Pages:
Author

Topic: What frequency works best for Antminer S7? - page 2. (Read 18263 times)

legendary
Activity: 3892
Merit: 4331
December 14, 2015, 04:01:22 PM
#49
Hello,

I have a antminer S7 from B6. After a week of running, one hashboard got only 36 chips running instead of 45. Anyone experience with this? Can I resolve this or are the other chips broken / dead??

Thx

five four ways to fix this:

1. go to pool settings and simply click on save and apply (one time I had the same and this simple procedure fixed it).
2. go to "advanced" in gui and simply click on save and apply.
3. #1 followed in 10-15 min by #2.
4. go to system, click on reboot
5. hard reboot (if you have access to PSU).

I would do either #1 or #2 first, then see what happens and do another option.
full member
Activity: 182
Merit: 100
To the MOOOON
December 14, 2015, 02:13:47 PM
#48
Hello,

I have a antminer S7 from B6. After a week of running, one hashboard got only 36 chips running instead of 45. Anyone experience with this? Can I resolve this or are the other chips broken / dead??

Thx
member
Activity: 84
Merit: 10
December 06, 2015, 02:45:43 PM
#47

Honestly for me every miner is a little different.  With the exception of the 2 batch 7's the firmware is on those factory installed and on one of them I left it at 625 MHz and it mined at the advertised 5TH, the second batch 7 at 625 was performing poorly and I was only getting around 3.8TH so I dropped the freq to 600 and then is jumped up to 5TH.  The 19 others I have are batch 2 (which were advertised to run at 4.66TH at 575 TH), I flashed them with the most recent Bitmain firmware that supports the 600Mhz setting.  When I had the time I found that really going through them slowly and taking my time tweaking with the frequency (and not always up) making small changes has made an incredible difference.  For instance, I have some of mine set at 606MHz and some set 593Mhz.  I do not think I have anything set over 625 but you can certainly try.  Just make sure you are monitoring it closely when you make changes like that. I guess the whole point I am getting at is there is not perfect setting for every single miner out there.  You can have rough guidelines that work for most but if you are trying to dial in your farm, then you need to play with the settings to see what works with your farm.  Also keep in mind not all miners are going to perform at the highest Th rating possible.  Some of my miners no matter what settings I have changed and close I think I have dialed it in, I have had to settle for 4.5Th when I should be getting 4.8  and that ok, its just the way it goes.  If you are interested you can check out my mining set up here


Very interesting. I am a firm believer that all miners are different and benefit from monitoring and adjustment to get the best from them. I have a couple of questions?  Smiley

If I understand correctly your approach seems to be to find the frequency at which the best hash rate is achieved? When doing this do you also look at the HW error rate and do you find when the Hash Rate is lower than expected for a Frequency that the HW error rate is high or is the error rate still ok?

Next have you considered that there could be a difference in your PSU's and that supply voltage that the miner has on the PCIe connectors could be different in some cases and that is influencing it's ability to hash at higher frequencies?

If you are not monitoring the voltage I would urge you to include it in your adjustment protocol as it may well add another dimension and enable you to achieve higher hash rates?


Rich


These are all good points Rich

Regarding the power supplies I stick with the same kind of power supplies up until just recently when I changed over to EVGA for my last 5 S7's.  I don't really factor my power supply's voltage bouncing as a reason for the difference variations in miner performance.  If I was have voltage problems then I would think my problems would be much bigger than what I have previously described.  (which by the way I do not consider a problem) I would lean towards mass produced miners over high end power supplies but that is just me.  Regarding HW errors, yes I do get them, we all get them and that is something that would take forever to watch over a long period of time.  The truth is there are several things you can look at to monitor your minors.  The most obviously thing that people look for is how fast can it hash.  Now if I can get it hashing fast but the results are garbage and I am getting a lot of rejects, hw errors, etc then of course that is when I am going to dig deeper.  I wasn't trying to go quite that in dept as the question was regarding frequency settings.  I am pretty sure if you polled people on the forum they would all agree that not all minors behave the same and they doesn't mean we need to start to digging into power supplies/PCIe connections over a couple Ghs.  However that is just my opinion, and please note I completely respect what you had to say it is just not something I would do if personally as my farm is running great.  Some of my miners are over achievers and others are a little under achievers(very little) but overall I am averaging a happy 105TH.  Thanks for your feedback. 
hero member
Activity: 588
Merit: 500
December 06, 2015, 01:43:13 PM
#46
One interesting observation re S7 Batch 6.

