Author

Topic: ANTMINER S7 is available at bitmaintech.com with 4.86TH/s, 0.25J/GH - page 133. (Read 528055 times)

legendary
Activity: 1694
Merit: 1002
Go Big or Go Home.....
My S7 are getting lots of HW errors and are underperforming
I have 4 of them ,only 1 is getting low HW errors and good performance

The rest of them perform like this one : https://gyazo.com/af319f55fa78b0ed11626611e76888b0
with HW errors around 0.03 % or higher

How can I fix this ?

Why do you think that 0.03% is too much? If you remember, previous generations of Bitmain ASIC miners (S2 for example) had way more HW errors, my S2 ran at 0.9%. But at the pool there was about 1+ Th. So did I lose any valuable amount of hashing power? I guess no.

Also I found that my S7 (Batch6) gives many HW errors right after start, it can be up to 3% in first seconds (I saw it once), but take a look at the HWE percentage after 20-30 minutes...

Because at original freq or under it is 0.0000% HW error rate. Once you go into higher speeds (although the later batches get it even at original speeds from what I see)you start hitting HW error rates. If there is more than 0.01%, IMO your miner is more irrational than stable and could impact the miner hardware or the income generated from it towards the pool.

On new gen hardware there should be pretty much no HW errors and comparing it to an old miner tech that is 3 generations or more previous is like saying that your 1990 Nissan Sentra sounded like a lawnmower and it is ok for a brand new 2015 model to have the same shitty sound.  Cheesy
legendary
Activity: 2030
Merit: 1076
A humble Siberian miner
My S7 are getting lots of HW errors and are underperforming
I have 4 of them ,only 1 is getting low HW errors and good performance

The rest of them perform like this one : https://gyazo.com/af319f55fa78b0ed11626611e76888b0
with HW errors around 0.03 % or higher

How can I fix this ?

Why do you think that 0.03% is too much? If you remember, previous generations of Bitmain ASIC miners (S2 for example) had way more HW errors, my S2 ran at 0.9%. But at the pool there was about 1+ Th. So did I lose any valuable amount of hashing power? I guess no.

Also I found that my S7 (Batch6) gives many HW errors right after start, it can be up to 3% in first seconds (I saw it once), but take a look at the HWE percentage after 20-30 minutes...
legendary
Activity: 2128
Merit: 1005
ASIC Wannabe
^I still suggest you try modifying the setting in cgminer.conf.factory and then hard-reset the device with the physical push-button. If that command can adjust the voltage, a reset might "force" the change and would possibly make a lot of sense depending how the chips are programmed initially for factory specs
hero member
Activity: 588
Merit: 500
It's fascinating watching you guys discus all of this. I wish I could contribute, but it's a good country mile above my head. Here's hoping a way to adjust voltage is found, since B8 definitely doesn't respond to increased 12V from the PSU as earlier batches do.

I am certain that there will be a way to adjust the voltage.  Smiley What is definitely known is that we have a Buck Converter supplying the voltage to the ASIC chain, that Converter has a Digital Pot which adjusts the value of the resistor from the FB input to ground (If you like a variable Digital Pencil Mod  Smiley controlled by a Micocontroller.

What is not known, at the moment, is exactly how this addition is being used by Bitmain. At one extreme it is just a Factory set fixed voltage used to make the best of the different batches of BM1385 at a given default frequency. At the other end of the scale the Microcontroller has connection to the System Controller and it's just waiting for a release of firmware to add adjustable voltage. Off at an angle from the extremes  Smiley we have  Cyper_BLC theory that it is a second layer of closed loop, safety based voltage control.

The reason B8 is not responding to a change in it's supply voltage is because as fast as you are winding the volts up the Buck Converter is winding them back down again  Smiley just like an S1 or S3.


Dearly wish that it was not so expensive to get an Batch 8 S7 to the UK as I feel that in the long run, after the halving when efficiency improvements become the name of the game, it will be the Batch to have.

Rich
legendary
Activity: 1596
Merit: 1000
It's fascinating watching you guys discus all of this. I wish I could contribute, but it's a good country mile above my head. Here's hoping a way to adjust voltage is found, since B8 definitely doesn't respond to increased 12V from the PSU as earlier batches do.
hero member
Activity: 588
Merit: 500

Smiley thanks, i can write a program for this pic and we can use a voltage adjuster but more reason to dont do that:
-i havent got b8
-safety first
-it can be adjust external, however i can make an plugin to beagle bone black with unused pins.

Agreed all sorts of changes and additions are possible, but first it's important to understand the original function of the circuit. Unfortunately with just the pictures I have got to the limit of the guesswork I can do so let's see what J4bberwock comes up with?

Rich
sr. member
Activity: 490
Merit: 270
Reverse Engineer

i dont touch that board, i dont see in my hands, i dont measure that, my posts for this subject are only guesswork ofcourse.
but engineer guesswork. if i made this board i use max813 not this one. this is bullshit compansate safety circuit.

Edit:
also dont forget the S7's boot-idle-working voltages are different. this closed loop circuit work for these too. Wink

OK good... I have a strong respect for "Engineer guesswork" I do it all the time... However when trying to understand a circuit I think it is important to differentiate between an educated guess and an actual check or measurement that has been made.

It's all to easy to make a mistake, for instance I agree and stated that it was likely that R21 & R22 are pull ups on the Data & Clock lines to 3.3V however when someone states that an unmarked resistor is 4.7K then I assume they have measured it as there is no "standard" pull up value for I2C.

We will have to wait for someone to makes some actual checks and measurements to establish if the function of the PIC & Digital Pot goes beyond a Factory setting of the Core voltage...



Rich

Smiley thanks, i can write a program for this pic and we can use a voltage adjuster but more reason to dont do that:
-i havent got b8
-safety first
-it can be adjust external, however i can make a plugin to beagle bone black with unused pins.
hero member
Activity: 588
Merit: 500

i dont touch that board, i dont see in my hands, i dont measure that, my posts for this subject are only guesswork ofcourse.
but engineer guesswork. if i made this board i use max813 not this one. this is bullshit compansate safety circuit.

Edit:
also dont forget the S7's boot-idle-working voltages are different. this closed loop circuit work for these too. Wink

OK good... I have a strong respect for "Engineer guesswork" I do it all the time... However when trying to understand a circuit I think it is important to differentiate between an educated guess and an actual check or measurement that has been made.

It's all to easy to make a mistake, for instance I agree and stated that it was likely that R21 & R22 are pull ups on the Data & Clock lines to 3.3V however when someone states that an unmarked resistor is 4.7K then I assume they have measured it as there is no "standard" pull up value for I2C.

We will have to wait for someone to makes some actual checks and measurements to establish if the function of the PIC & Digital Pot goes beyond a Factory setting of the Core voltage...



Rich
legendary
Activity: 1120
Merit: 1001
I have batch 8, running at stock 700 Mhz
legendary
Activity: 3892
Merit: 4331
My S7 are getting lots of HW errors and are underperforming
I have 4 of them ,only 1 is getting low HW errors and good performance

The rest of them perform like this one : https://gyazo.com/af319f55fa78b0ed11626611e76888b0
with HW errors around 0.03 % or higher

How can I fix this ?

3min is way too short a time to make conclusions. what batch is this?
see the dynamics for longer (at least 1hr). Sometimes they start slow, then improve.
sr. member
Activity: 490
Merit: 270
Reverse Engineer
My S7 are getting lots of HW errors and are underperforming
I have 4 of them ,only 1 is getting low HW errors and good performance

The rest of them perform like this one : https://gyazo.com/af319f55fa78b0ed11626611e76888b0
with HW errors around 0.03 % or higher

How can I fix this ?

change the power supply
or decrease the freq.
legendary
Activity: 1120
Merit: 1001
My S7 are getting lots of HW errors and are underperforming
I have 4 of them ,only 1 is getting low HW errors and good performance

The rest of them perform like this one : https://gyazo.com/af319f55fa78b0ed11626611e76888b0
with HW errors around 0.03 % or higher

How can I fix this ?
sr. member
Activity: 490
Merit: 270
Reverse Engineer
these resistors as you shown are i2c standart 4.7k serial data-clock pullup resistors on 3.3 Vcc line.

So you have checked that the PIC is completely stand alone from the rest of the board with no connections  to the LM75 I2C or the Controller connector?


Rich

no i dont, i said digital pot connected to pic via i2c bus with normal pullup resistors, if Dpot connects to lm75 i2c line, it must have a adress like:
                    Ta0 Ta1 Ta2
hashboard A-  0     1     0
hashboard B-  1     1     0
hashboard C-  0     0     1

but it hasnt got an id pins.
i said before this is closed loop safety circuit.


So can you just confirm that everything you are saying is just guesswork as opposed to checks and measurements you have made on a hash Board?

I have given my comment above on the "safety circuit".




i dont touch that board, i dont see in my hands, i dont measure that, my posts for this subject are only guesswork ofcourse.
but engineer guesswork. if i made this board i use max813 not this one. this is bullshit compansate safety circuit.

Edit:
also dont forget the S7's boot-idle-working voltages are different. this closed loop circuit work for these too. Wink
hero member
Activity: 588
Merit: 500
these resistors as you shown are i2c standart 4.7k serial data-clock pullup resistors on 3.3 Vcc line.

So you have checked that the PIC is completely stand alone from the rest of the board with no connections  to the LM75 I2C or the Controller connector?


Rich

no i dont, i said digital pot connected to pic via i2c bus with normal pullup resistors, if Dpot connects to lm75 i2c line, it must have a adress like:
                    Ta0 Ta1 Ta2
hashboard A-  0     1     0
hashboard B-  1     1     0
hashboard C-  0     0     1

but it hasnt got an id pins.
i said before this is closed loop safety circuit.


So can you just confirm that everything you are saying is just guesswork as opposed to checks and measurements you have made on a hash Board?

I have given my comment above on the "safety circuit".


sr. member
Activity: 490
Merit: 270
Reverse Engineer
these resistors as you shown are i2c standart 4.7k serial data-clock pullup resistors on 3.3 Vcc line.

So you have checked that the PIC is completely stand alone from the rest of the board with no connections  to the LM75 I2C or the Controller connector?


Rich

no i dont, i said digital pot connected to pic via i2c bus with normal pullup resistors, if Dpot connects to lm75 i2c line, it must have a adress like:
                    Ta0 Ta1 Ta2
hashboard A-  0     1     0
hashboard B-  1     1     0
hashboard C-  0     0     1

but it hasnt got an id pins.
i said before this is closed loop safety circuit.
legendary
Activity: 2294
Merit: 1182
Now the money is free, and so the people will be
I switched to Kano pool, looks good, I like that stats update really fast, and you have lots of them.  UNFORTUNATELY, I had to change settings in the antminer batch 8, big mistake, as it took many "soft reboots" and 2 actual "reboots" to get it at 4720ghs, with good hw.  I also see lots of X's for the chips, but thats just a bad reading or something because the hashrate is good and it goes away after 5 minutes.  

Note to self:  dont ever change anything in the settings ever again.  ever.
hero member
Activity: 588
Merit: 500
these resistors as you shown are i2c standart 4.7k serial data-clock pullup resistors on 3.3 Vcc line.

So you have checked that the PIC is completely stand alone from the rest of the board with no connections  to the LM75 I2C or the Controller connector?


Rich
sr. member
Activity: 490
Merit: 270
Reverse Engineer
I think the PIC & Digital Pot are at the moment in place to allow a Factory adjustment to the Core Voltage to be made either to cope with batch variations in the BM1385 and / or to optimise each hash board for it's default frequency?

What does not quite make sense if this was the only objective is why they did not just use an EE as opposed to a RAM based Digital Pot and not bother with the PIC?

My best answer to this is that there is connection from the Controller to the PIC to allow at some future date a version of firmware to be released that allows setting of the Core voltage or even an automatic selection of core voltage with a frequency change?

When someone gets a chance tracing the connections to the PIC will reveal what is or could be possible?


I have traced what I can from the pictures and everything makes sense around the PIC & the Digital Pot as regards the PIC being factory programmed from the row of contacts on the board edge.

What would be worth investigating is where if anywhere the I2C data & Clock connections go to on the rest of the board,?



The red traces show the Data & Clock connections from the PIC to the Digital Pot, however there is also through R21 & R22 additional connections that connect to the inner layer on the PCB, or are these just termination resistors. Question is do they pick up the main I2C bus used by the LM75 Temperature IC or go to separate connections on the Controller connector?

Rich

these resistors as you shown are i2c standart 4.7k serial data-clock pullup resistors on 3.3 Vcc line.
legendary
Activity: 1120
Merit: 1001
My s7 was held by customs because i have imported 4 of them and max is 3 per person Sad they will not release them all i can do is ship them back.do you think bitmain will ever refund me or what will happen? They could also re ship them but with 2 per package not all 4 of them

I thought that Romania is now in the European Union, but you still live in  Nicolae Ceaușescu  time.

You need a tax specialist. In EU there are no such restrictions, someone is lying to you.

I've bought past 4 months 44 Antminer (3x S5+ and 41x S7) from Bitmain and paid VAT over 13000 euro . There are no limits if they are for personal use only.

I finally managed to get them out of customs, after long long struggles !!


Every country has it's own rules regarding imports, I belive Romania has some weird rules where you can't import big expensive products on personal name, only on company is allowed
hero member
Activity: 588
Merit: 500
I haven't been on this forum long, my interest in mining is what brought me here. The amount of knowledge present in this forum leads me to believe that a kick-starter or grassroots movement to build mining equipment as an alternative to the Chinese companies is not far-fetched. I will start a separate thread and see what others have to say. Pessimists need not apply and may remain here to bitch about Bitmain, ROI, etc.

Please don't start a new thread, have a read of this one and add your thoughts to it....

What does it take to start manufacturing miners?
Jump to: