Topic: [Hands On] Bitmain AntMiner S7 - Batch 8 - Notlist3d (Read 15166 times)

Have you identified where on the controller board the PIC connections go to? A thing worth trying is to read the code from the PIC, manufacturers do not always remember to code protect. If you have the means to do this please give it a go, if not PM me and I will send you a PicKit2 which I have several of left over from a previous project.

AS this is getting a bit detailed I wonder if you should start a separate thread for this stage of the project?


Rich

Thank you Rich for identifying U2. Now it is clear that PIC can control buck converter voltage and PIC is connected to control board - two pins from PIC (pin #2 and #4) go to controller board (TX and RST2).

My board has currently malfunctioning buck converter, it has only 8.7V across the chain which I belive is the lowest voltage set without digital pot, so my best guess is that we should be able to set PIC digital pot to control voltage from about 9V and voltage can be set even more that 10.59V.

PIC is controlling U2 which is an I2C Digital Pot. MCP4017 50K 128 Steps.

The unknown was if there is any connection from the PIC to the main controller board in the S7 or as I suspect the PIC is set up by the factory from the board edge connections to a preset resistance value for the Digital Pot & hence the Core Voltage?


Rich

I did some deeper investigation, hope you will help me with some points.

1) Buck converter is set at 10.59V
2) PIC controller IS connected (through U2) to Feedback line of buck controller, so PIC controller is either controlling (more probable) or reading voltage. Or both.
3) Still don't know what U2 is.

U2: This IC uses 3 pins. Two coming from PIC, one goes to FB. Still unclear which ports of PIC are used - PIC has SW configurable pins, so we can guess about configuration.
PIC: no pin goes to buck converter, two pins go to controller board (TX and RST2)

Pin connection:

U2:

1: 3.3V
2: GND
3: PIC pin 6
4: PIC pin 5
5: GND
6: line to feedback of buck convertor

PIC:

1,8: Vdd, Vss
2: TX
3: P2 pin 1
4: RST2; P1 pin 1
5: U2 pin 4
6: U2 pin 3
7: P1 pin 4

http://pantin.cz/20160209_155344.jpg

So, I think PIC pin #2 receives commands or simple voltage settings through USART.

Rich, I send you a PM, could you please read it.

Yes I have seen that a couple of times on my S5. I assume that a counter has wrapped round or overflowed into another variable.

Rich

In reality no, but after it runs for about a month it is a low number and the equation behind it gets confused it seems.

Is that a negative ) on the % of hw errors? ))

Vertical and horizantal orientation is directly related to aerodynamics.The importan point is hot air rises up, if you have a some device to transfer heat out it is going to be very efficient.

The heat sink's falling problem is not the gravity problem but still an aerodynamic problem.Some points can not be cooled enough, these not enough cooled extra hot points reason not well oriented and applied heat sinks to fall.Not well oriented and applied means if the chemical is not perfectly under or less then the neccesarry enough amount.

S7 Batch 8 is a good miner for me on up-time.  Over 27 and 1/2 day's uptime still hitting average of 4.7T.   I am using the bitmain PSU on it.


One thing which I saw in S5+ aswell is it seems HW error can go negative if run for around a month like this.   So that is impossible.  But the main thing is the S7 is stable and running good.

I'm building a vent in the ceiling of the Garage so figure the rack being 72" high, I can just have the S7 exhaust fan pointed up.   It would pull the cooler air and push the hot air up, since hot air rise anyway, it would be the most efficient.  Yeah, Mark did mention about the Heat sink issue where they can fall and jam a FAN causing it to over heat.  I'm just wondering how it can burn down the unit if the system is designed to turn off when it's 80C.  I would think the cooling of the unit via the fans would be exactly the same as it's thru the same short tunnel.   If I have it lay on the side then I'll need to create a Plenum to funnel the hot air to the vent better.   

Deleted a post that as pretty much double post from regular S7 thread.   Don't delete many so I try to mention when I do.

Not much reason to have 2 conversations https://bitcointalksearch.org/topic/m.13564443 is where other one is if anyone want's to comment.

I have had good luck with laying regular.   I do use some extra fan's to pump all the hot air though the miners quickly and twords my exhaust.  So far have had good luck with it (winter has helped to ofcourse).

It's a interesting idea on risk of heat sink falling being more likely.  I'm not sure that the weight of being turned up would do it or not.  But it would be getting the board hotter exhaust on back of boards instead of top so depending on how good the adhesive is you might have problems from the heat in a different pattern then they tested.   

Do you no have the space to push the air through it normal with laying down?   Is space the issue where you have miners? Or are you just wanting to do a different setup?

Anyone have experience using the S7 in an upright position where it's sitting on a wire shelf with the exhaust fan pointing to the ceiling and the intake fan at the bottom?  I'm hearing that this would increase the risk of the heat sink falling and burning out the boards but when I look at the pictures of the internal, regardless of what position there will always be some heat sink that will be in a vertical position.  Any other reason why I should avoid this particular configuration?  Intake air wont be block as it's wire shelves below have plenty of gaps for air.  My goal is to point all the exhaust fans to the ceiling Vent for better efficiency.

yeah, its great in the winter, i usually just put a few antminers near a slightly-open window and they run quietly and without being too hot

It's getting colder and colder here.  They are still working in 20F or below here.   So it's good to see cold does not seem to have an effect.

I have also turned off some fan's.  Winter mining is a heck of a lot easier then summer mining when mining a decent scale.

its around -10C outside here. i have a quiet 120mm fan pulling cold air directly to the S7 intake

S7, B8 fans are 15%, 691MHz, and I have temperatures of 61-68C

I've ran some miners where the chip were at -6~-10c, they worked fine. I'm guessing as long as no condensation freeze at temp variance there would be no problems.

No problem at all.  This is for any batch 8.  That is very low on testing but seems to work fine which is nice.

Lowest I have got is around 30F or so but no issues running in it.   Also started mining around December 19, 2015. or so and no fault's or errors which is nice.  So guess were like 1/2 month of running great.

hope you dont mind me posting this:

5min test at -10C  

Oh sorry, my mistake, other side of PCB is PIC12F1572 controller Buck controller seems to be LM27402 driving pair of toshiba TPHR9003 high side and only one at low side.

The pictures show a PIC and a Digital Pot on the Batch 8 Boards. What is unknown at the moment is if it is stand alone /Factory Pre-set or if it is controlled by the main controller?

Rich

So, B8 has no PIC controller. Voltage is set at 10.5V. I don't think it will be SW controlled.

0.0078%/0.0186% only

and power consumption from the wall? (and what type PSU?)

HW errors must be high

I have two B8 OCed at 750MHz hasing at 5THs and 5.03THs. The garage is very hot (total about 15kW) and miners are running at 70/71C.

I think this is chinese - english translation error. Bitmain's reply is explaining that INPUT voltage can be variable (adjusted), while frequency remain fixed and miner should work.

I don't think anyone has pushed it hard.  At 700 MHz on default it just is set pretty high already compared to other batches.

I don't see people pushing them past that to much.    Risk/reward just is not there especially with losing warranty if caught doing it.

Ive tried to go as high as 712MHz successfully with errors rate of ~0.015% and temps of 55-65.  Didn't go any further because i dont have a kill-a-watt and am only using a single 1350W (enermax gold) PSU.

I suspect bitmain instilled the higher voltage in order to maximize hashrate/chip for short-term ROI. I suspect a B8 unit could exceed 800MHz/1420wDC/~1500wAC

a higher voltage means less amps drawn for the same wattage. so whatever overclocks a lower-voltage chip can achieve should be possible to exceed with a higher voltage in the B8

IIRC S2 had *some* issues with that though, due almost entirely to the beaglebone SD card (same reason for almost every BBB/Rpi+SD-based miners failing after power cycle)

Did someone made some OC on the batch 8?

Bitmain's official response would support your idea of factory set fixed voltage:


It's been a bit since they sent that to me.  But it did mention fixed frequency.

It is because of this sort of problem that I suspect that Bitmain have gone for a factory set fixed voltage. it is however possible that they have chosen a different set voltage for Batch 6 v Batch 8/9?

One thing that keeps alive the possibility of a dynamic voltage setting is that if they had just wanted a factory set Core Voltage they could just have used an EE based digital pot with no PIC?

However anything is possible right up to a programmed selection of voltage dependant on the frequency selected? Some tests and measurements will be good.  


Rich

I will try to do my own investigation tomorrow, but C1 series had similar DC-DC and I think also with PIC. C1s DC-DC was very unreliable, some of them failed to set-up correct voltage and overvolted chips and shut down whole board. It is very strange as C1 DC-DC set-up voltage at 0.75V, but some voltage comands (after cgminer starts) result in wrong voltage setting. This happened over the time and some of blades can't set up even default 0.76V. Bitmain supplied "hot patch" by unsoldering buffer chip, so voltage stays at start-up level.
 

The batch is doing well in the cold:


It's low 30's outside and that air is going into my mining area.  So far no issues with S7's with dealing with the cold.   It is safely where no water will effect I should mention to.

There is a PIC Microcontroller and a Digital pot controlling the Fb ground resistor. What is unknown at the moment is if this arrangement is tied into the main Controller in any way or if it is stand alone and either factory pre-set (most likely) or monitoring some parameter and adjusting accordingly?

I do not own a Batch 8 S7 so can run no checks, most likely that will be done by J4bberwock in the New year?



Rich

So, there is digital pot or digital dc-dc as seen in C1? So if it is digital, we may be able to controll it via SW or resistor VID settings?

underclcok it down to 100M sure quiet.

A mod of some sort will be possible, but we need to fully understand exactly how they are using the Buck Converter & Digital Pot first.

Rich

Anybody tried to adjust voltage on buck converter of S7B8? Still waiting for my piece. Maybe another pencil hack will do the trick?

I am considering to undervolt to keep noise below certain level...

I never noticed any issues when hard switching off/on any antminer. All series boots up perfectly.

I had a power outage.  The good new's is the S7 seemed to handle it great.  No issues from turning off in not the greatest way, and it booted up and worked fine after power was on.

So it get's some point's for being reliable which is important.

What I got out of it and I could be wrong this is my opinion.  They have it where it get's 4.7TH they seem to not have any issues with being to far off that I have seen.

So they kinda locked it down to where you get what is advertised.  It seems like they are keeping design longer then others it seems.  We know for sure batch 9 will use the same, and it could be possibly longer.

Yes but we know that batch 8 , as are all batches, is frequency adjustable and that it is not psu adjustable at the PSU level because of the buck converter.. so....  they must have made a mistake and the only half sensible explanation is that they meant that " currently it is fixed voltage but the frequency can be adjusted"? Implying that at some point in the future it will be voltage adjustable? Or... have I been drinking too much as it's Xmas Eve?


Rich

Yes should be batch 8 I asked specifically about it.  They apologized about how long it took.  But said one of the engineers was consulted to get a anwser.

So it should be petty valid info.

try an adjustable psu like this

https://www.trcelectronics.com/ecomm/pdf/rsp1500.pdf

or the finksy/j4bberwock one like this  note the 12.17 volt is middle speed.

As long as this is referring to batch 8 perhaps they just got voltage & frequency the wrong way round? However might imply that in the future Voltage will be adjustable?

Rich

They say the voltage can be adjusted? That make it sound like the software control is back in. But what do they "actually" mean, since we know they blatantly mislead with whatever they say?

That you can undervolt it at the PSU level?

That's funny and a little misleading if you look closely.

Instead of saying the power consumption is ".275 J/GH" they just say ".25 J/GH + 10%"

Got the official response back on efficiency:  "Currently, S7 is fixed with frequency to make sure the hash rate is more stable. but the voltage can be adjusted. The power consumption is 0.25 J/GH + 10% as marked in website."

So it appears the 4.7 is constant.  And that efficiency is the +10 marked in specs.

You're right alh, he did do it on older Batch, I'm confused as Phil mentioned he got 4 Batch 8 on order.

I'd swear that Phil had a Batch #1 S7, certainly not a Batch #8. He can correct me if I am wrong. Preliminary photos suggest that there is a different voltage regulator on the Batch #8 S7, that would likely nullify any fiddling with the 12V actual voltage (i.e. a small bump to say 12.5V).

Kilo17,

I changed my Batch 8 to 775 and 800M Freq and only got more hardware errors, no additional performance gains.  This is on the 2K IBM 220V PSU without any voltage changes.  Phil on the other hand bump it up to 12.55 V and he saw a small uptick in performance on the S7 and a huge uptick on the Avalons so it seems that higher voltage is required.  These at stock are already at the highest default OC level compare to other batches so quite frankly I think I will just leave them alone until others have shown stability over a month or so.  Not worth frying them before Warranty is over.

Have not got an official response on electricity usage I know some were waiting on it.  So of you measure usage it is appreciated if you put usage and PSU you used.

But it appears it is less efficient then the batch before it.   Still a more efficient miner overall but not to what some were experiencing on some batches.

I'm more into providing information and i let everyone decide on their own.  I don't have a Buy/Don't buy as I feel i overstep a review if I start to say this.  Also I make nothing if you buy or not, this is on purpose.  I don't use short links, or referral in these as I don't have a percent I take off sales.   I try to be impartial you will see direct link to official site.

And it is ever-changing as far  as prices.   So one buy/don't buy does not work out forever.  As of today 12/21 the lowest price per hash power is the S7 Batch 8.  I have been waiting on official statement as far as efficiency, have not got it yet.  So if your bottom line is hashing power, not sound chances are as of right now lowest price per hashing power.

If you need to have it in house as you mention if sound is important Avalon 6, and quieter PSU might be better option.  So it depends on your situation in all honesty.  It's not one size fit's all.

so a bottom line cliff notes?

Pros and cons of each, or just an outright opinion on which you would buy if looking to buy another miner?

Also thanks for your review of the Avalon 6, I got one and am quite happy any my house is much warmer 

My temps are pretty low with temperature's currently it's winter weather.  I need to slow down some fan's just haven't got to it yet in mining area.  Mine are pretty consent 4.7TH.   The hashing average rate is pretty constant that it does not seem to vary much at all for me. Below is running currently:

Been Running my Batch 8 for 1d 10 Hours and the average is 4719.56 here.

My S7 Batch 3 averages 4855.18 and the temps are better on the Batch 3 as well.

For some reason on my Batch 8 the first card consistently runs 5C hotter than the other two.

I think most will leave at default frequency. I don't expect anyone to really push it with it being at so high already.  It is pushing freq on default as shown below:


I do not think they will keep jumping all over on range for near future, but I could be wrong.  They have gotten a design that I think they will stick with, Batch 9 has same specs as Batch 8.   Also still on issue of running I sent that email asking on efficiency I am hoping Monday i get something back.

Any chance that you have messed with the Default Frequency?  The range has been all over the place on each batch and I was wondering if the Default Frequency of 700 on Batch 8 limits the overclocking ability?

thanks....

The s-7 is way cheaper then the Avalon six is and does more hash.

These are the only two you can buy with .25 to.32 watts a gh


I have an S-7 and two Avalon sixes.  All three overclock well.  But if you were in my garage and asked me to tell you phil should I buy the s-7 from you or the avalon six from you.

Based on money I would tell you buy the s-7.
Based on sound I would tell you buy the avalon sixes.

 All in all. Do you think it's the best choice for the money??

Here is running at around 17 hour's.  Getting right around 4.7 TH all default settings.

Well those a pretty beefy mosfets, so just need some sort of controller chip and we are back to perhaps having a way to adjust the voltage?

Would be great for someone to measure the voltage across the full chain? but understand the warranty issues.

Rich

Right behind the inductor. There are 3 TPHR9003NL mosfet. I didn't take a photo when I took out the hashing board

If you have the fans off again can you do a picture at the best angle you can get to show the back of the board on the other side from the new big inductor? I am beginning to have doubts about there being a buck converter chip there? Various reasons, which I won't go into for the moment.


Rich

Mine has still not been on very long around 4 hours my average is: 4,710.30 .   It is using the bitmain PSU on a 240.  Still don't have the tools to measure watts yet.

I will let it run over the night and post a screenshot in morning a 12 hour test hopefully will be a little faster then my 4 hour one but chances are it will remain close to what it is now.

Barely 4.73th as advertised. Getting about 4.6-4.7th

Thank You! I appreciate measuring it.  With my current 240 setup my normal means of measuring are not working.  What speed are you getting at that wattage?

I'm getting about 1440watts. Using 2 ibm 1440watter to power the miner.

Part of the reason for this thread is that things have changed since Batch 1's so much.  The reason I don't have a wattage like I normally do is it's on a 240 which with current equipment closest I can get is in amps... so not to good.  We want more accurate then me guessing, and I expect that to come.  

I realize you want exact and correct numbers so were on same page as far as what we want.   Just give it a few day's or little time before bashing it to much.  The goal is to get accurate information as most of reviews were based on old batches (1 was most).  I think we need a more up to date look at it.   And as people report usage with different PSU's it will come I think.

Also put the ** above in specs showing it is being looked at.  I hope that will allow us a little time to get different PSU's and efficiency. 

Bah, why not check with a wattmeter what the draw at the wall is for Batch 8? For example the S5 says 0.51J/W, and i get 0.49J/W actual. If for the S7, its 0.27J actual at the wall, then its not 0.25J efficiency at all. If you add a 10% to that, you're almost at 0.3J/W efficiency. Either way it sound like its not 0.25J anymore, since the first few batches which actually were.

For previous batches people reported that this time around (For the S7 overall), the efficiency rating listed was DC, not actual consumption. And that J/GH has been getting worse as later batches came to be.

I should get an official anwser in email, but I expect it to be during the business week most likely.  But your input is very much appreciated if you do measure it and post your results in thread.