After ~2-3 min of initial start and any hard reboot, suddenly there are a lot of XX appearing, but speed does not drop at all, remarkably.
In a few seconds all XX are cleared and speed stays or increases slightly to reach the steady-state.
I think that this shows the effect of the buck converter in action.
I wonder if others saw the same, but it is difficult to catch as it happens only at that time (2-3 min after hard restart).

I wrote about this effect earlier. November 30, 2015,
https://bitcointalksearch.org/topic/m.13113139


Very strange effect, and not one I have observed on an S5 or seen reported on the 162 chip S7? It is difficult to speculate on reasons, until we understand exactly what has changed in Batch 6/8. However... one fanciful reason could be that the firmware powers up with the Buck Converter set to a voltage, then if there are xx it ups it a bit?

Rich
legendary
Activity: 1218
Merit: 1003
December 06, 2015, 01:16:15 PM
#45
One interesting observation re S7 Batch 6.

After ~2-3 min of initial start and any hard reboot, suddenly there are a lot of XX appearing, but speed does not drop at all, remarkably.
In a few seconds all XX are cleared and speed stays or increases slightly to reach the steady-state.
I think that this shows the effect of the buck converter in action.
I wonder if others saw the same, but it is difficult to catch as it happens only at that time (2-3 min after hard restart).

I wrote about this effect earlier. November 30, 2015,
https://bitcointalksearch.org/topic/m.13113139
legendary
Activity: 3892
Merit: 4331
December 06, 2015, 12:42:38 PM
#44
One interesting observation re S7 Batch 6.

After ~2-3 min of initial start and any hard reboot, suddenly there are a lot of XX appearing, but speed does not drop at all, remarkably.
In a few seconds all XX are cleared and speed stays or increases slightly to reach the steady-state.
I think that this shows the effect of the buck converter in action.
EDIT: @Tupsu saw it on 11/30 I wonder if others saw the same, but it is difficult to catch as it happens only at that time (2-3 min after hard restart).
legendary
Activity: 3892
Merit: 4331
December 06, 2015, 12:36:01 PM
#43
Damn guys! Nice speeds!

As for B6, I only have one running atm and I'm at 625 frequency with no drama. Newer batches of presumably the exact same miner are coming out at 700 MHz

Well that is the $64000 BTC160 question. I see a few possibilities.

1) There is no difference and batch 6 will run at 700MHz
Possible, but maybe with a lot of errors, but it does run at 650 (see below).

2) They are selecting BM1385 and the Batch 8 ones are better than Batch 6
Most likely small mods to boards or yours #3

3) They have set the Buck Converter voltage higher in Batch 8
Probably, but we won't know what is the error rate on B8 until it is here

4) Something else.

S7 continues to be more of a mystery than any of it's predecessors and I look forward to the Day when a 135 chip version can be taken apart and some pictures taken and measurements made.  Smiley

Rich

S7 batch 6 at 650 (at 600-0.0001% error)

hero member
Activity: 588
Merit: 500
December 06, 2015, 12:21:45 PM
#42
Damn guys! Nice speeds!

As for B6, I only have one running atm and I'm at 625 frequency with no drama. Newer batches of presumably the exact same miner are coming out at 700 MHz

Well that is the $64000 BTC160 question. I see a few possibilities.

1) There is no difference and batch 6 will run at 700MHz

2) They are selecting BM1385 and the Batch 8 ones are better than Batch 6

3) They have set the Buck Converter voltage higher in Batch 8

4) Somthing else.

S7 continues to be more of a mystery than any of it's predecessors and I look forward to the Day when a 135 chip version can be taken apart and some pictures taken and measurements made.  Smiley

Rich
legendary
Activity: 1946
Merit: 1006
Bitcoin / Crypto mining Hardware.
December 06, 2015, 12:20:32 PM
#41
I've been experimenting too.


I didn't measure at the wall yet, but I did measure DC power coming from the PSU. Each blade is burning 500w, so 1500w / 5770gh = 0.259w/gh at the PSU. What's being pulled at the wall will depend on the quality of your PSU, obviously. I don't see why you're saying efficiency is bad, it's incredibly good considering it's running almost 1TH/s faster than stock.

To compare, stock 600MHz each blade pulls 412w. 1236w / 4860gh = 0.254w/gh at the PSU.


I'm overvolting the S7. PSU is outputting 12.58v unloaded, 12.48v loaded. running 600MHz = 12.4v at the blade, 737 = 12.33v at the blade.

I don't suggest anyone attempt this. It's definitely not good for the S7, and you risk killing it.
You got nice OC there  Smiley

Ahh, I got 0.28W / GH @ wall. For some reason I thought the efficiency should be 0.25W / GH at wall.
legendary
Activity: 4256
Merit: 8551
'The right to privacy matters'
December 06, 2015, 12:16:31 PM
#40
Damn guys! Nice speeds!

As for B6, I only have one running atm and I'm at 625 frequency with no drama. Newer batches of presumably the exact same miner are coming out at 700 MHz

yeah like the s-3 the new batches run much higher freq.

I am happy to have one really good one  and one so so one since I was early batch.

Once I get set with solar project I am hoping to run 20 kwatts worth of gear.
legendary
Activity: 1596
Merit: 1000
December 06, 2015, 11:41:22 AM
#39
I've been experimenting too.



I didn't measure at the wall yet, but I did measure DC power coming from the PSU. Each blade is burning 500w, so 1500w / 5770gh = 0.259w/gh at the PSU. What's being pulled at the wall will depend on the quality of your PSU, obviously. I don't see why you're saying efficiency is bad, it's incredibly good considering it's running almost 1TH/s faster than stock.

To compare, stock 600MHz each blade pulls 412w. 1236w / 4860gh = 0.254w/gh at the PSU.


I'm overvolting the S7. PSU is outputting 12.58v unloaded, 12.48v loaded. running 600MHz = 12.4v at the blade, 737 = 12.33v at the blade.

I don't suggest anyone attempt this. It's definitely not good for the S7, and you risk killing it.
legendary
Activity: 1946
Merit: 1006
Bitcoin / Crypto mining Hardware.
December 06, 2015, 11:18:10 AM
#38
Playing around with the clocks today. Efficiency is bit poor for a S7 though: Power@Wall=1600watts.
full member
Activity: 182
Merit: 100
To the MOOOON
December 06, 2015, 07:32:54 AM
#37
What is best settings for 3x 45 Asics is AntminerS7  batch 6.

The firmware from the website from Bitmain?
Or which frequency?

Thx in advance.


There is no firmware for Batch 6 yet. Default frequency should be 600MHz.

Rich

Yes the default I know, which is 600MHz.

But some hash at 687MHz, around 4.7 TH/s. While others only hash around 3 TH/s with this frequency. Some other experience with this?
hero member
Activity: 588
Merit: 500
December 06, 2015, 02:27:10 AM
#36
What is best settings for 3x 45 Asics is AntminerS7  batch 6.

The firmware from the website from Bitmain?
Or which frequency?

Thx in advance.


There is no firmware for Batch 6 yet. Default frequency should be 600MHz.

Rich
full member
Activity: 182
Merit: 100
To the MOOOON
December 06, 2015, 02:18:32 AM
#35
What is best settings for 3x 45 Asics is AntminerS7  batch 6.

The firmware from the website from Bitmain?
Or which frequency?

Thx in advance.
legendary
Activity: 4256
Merit: 8551
'The right to privacy matters'
December 05, 2015, 04:15:00 PM
#34
legendary
Activity: 1022
Merit: 1003
December 05, 2015, 04:01:21 PM
#33
hero member
Activity: 588
Merit: 500
December 05, 2015, 12:26:47 PM
#32

Honestly for me every miner is a little different.  With the exception of the 2 batch 7's the firmware is on those factory installed and on one of them I left it at 625 MHz and it mined at the advertised 5TH, the second batch 7 at 625 was performing poorly and I was only getting around 3.8TH so I dropped the freq to 600 and then is jumped up to 5TH.  The 19 others I have are batch 2 (which were advertised to run at 4.66TH at 575 TH), I flashed them with the most recent Bitmain firmware that supports the 600Mhz setting.  When I had the time I found that really going through them slowly and taking my time tweaking with the frequency (and not always up) making small changes has made an incredible difference.  For instance, I have some of mine set at 606MHz and some set 593Mhz.  I do not think I have anything set over 625 but you can certainly try.  Just make sure you are monitoring it closely when you make changes like that. I guess the whole point I am getting at is there is not perfect setting for every single miner out there.  You can have rough guidelines that work for most but if you are trying to dial in your farm, then you need to play with the settings to see what works with your farm.  Also keep in mind not all miners are going to perform at the highest Th rating possible.  Some of my miners no matter what settings I have changed and close I think I have dialed it in, I have had to settle for 4.5Th when I should be getting 4.8  and that ok, its just the way it goes.  If you are interested you can check out my mining set up here


Very interesting. I am a firm believer that all miners are different and benefit from monitoring and adjustment to get the best from them. I have a couple of questions?  Smiley

If I understand correctly your approach seems to be to find the frequency at which the best hash rate is achieved? When doing this do you also look at the HW error rate and do you find when the Hash Rate is lower than expected for a Frequency that the HW error rate is high or is the error rate still ok?

Next have you considered that there could be a difference in your PSU's and that supply voltage that the miner has on the PCIe connectors could be different in some cases and that is influencing it's ability to hash at higher frequencies?

If you are not monitoring the voltage I would urge you to include it in your adjustment protocol as it may well add another dimension and enable you to achieve higher hash rates?


Rich
member
Activity: 84
Merit: 10
December 05, 2015, 12:03:16 PM
#31
Please post your results, there are a lot of frequencies and setups, so please post the following in this format.



Batch: Batch#
Frequency: Freq#
Hashrate: Hashrate#
Power consumption: PowerConsumption# (If possible)
Fan Speed: FanSpeedPercentage#
Avg Blade Temperature: TempInCelcius#
Hardware Errors Per Day: HWerrors#


For example:


Batch: 6
Frequency: 575
Hashrate: 3.82TH/s
Power consumption: N/A
Fan Speed: 35%
Avg Blade Temperature: 54C
Hardware Errors Per Day: 3/day



Honestly for me every miner is a little different.  With the exception of the 2 batch 7's the firmware is on those factory installed and on one of them I left it at 625 MHz and it mined at the advertised 5TH, the second batch 7 at 625 was performing poorly and I was only getting around 3.8TH so I dropped the freq to 600 and then is jumped up to 5TH.  The 19 others I have are batch 2 (which were advertised to run at 4.66TH at 575 TH), I flashed them with the most recent Bitmain firmware that supports the 600Mhz setting.  When I had the time I found that really going through them slowly and taking my time tweaking with the frequency (and not always up) making small changes has made an incredible difference.  For instance, I have some of mine set at 606MHz and some set 593Mhz.  I do not think I have anything set over 625 but you can certainly try.  Just make sure you are monitoring it closely when you make changes like that. I guess the whole point I am getting at is there is not perfect setting for every single miner out there.  You can have rough guidelines that work for most but if you are trying to dial in your farm, then you need to play with the settings to see what works with your farm.  Also keep in mind not all miners are going to perform at the highest Th rating possible.  Some of my miners no matter what settings I have changed and close I think I have dialed it in, I have had to settle for 4.5Th when I should be getting 4.8  and that ok, its just the way it goes.  If you are interested you can check out my mining set up here

https://www.youtube.com/watch?v=K1C2LQnBr2Y

If you like it, subscribe as I am always adding new content and I love audience feedback.  I look at bitcoin as a group learning effort.  In fact your thread has kinda of inspired me to create some kind of video addressing this very issue because I have seen this question or variations to this question come up before where people have had performance issues when dealing with several of the same type of bitmain products.  They are frustrated why two or more of the exact same product do not operate the exact same way. 

I hope this help!
legendary
Activity: 4256
Merit: 8551
'The right to privacy matters'
December 05, 2015, 09:24:47 AM
#30
Amen to that...  Would love to know if efficiency has decreased at all at that clock speed, if you ever get a chance to measure ATW consumption!

when I used this psu with the s-7

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


I set volts to 13 and ran both the s-7 and the avalon 6

in both cases  efficiency drops a bit the s-7 is .26 on this psu at 600 clock and drifted up to .29 running at freq 631.25 my batch 1 got pretty hot and loud so I can't do long runs with fans blowing at 4200 rpm.

This psu was purchased for testing  the s-7 and avalon 6 for underclock not overclock.  And right now under clock under volt is a no go.

But I have determined this.  both the s-7 and the avalon 6 should have at least 12.2 volts  not 11.9  this means finksy + j4bberwock

has the best units to mine the 2x 2000  units with a few volts setting are the best gear (240 volts fuck the sound).

At finksy pm me when I can get one of them.

I no longer tell anyone to use ATX psu with this gear.  Unless you have only 120 volts power
Pages:
Jump to: