Topic: undervolt antminer s3 - page 3. (Read 26411 times)

okay so just messing with adjusting the voltage of the ASIC chips my results are all over the place and I think I know why.at this point I'm assuming on the architecture of how the device functions but if I was building a cheap control system for voltage I would use some sort of processor to send out a control signal such as pulse width modulation or a 1 to 5 voltage as a referenceto some sort of voltage regulator. This appears to be the case if looking at the ASIC blades. There are several banks of voltage regulators leading down the card to several groups of chips looking at thisI have to assume that it's an open loop voltage control system. Meaning there is not feedback from each individual voltage regulator to allow the control system to know whether it's too high voltage or too low voltage and then adjust appropriately. Therefore I have to assume setting a voltage in the firmware is not precise control of the voltage to each individual ASIC chip. Also if you look at the design the board is laid out from left to right or right to left depending on how you look at it. This means the intake air comes in on one side of the board and then exits on the opposite side of the board therefore the temperatures across all these voltage regulators will vary greatly.  temperature differences will make voltage regulators drift in their output voltage. After observing this it would be my recommendation 2 leave the voltage at stock or to set it at the midway point. Setting it at the midway point would put the chips in the center at the midway point the chips at the inlet. At a lower voltage and the chips at the highest temperature location next to the exit at the highest point. By averaging all these together you would get a average efficiency across all the chips.

for the rest of my testing I set the voltage to. 7 volts. I was able to get a higher wattage efficiency out of the units by adjusting the clock more so than adjusting the voltage I attached a copy of some data I produced during these tests above.

by keeping units pretty well stock it appears that he can only fit about 7 units on a 3 kilowatt circuit.you can get the chip efficiency up by decreasing the clock but by decreasing the clock your overall hashrate decreases but efficiency increases. At the most efficient. With a clock of 100 he can place 18 minors on a 3000 watt circuit this would give him a higher overall hashrate than just using 7 miners at stock but would only lead to about 200 more Giga hashes. His return on investment would be ridiculous. Even adjusting the clock to 200 he could add 9 units to the circuit and would gain approximately a hundred and fifty more Giga hashes but with the price of ant miners he would be adding 600 to 700 more dollars to his setup which is just not a good return on investment.

I think the best way to optimize these units would be to set them up default and then increase the clock until you start seeing errors then decrease the clock so you don't get errors. Hardware errors in this scenario will lead 2 more wasted watts due to the fact units having to redo calculations.

I did observe that with higher temperatures the hash rate begins to decrease I was able to prove this by blocking the vents and watching my hash rate dropped by about a hundred Giga hashes. Therefore I have to assume that a more effective way of increasing efficiency would be to keep these units as cool as possible. Sometime in the future I will start experimenting with supercooling these units by using A heat exchanger inside of a 4 inch pipe feeding into the units. I will also try experimenting with other ways to cut wattage off of the units. I'm starting to think there are a lot of wasted wats inside of the circulation fans. Example would be that your power supply is not 100% efficient and DC motors are not the most efficient either therefore I have to assume y would you use clean power to run DC fans. I will attempt to remove the fans and replace them with AC units in an attempt to he could increase efficiency.so far I've been able to get objects up to a hundred degrees by just using the exhaust air on these units. Which leads me to think we may be able to reclaim wattage with the use of a device such as a thermal electric generator.

I will continue to try to work on my units to make them the most efficient as possible in my spare time its in my best interest to do so because I plan on being independently rich using Bitcoin miners.until then donations are greatly appreciated to help further my research which I will continue to share with everyone on the post form. When I get something new I will post it in this post forum as well as maybe make a new post.

16readkuKHgLPo7ii24VxVstJtM3VnRPrn

clock   volts   hash 1   hash2   hash total   watts   wats per hash      units on 3kw      test    test hash
250   0.6   493.19   466.44   959.63   839   0.874295301      7.151370679      7   3358.705
250   0.7   497.31   486.55   983.86   847   0.860894843      7.083825266      7   3443.51
250   0.8   499.46   491.23   990.69   836   0.843856302      7.177033493      7   3467.415
                                 
200   0.7   397.99   399.31   797.3   645   0.808980309      9.302325581      9   3587.85
100   0.7   213.2   190.19   403.39   325   0.80567193      18.46153846      18   3630.51

good point on the 240. if so lucky them. thing run better at the higher voltage. 

i have not found a refferance to the defult voltage.

in the  /etc/config/cgminer the last line is blank but after putting in a voltage on the GUI it is added as the last line.   no line default voltage. what that is ?   not sure currently trying to reverse engineer this thing but work is getting in the way

https://bitmaintech.com/files/download/BM1382_Datasheet_v3.0.pdf
page 3 shows the chip speck

Sounds like they are from Europe (Portugal?) where the voltage is 240V, so that would be only 12 Amps.

By the way, where do I find what the stock voltage is?

3000w/120volts=25amps
keep in mind that as you volts drop you watts go down and amps go up

curently at my house i get 117 volts. a good idea would be to put a volt meter on your plug to see what you volts are at different times of the day and different days.

117v * 25 amps = 2925 watts
115v * 25 amps = 2875 watts

do you have a 25 amp beaker. that seams odd. residential is usealy 15 or 20???

ill put a report together on this. i have a water meter at the wall 1 PSU and 2 s3+

i have a DC what meter on order right now. so every thing i post right now just a estument.

right now i see at 10 watt swing with ambent temp. so he need to keep them cool for lower power use

miner #2 was not stable at 250 & .80 had to retard it to 243 to be stable but i and only testing volts right now and then will test clock

250 & .08 / 499.46 491.01 ghs / 834 watts / 7.19 amps

ill minn in large chunks to get good averages and use 2 minners to rule out defects in asic chips.

that's alot to lug around    Please tell us how it goes.  Do you have a way to measure how much power it draws?

ok so i went all the way up to 1.00 for the volts and 250 on the clock. it spiked up to 611.72 GH/s but is up and down. ran 16 min with 1 HW error. fans are 2,400 and 2,100. temp 41 and 37    all on room air.   

 i live in hotel rooms so i travel with 2 antminer s3+ free electric. ill get these two in the A/C server room this weekend and see how far i can push them.

SSH in to the unit and run this cmd
vi  /etc/config/cgminer
this is the location that the new FM uses to store pools, freq, time out, volt, ect.

vi /etc/asic-freq is no longer used
\

you have to set volts in the GUI first before it shows in the config file.

So im not sure how the voltage should be applied 0.55 or .55 or .55v - is there a way to check that is does something? .55v 218mhz and iget 0.01% hw. Those have burned alot electricity for nothing and keeped some noice too...

Nope, there isnt.

Are you doing this with the s3 or s3+?  Do you know if there is a difference?

With the new antMiner_S320141013 update on Bitmain's website you can easily modify the voltage under "miner configuration" then "advanced settings". I have played around with .63v but I get more hw than I would like. A voltage of .72v at 243.7mhz gives me a solid 490gh/s while using about the same wattage as stock.

I am very interested in this topic.

well I have 6 of them so I would be willing to tear one down.

Yes they look very similar, the man difference is that you have to take the S3 pretty much completely apart.
You have to get access to the back side of the blades since the resistors ar on the back not in the front like on the S1. So just popping of the case won't be enough.

sent pm.

i did the s-1's i would do the s-3's

Hi Phillip,
Correct, I run 6x S3 at 150Mhz undervolted at exactly 1090w on the wall (GHash.io reports a 1.85 TH/s dally total ). Not sure if 200 watts headroom is save ( I think your suppose to stay 25% under the maximum rating at continues usage ).

I tried undervolting 193MHz-200Mhz but it only yielded 380Gh/s at 280w was not worth it in my opinion. ( the gain was only 20w ) and noise and heat was not much lower.

If you shot me your email in a pm I canemail you some photos and simple instructions. If you done a S1 pencil mod you should be fine, but if not it might be a little more you bargaind for.

so 180/310 =  .58 watts   at freq clock 150

if I do advanced tab now at freq clock 200 and 400 gh  just a down clock I get   333watts  so that is 333/400 =  .83 watts

  frankly  I would be tempted to do this. as I could run 6 s-3 on 1 evga 1300  

 6 x 180 = 1080 watts  and get 1860 gh

could you post photos of the resistor to be pencil modded.

@ neegeeboo

setting for 150 are

  #option 'freq_value'    '0b83'  #150M
  #option 'chip_freq'     '150'
  #option 'timeout'       '27'                feel free to try 26 or 28

I used the new firmware with the "Advanced" tab to adjust the clock frequency down to 150mhz, with that the consumption goes down to 240-250w on the wall ( at 310 Gh/s ).
The resistors are located on the back, or the inside where the big sink is attached. You will need to completely remove the board to get proper access to the resistors.
The default value is 2.083 Kohm. And I use the pencil mod to bring the value down to 0.900 to 1.050 Kohm. I have most of mine set close to 0.950 Kohm.
After that the consumption will be down to 180w on the wall ( still at 310 Gh/s ). Noise from the fans and temp will be way down after this mod. And you can use a single PCIE cable to power one S3 ( two hashing boards )