[GUIDE] Undervolt antminer s1 [1.19W/GH at the wall]

[grn]

I just did my very first under volt under clock.

I have a good multi meter all resistors read 4.46 or 4.47 to start.

adding graphite is not exact and I had variance from 2.81 to 3.42    2 machines so 16 resistors.

one machine does not like running at freq  300 it had 20 percent   error rate.  so I dropped it to freq 275

 these two machines with just under clock to 300 each pulled 616 watts. and hashed at 305gh

once I did both under volt and under clock

I am now hashing at 295gh with 399 watts

at over clock of 393 I had  398gh at 818 watts.

To run at 300 with little error you have to make sure you get 3.4 - 3.5kohm, that will give you 153 - 155ghs per ant at about 210 watts

I currently have 2.15 Ths at under 3kw at the wall, only 4 power supplies 3 x 850, 1 x 1300

[Beaflag VonRathburg]

If you're going to do this and read the resistance values please post the voltage readings you have for the units under load as well. It is rather straight forward and makes a lot more sense as even at the same resistance values the voltage can vary. Everyone is used to seeing the correlation between XYZ frequency and 1.abc voltage.

[philipma1957]

thanks I just carried the 2 of them to my buddies office and we have 300gh   at free power.   but pulling 400 watts.    


we can do up  to 800 free watts as per his leasing contract.  

  so by next week 4 s-1's  running at 600gh 800 watts and the watts are free!

to beaflag will do so on the next 2.

If you're going to do this and read the resistance values please post the voltage readings you have for the units under load as well. It is rather straight forward and makes a lot more sense as even at the same resistance values the voltage can vary. Everyone is used to seeing the correlation between XYZ frequency and 1.abc voltage.

[grn]

If you're going to do this and read the resistance values please post the voltage readings you have for the units under load as well. It is rather straight forward and makes a lot more sense as even at the same resistance values the voltage can vary. Everyone is used to seeing the correlation between XYZ frequency and 1.abc voltage.

voltage under load = .875 - .91

[Bitcoin++]

Here's a closeup of my Antminer.

http://imgur.com/SglhzeV

As I've understood the R66 transistor is where I should add pencil graphite (and there are 16 of these albeit with different numbers).

It has kind of two blank edges and the middle is matte dark gray. Does it matter where I apply the graphite?

What happens if I apply too much? Will only the HW error increase, or can the hardware get permanent damage?

What happens if I don't apply enough?

It should be applied when the miner is off, right? Should I do all 16 at the same time?

Is it necessary to use a voltmeter? Can the measure be made when the antminer is off? If I don't use the meter, is there another way to determine if the adequate amount of graphite has been applied?

[Raptor2213]

Here's a closeup of my Antminer.



1 - As I've understood the R66 transistor is where I should add pencil graphite (and there are 16 of these albeit with different numbers).
2 - It has kind of two blank edges and the middle is matte dark gray. Does it matter where I apply the graphite?
3 - What happens if I apply too much? Will only the HW error increase, or can the hardware get permanent damage?
4 - What happens if I don't apply enough?
5 - It should be applied when the miner is off, right? Should I do all 16 at the same time?
6 - Is it necessary to use a voltmeter? Can the measure be made when the antminer is off? If I don't use the meter, is there another way to determine if the adequate amount of graphite has been applied?

1 - Yes
2 - Not really - so long as you join the two ends.
3 - You erase and try again.
4 - You put more on.
5 - Yes
6 - Absolutely.  It is impossible to apply this mod without a voltmeter.  You can measure the ohms when the miner is off to determine if you put on enough or too much.

[klondike_bar]

Here's a closeup of my Antminer.



1 - As I've understood the R66 transistor is where I should add pencil graphite (and there are 16 of these albeit with different numbers).
2 - It has kind of two blank edges and the middle is matte dark gray. Does it matter where I apply the graphite?
3 - What happens if I apply too much? Will only the HW error increase, or can the hardware get permanent damage?
4 - What happens if I don't apply enough?
5 - It should be applied when the miner is off, right? Should I do all 16 at the same time?
6 - Is it necessary to use a voltmeter? Can the measure be made when the antminer is off? If I don't use the meter, is there another way to determine if the adequate amount of graphite has been applied?

1 - Yes
2 - Not really - so long as you join the two ends.
3 - You erase and try again.
4 - You put more on.
5 - Yes
6 - Absolutely.  It is impossible to apply this mod without a voltmeter.  You can measure the ohms when the miner is off to determine if you put on enough or too much.

to clarify number 6: a multimeter is used to check the resistance (3.4-3.5 Kohm). Measure after each pencil mod.

checking the voltage isnt *really* needed for each chip secton - if all the resistances are the same checking the voltage in one section should mean all the other sections are the same voltage.

[basil3legs]

Just gave this a go.

I reduced the resistances to between 2920 and 2950 and the chip voltage range is now between 855mV and 870mV, so pretty consistent although I only measured the top four on each board.

Having set the frequency to 275MHz I am getting around 0.25% errors (although this appears to be slowly creeping up) and my meter is showing 180W (240V).  With the sudden increase in the price of Bitcoin I calculate I am losing around 5% of my income after allowing for the cost of the electric which is pretty good and that will decrease should the price go down again (or when the difficulty goes up next).  The plus side is a reduction of around 10C in chip temperature and it is a lot quieter.  Now to decide which other ones to do as that S2 would only really run at 375MHz anyway, although at that speed there were little to no errors, any quicker and they really racked up!

I have another one that only likes 375MHz and gives a few more errors so might as well do that one as well!

Thanks for the research on this guys.

[philipma1957]

here is one photo  4 resistors on each side 8 per miner.

[Hqen2000]

I just set my 3 miners all to 3.4-3.5 ohms.. 300mhz, the load went from 1028watts, down to 690watts. Same exact hash rate. I was running 300mhz prior due to heat issues. So now I am saving 338watts with zero penalty!!! And they are sooooo quiet now!

The difference in heat output is huge. I'm also buying molex adapters to power all 3 miners from a single 850watt power supply. I'll probably get down to about 650watts to power 3 miners all hashing at 155.

All watts are measured at the wall.

Looking at cost:

$2250
S2 1100watts / 1000g/h

~$290 (depending on power supply, etc.)
S1 230watts / 155g/h

7x S1 1085g/h
1610watts
$2,030

You could probably get that 2,030 down to 1800 or so if you were very cost effective on the power supplies and buying the miners. But the S2's value is pretty good. Especially since you are talking about about 500 additional watts from the S1.

I think the S2 is slowly starting to become a better value.

[crazyearner]

Has anyone done this to get  200GH per ant 100GH per blade etc with reduced voltage. So less voltage more GH am thinking to do but just checking here first to see if anyone's managed to reduce the voltage and have a good hash speed

[klondike_bar]

Has anyone done this to get  200GH per ant 100GH per blade etc with reduced voltage. So less voltage more GH am thinking to do but just checking here first to see if anyone's managed to reduce the voltage and have a good hash speed

Imo the 150GH/210W target is pretty ideal. Good payoff between time spent modding and the power savings. If you did 180GH/300W you would find yourself re-modding a few weeks later.

(that said, 3.9-4.0 kohm might give 180GH / 300W if you are lucky)

[DimensionsOfHell]

Is there any way to have the fans always at 100%? I want to reduce the heat as much as possible. Summer here in FL gets pretty hot, and the temp in my miner room is 98F right now, and its only going to get hotter.

I've got 16 AntMiner S1's, 2 Dragon 1T miners, and 1 KNC Saturn with 3 boards.

If I can get the Ants to drop some massive temp, I could add 1 or 2 S2's. I'm out of power & have excessive heat.

[philipma1957]

Is there any way to have the fans always at 100%? I want to reduce the heat as much as possible. Summer here in FL gets pretty hot, and the temp in my miner room is 98F right now, and its only going to get hotter.

I've got 16 AntMiner S1's, 2 Dragon 1T miners, and 1 KNC Saturn with 3 boards.

If I can get the Ants to drop some massive temp, I could add 1 or 2 S2's. I'm out of power & have excessive heat.

you need to add really good pull fans. I posted some shots of them and where I got them on ebay.

add these and vent the heat out of the room


http://www.ebay.com/itm/151276722479?ssPageName=STRK:MEWNX:IT&_trksid=p3984.m1439.l2649

   he sold me 14 at 107  I will find the hook up photos


I used 4 pipe cleaners to attached the fan on the pull side.




see the fan pipe cleaners and nylon nuts for spacing

[Beaflag VonRathburg]

Is there any way to have the fans always at 100%? I want to reduce the heat as much as possible. Summer here in FL gets pretty hot, and the temp in my miner room is 98F right now, and its only going to get hotter.

I've got 16 AntMiner S1's, 2 Dragon 1T miners, and 1 KNC Saturn with 3 boards.

If I can get the Ants to drop some massive temp, I could add 1 or 2 S2's. I'm out of power & have excessive heat.

I'm at 106* right now with both sides open and 3x 500cfm duct fans running. I understand completely.

[Beaflag VonRathburg]

I was noticing that the hashrate for my antminers had been dropping off. Logged into them and noticed they had some pretty high error rates:

18 Hours 30 Minutes
Ant1 275 = 12.763361189114817696879475815092
Ant2 250 = 0.10015846041845231355797176812215
Ant3 325 = 4.4344988816100151774557453763173
Ant4 275 = 8.9107904850450710092977864238739

Pulled them onto my test bench to remeasure resistance values to see if it had changed any. Sure enough all of the miners, but one had massive resistance drops. Used a pencil to readjust them back to nominal and things are running properly again.

Ant1

Code:

2.675 kOhms .802v top / .802v bottom
2.480 kOhms .800v top / .800v bottom
2.270 kOhms
2.674 kOhms
-
R3  Resistance 2.838 kOhms 0.851v top / 0.847v bottom
R66 Resistance 2.842 kOhms 0.858v top / 0.853v bottom
R38 Resistance 2.944 kOhms 0.861v top / 0.856v bottom = *added*
R52 Resistance 2.876 kOhms 0.859v top / 0.854v bottom

2.705 kOhms
2.566 kOhms
2.633 kOhms
2.730 kOhms
-
R3  Resistance 2.860 kOhms 0.854v top / 0.850v bottom
R66 Resistance 3.109 kOhms 0.881v top / 0.876v bottom = *added*
R38 Resistance 2.964 kOhms 0.867v top / 0.863v bottom = *added*
R52 Resistance 2.850 kOhms 0.853v top / 0.849v bottom

Frequency = 275
Power Consumption = 176 watts
Roughly =  139.73 GH
30 Minute Error Rate = 0.55135273753045262213104244133863

Ant2

Code:

No issues

Ant3

Code:

3.640
3.578
3.620
3.259
R3  Resistance 3.870 1.010v top / 1.004v bottom
R66 Resistance 4.000 1.035v top / 1.028v bottom
R38 Resistance 3.900 1.014v top / 1.008v bottom
R52 Resistance 3.930 1.010v top / 1.006v bottom = Added
-
3.478
3.475
3.750
3.560
R3  Resistance 3.960 1.008v top / 1.001v bottom = Added
R66 Resistance 3.920 1.008v top / 1.001v bottom = Added
R38 Resistance 3.920 1.014v top / 1.007v bottom
R52 Resistance 3.999 1.012v top / 1.007v bottom = Added

Frequency = 350
Power Consumption = 280 watts
Roughly =  167.65 GH
30 Minute Error Rate = 0.2453540629175270253856431434471

Ant4

Code:

2.670
2.599
2.533
2.481
R3  Resistance 2.916 kOhms .860v top / .856v bottom =  Added
R66 Resistance 2.875 kOhms .860v top / .856v bottom =  Added
R38 Resistance 2.855 kOhms .851v top / .847v bottom =  Added
R52 Resistance 2.959 kOhms .864v top / .860v bottom =  Added
-
2.659
2.912
2.591
2.674
R3  Resistance 2.871 kOhms .854v top / .848v bottom =  Added
R66 Resistance 2.912 kOhms .865v top / .860v bottom
R38 Resistance 2.875 kOhms .861v top / .855v bottom
R52 Resistance 2.989 kOhms .870v top / .865v bottom


Frequency = 275
Power Consumption = 174 watts
Roughly = 139.49 GH
1 Hour Error Rate = 0.8297014898156421524827221502945

[klondike_bar]

I was noticing that the hashrate for my antminers had been dropping off. Logged into them and noticed they had some pretty high error rates:

18 Hours 30 Minutes
Ant1 275 = 12.763361189114817696879475815092
Ant2 250 = 0.10015846041845231355797176812215
Ant3 325 = 4.4344988816100151774557453763173
Ant4 275 = 8.9107904850450710092977864238739

Pulled them onto my test bench to remeasure resistance values to see if it had changed any. Sure enough all of the miners, but one had massive resistance drops. Used a pencil to readjust them back to nominal and things are running properly again.

Ant1

Code:

2.675 kOhms .802v top / .802v bottom
2.480 kOhms .800v top / .800v bottom
2.270 kOhms
2.674 kOhms
-
R3  Resistance 2.838 kOhms 0.851v top / 0.847v bottom
R66 Resistance 2.842 kOhms 0.858v top / 0.853v bottom
R38 Resistance 2.944 kOhms 0.861v top / 0.856v bottom = *added*
R52 Resistance 2.876 kOhms 0.859v top / 0.854v bottom

2.705 kOhms
2.566 kOhms
2.633 kOhms
2.730 kOhms
-
R3  Resistance 2.860 kOhms 0.854v top / 0.850v bottom
R66 Resistance 3.109 kOhms 0.881v top / 0.876v bottom = *added*
R38 Resistance 2.964 kOhms 0.867v top / 0.863v bottom = *added*
R52 Resistance 2.850 kOhms 0.853v top / 0.849v bottom

Frequency = 275
Power Consumption = 176 watts
Roughly =  139.73 GH
30 Minute Error Rate = 0.55135273753045262213104244133863

Ant2

Code:

No issues

Ant3

Code:

3.640
3.578
3.620
3.259
R3  Resistance 3.870 1.010v top / 1.004v bottom
R66 Resistance 4.000 1.035v top / 1.028v bottom
R38 Resistance 3.900 1.014v top / 1.008v bottom
R52 Resistance 3.930 1.010v top / 1.006v bottom = Added
-
3.478
3.475
3.750
3.560
R3  Resistance 3.960 1.008v top / 1.001v bottom = Added
R66 Resistance 3.920 1.008v top / 1.001v bottom = Added
R38 Resistance 3.920 1.014v top / 1.007v bottom
R52 Resistance 3.999 1.012v top / 1.007v bottom = Added

Frequency = 350
Power Consumption = 280 watts
Roughly =  167.65 GH
30 Minute Error Rate = 0.2453540629175270253856431434471

Ant4

Code:

2.670
2.599
2.533
2.481
R3  Resistance 2.916 kOhms .860v top / .856v bottom =  Added
R66 Resistance 2.875 kOhms .860v top / .856v bottom =  Added
R38 Resistance 2.855 kOhms .851v top / .847v bottom =  Added
R52 Resistance 2.959 kOhms .864v top / .860v bottom =  Added
-
2.659
2.912
2.591
2.674
R3  Resistance 2.871 kOhms .854v top / .848v bottom =  Added
R66 Resistance 2.912 kOhms .865v top / .860v bottom
R38 Resistance 2.875 kOhms .861v top / .855v bottom
R52 Resistance 2.989 kOhms .870v top / .865v bottom


Frequency = 275
Power Consumption = 174 watts
Roughly = 139.49 GH
1 Hour Error Rate = 0.8297014898156421524827221502945

pencil mods have a tendency to 'bake in'.  Within the first 24 hours the resistance may drop slightly and thus the voltage may slip below what worked at first

[DimensionsOfHell]

......

I actually have all my ants on a push-pull configuration. I have some Rosewell fans on them (Was on sale for $5 with free shipping)

http://www.amazon.com/gp/product/B003WM184A/ref=oh_details_o01_s00_i00?ie=UTF8&psc=1

which isn't as high as the ones you've linked to, but its not that bad. It's about 75CFM. I just took out two of the screws from the original fan (one from bottom left side of the fan, and one from top right side of the fan), and used it on the new front fans (top left/bottom right I think. You can't use the long screws on one side due to a capacitor blocking the way). Two screws is more then enough to hold them in place, and that way you don't have to use the pipe cleaners, and no extra money spent.

[philipma1957]

......

I actually have all my ants on a push-pull configuration. I have some Rosewell fans on them (Was on sale for $5 with free shipping)

http://www.amazon.com/gp/product/B003WM184A/ref=oh_details_o01_s00_i00?ie=UTF8&psc=1

which isn't as high as the ones you've linked to, but its not that bad. It's about 75CFM. I just took out two of the screws from the original fan (one from bottom left side of the fan, and one from top right side of the fan), and used it on the new front fans (top left/bottom right I think. You can't use the long screws on one side due to a capacitor blocking the way). Two screws is more then enough to hold them in place, and that way you don't have to use the pipe cleaners, and no extra money spent.

that does not work for the fans i linked they are 38mm just like the stock push fan.  they move 160 cfm.  if you put them in with the screws you mention the blades do not clear the assembly nuts on the heat sinks.  my 4 pipe cleaners allowed for the spacers and let the fast fans blades work.

your idea does work if the fan blades clear the heat sink nuts.

the pipe cleaners cost be 99 cents for 50.  i had the nylon spacers.

   if you are in the usa i can mail you one to see if it is better then your fans.  it fits in a padded flat rate envelope. cost 6 bucks to ship. i have extra. so i would send it for the shipping fee.

[BITMAIN]

This guideline is for under clocking modification to get 1W/GH/s. In order to get 1W/GH/s, you shall modify the hardware and software.

1 Hardware Modification

1.1 Resistor Typical Value
Package Type: 0603


1.2 Power Consumption and Hash Rate


1.3 How To Modify
Step 1: Remove the existing components: R3, R66, R38, R52.
Step 2: Refer to Resistor Typical Value to replace the R3, R66, R38, R52.
Step 3: Clear the excess of rosin and solder tin slag around resistors.

1.4 Refer to below locations to do modification


2 Software Modification

2.1 Connection Miner
Though SSH protocol on your PC to connect S1, user name and password are ‘root’ by default. Then modify ‘freq_value’, ‘chip_freq’ and ‘timeout’ in the etc/config/asic-freq file.

2.2 Modification Value
2.2.1 200M*8*32*2 = 102.4GHash:
        

Code:

option 'freq_value'     '0781'  #200M
        option 'chip_freq'      '200'
        option 'timeout'        '60'

2.2.2 225M*8*32*2 = 115.2GHash:
        

Code:

option 'freq_value'     '0881'  #225M
        option 'chip_freq'      '225'
        option 'timeout'        '55'

2.3 Restart Miner
Power off and restart Miner, check the configuration via Web GUI ‘Miner Status’. If the Miner doesn’t work normally, please double confirm the correction of frequency configuration file.