Don't bother, they aren't to be trusted. We need to measure these things our selves. I'll eventually get around to measuring DC and AC on my B8 and post results.

They do give it a percentage where at the wall it could be higher like you mentioned.  

1. Hash Rate: 4.73 TH/s ±5%
2. Power Consumption: 1293W + 10% (at the wall, with APW3, 93% efficiency, 25C ambient temp)
3. Power Efficiency: 0.25 J/GH + 10% (at the wall, with APW3, 93% efficiency, 25°C ambient temp)

I will send them a email to verify if it is closer to .27 on actual at wall. (It's weekend in China I might not have a instant anwser).
*Sent them a email asked on 240 with recommended PSU (The Bitmain PSU) well update once I get anwser.

Here's something to get out of the way right away;

1. Hash Rate: 4.73 TH/s ±5%
2. Power Consumption: 1293W
3. Power Efficiency: 0.25 J/GH

1293/4730 = 0.27~ Efficiency. So it is 0.25 efficient or not. It says +10% at the wall, as if you have to add it to the 1293W, then that would be even less efficient. So what is the truth?

0) General information
1) Unpacking
2) The Miner
3) Install
4) Running the miner
5) Conclusion
6) How to purchase

0) General Information
The Bitmain S7 has been a very popular miner with current gen miners. With most reviews covering the earlier batch (most batch 1's) I felt it was worthy to do a more current one with a batch S7 from batch 8. They have now went to a specification which I expect them to keep for at least the near future and possibly longer. They currently are shipping out within 5 day's after payment.  The efficiency and recent drop in price make this a real contender in current gen miners.  This miner consists of 3 main modules with 3 blades in each.  Together they bring a hashing speed of 4.73 TH at only 1293 watt's, making it run approximately at a efficiency of 0.25 J/GH. All of this with the familiar Bitmain GUI you have come to expect, making setup very easy.

Specifications:
1. Hash Rate: 4.73 TH/s ±5%
2. Power Consumption: 1293W + 10% (at the wall, with APW3, 93% efficiency, 25C ambient temp)
3. Power Efficiency: 0.25 J/GH + 10% (at the wall, with APW3, 93% efficiency, 25°C ambient temp) **Has been questioned as shown below.  Email sent also asking for people to report their efficiency.
4. Rated Voltage: 11.60 ~13.00V
5. Chip quantity per unit: 135x BM1385
6. Dimensions: 301mm(L)*123mm(W)*155mm(H)
7. Cooling: 2x 12038 fan
8. Operating Temperature: 0 °C to 40 °C
9. Network Connection: Ethernet
10. Default Frequency: 700M

1) Unpacking
The miner is very well packed as you can see below.  It is within a special styrofoam packing, that helps keep the miner in the center of the box.  This helps keep the miner safe during shipping, and there is a message to owner that is on-top so you read it before running miner.



2) The Miner
Below are images to showcase the miner.




3) Install
One of the big things before install is to make sure all heat-sinks stayed in place during shipment.  There are different way's to do this. I did not want to void warranty during my check.  I chose to take fans off on both ends and look inside to ensure no heat-sinks had fallen during shipping.



The install can vary greatly for each person on electricity. For those of you familiar with my guides I used the same 240 volt 30 amp connection to a 30 amp PDU I used in my S5+ review.  From this PDU I used one of the Bitmain server PSU suggested which came with the needed 10 PCIe cables, making the install very easy. ( https://bitmaintech.com/productDetail.htm?pid=000201505040743496917U7kGsCm0694 ).  

The install time of S7 and PSU went very quick plugging into PDU (I did turn off power to PDU to not be plugging in cables that are live).  This was done in only a few minutes.  After install I did set a static IP as I always do then went to the very familiar and easy GUI to set up my pool information.

Below is a picture of the S7 installed with the Bitmain PSU:



4) Running the miner
Was very easy to set up and get running if you are familiar with the AntMiner GUI.  At this point I changed network settings to set up with a static IP, and my pools are inputed after. Below you can see the S7 hashing away after setup.
 
Below shows the miner actually mining:



5) Conclusion
This miner takes the BM1385 chip and makes a powerful and efficient miner.   For me it was a very simple and easy install, there were no issues during it.  Using the recommended Bitmain psu really does help make it a simple install for those with 220/240. The hashing power you get with efficiency makes the new batch 8 with the price drop from batch 7 a great choice currently.  I will continue to run and test out the S7 batch 8 and update this, but I expect there to be no issues.

6) How to purchase
You can go directly to the Bitmain's website: https://www.bitmaintech.com/productDetail.htm?pid=000201511170341298180m44675v0613