardware, so its safe to try and i think max is the default setting from the factory.  the power stepping is a reduction from maximum performance in case of some issue like lack of local power supply or ambient temp too warm, (or some hardware problem).

if youre going to run it on a 240 supply you should have plenty of power, as the problems of lack of power come with under voltage not over voltage.
legendary
Activity: 2128
Merit: 1073
I doubt that anyone has any reliable data of this kind.

The key component of a miner is an un-binned CMOS chip where expected manufacturing tolerances are probably +/-25% or even +/-50%. I don't think that any of the vendor does any bin-sorting of their chips, anything that hashes and doesn't smoke is getting shipped to the customers.

Instead of investing into more true power meters I would suggest splicing sub-1-ohm resistors into the neutral wire of the extension cords. Just measure the voltage drop on them with any half-decent AC multimeter.

If you have reasonably US-electrical-code-conformant wiring this would be quite safe from accidents, especially after all I've seen here about GPU mines. But please don't follow my advice only if you are reasonably sure that children can't walk into your mine.

The child-safe option is to get a clamp-on magnetic AC current probe for your multimeter and just cut the outer layer of the power cord isolation to be able to clamp on a single wire out of three.

staff
Activity: 4284
Merit: 8808
Has anyone charted out the cointerra power stepping vs wattage (vs hashrate too, perhaps)?

I am precariously close to over-provisioning some breakers and short on power meters. It would be handy if someone had charted this stuff out so I can set it in a way thats more reliable than trial and error. Smiley
Jump to: