S17+ S9k/SE T17 S17/Pro DR5 B7 Others All Overclock per chip and Voltage Control

[Bran5471]

Keep up the good work Chipless,I see your just trying to get our aisc's control back,Shrug off anyone trying to put you down,or discredit your work. I will be following/using your CFW's.

[chipless]

Keep up the good work Chipless,I see your just trying to get our aisc's control back,Shrug off anyone trying to put you down,or discredit your work. I will be following/using your CFW's.

Thank you, the latest tweak has it up over 154 k/sol ,,,,, Baby steps I dont want to screw a 2k machine by pusing it too far. Takes about an hour to see what it tops off at so every tweak can take as much as 1 hr running


Elapsed    KSol/S(RT)    KSol/S(avg)
1h2m17s    164.96               154.3

2 hrs runtime... Thats where that tweak peaks out on my machine
Elapsed   KSol/S(RT)   KSol/S(avg)
1h53m26s   158.9             154.41

[weremy73]

Just sharing some observations. The beta 3 firmware held over 151-153 ksols for most of the day, but settled back down to 145 ksols recently with chip temps in the 81c-85c with fans set to 100%. Another Z11 twelve inches away running Bitmain 4/23 FW is at 70c-73c with fans at 95%. Re-releasing an April 23 version to cut down on hardware errors at the very least (and any other positive unintended consequences) seems like a no-brainer.
--------------------------------------------------------------------------------------------------
Beta 3:

Elapsed          MH/S(RT)          MH/S(avg)
8h44m23s      146.45              145.66

HW Errors: 7,813

Power Draw: 1780w @ 120v
---------------------------------------------------------------------------------------------------
Bitmain 4/23:

Elapsed          MH/S(RT)          MH/S(avg)
8h59m26s     132.37              137.99

HW Errors: 649

Power Draw: 1492w @ 120v
----------------------------------------------------------------------------------------------------

[Bran5471]

No problem man,overclocking an voltage adjustment is a touchy thing,one 1Mhz higher crashes when 1 lower is perfect,and I'm sure theirs more to it with asic chips. Bitmain locking them down was a pretty sad day,I was hoping they'd open source their FW,but that was dream. And thanks @weremy73 for the updates!

[efudd]

@chipless - 

9h40m16s   158.57   151.34

I give you props on bypassing the voltage limits you thought were in the PIC -- now we just gotta figure out what the voltage maximums are.

-j

[chipless]

@chipless -  

9h40m16s   158.57   151.34

I give you props on bypassing the voltage limits you thought were in the PIC -- now we just gotta figure out what the voltage maximums are.

-j

Anyone having throttle back down problems should try to set the fan manually to 75% or higher. I run mine at 90%, many problems with the miners is the voltage chip gets warmer and cuts out. The only way to keep it cooler is to turn up the fan. This scenario can happen without any errors being set in the log. Bitmain needed small heatsinks on the voltage regulator.

I guess I have to say thanks....I guess...Coming from you thats odd but.. you are also welcome for finding the spot so you can now volt/clock per asic. I would like to know the file format for the data in chips_freq.config, i was considering trying to enable it to read the volt or freq thru the file...I had it as high as 1200mV so far running in 75-90c range. I have all my miners in an air-conditioned 12x12 building with all the units vented out the side. So the heat may be a problem for some until freq and voltage are added back to the config file for some control. I am in the deep south where it has been in the high 80's and humid.

Anything below 870mV you lose speed

I am going to get it to 156k for free then after that paids users....I have a plan that will be fair and work for everyone's budget. The details will be laid out on the first post when it is all ready

[lenovop-70]

Wow, i found many nice information here, i have antminer z11 too in my home, ill try what peoples said above, i hope it will increase my miners speed as well, thanks for the OP for this great information, bless for you.
I will try it and post it here to share with you all.

[efudd]

@chipless -  

9h40m16s   158.57   151.34

I give you props on bypassing the voltage limits you thought were in the PIC -- now we just gotta figure out what the voltage maximums are.

-j

Anyone having throttle back down problems should try to set the fan manually to 75% or higher. I run mine at 90%, many problems with the miners is the voltage chip gets warmer and cuts out. The only way to keep it cooler is to turn up the fan. This scenario can happen without any errors being set in the log. Bitmain needed small heatsinks on the voltage regulator.

I guess I have to say thanks....I guess...Coming from you thats odd but.. you are also welcome for finding the spot so you can now volt/clock per asic. I would like to know the file format for the data in chips_freq.config, i was considering trying to enable it to read the volt or freq thru the file...I had it as high as 1200mV so far running in 75-90c range. I have all my miners in an air-conditioned 12x12 building with all the units vented out the side. So the heat may be a problem for some until freq and voltage are added back to the config file for some control. I am in the deep south where it has been in the high 80's and humid.

Anything below 870mV you lose speed

I am going to get it to 156k for free then after that paids users....I have a plan that will be fair and work for everyone's budget. The details will be laid out on the first post when it is all ready

Eh, the chips_freq.config is stale code. I futzed with it some.

I reverse engineered the voltage control communications to the PIC and ISL this evening and by passed the factory stuff in favor of my own initialization code for frequency and voltage on a per-ASIC (not just per hashboard) basis (I already have the frequency stuff done from my previous work).

I am still not happy with the predictability of results, unfortunately. While it is possible I am testing with a machine that just isn't up to par (which would change the equation if so), the more likely scenario is that holding a stable overclock across a broad range of systems is improbable.

Who knows...

-j

[chipless]

@chipless -  

9h40m16s   158.57   151.34

I give you props on bypassing the voltage limits you thought were in the PIC -- now we just gotta figure out what the voltage maximums are.

-j

Anyone having throttle back down problems should try to set the fan manually to 75% or higher. I run mine at 90%, many problems with the miners is the voltage chip gets warmer and cuts out. The only way to keep it cooler is to turn up the fan. This scenario can happen without any errors being set in the log. Bitmain needed small heatsinks on the voltage regulator.

I guess I have to say thanks....I guess...Coming from you thats odd but.. you are also welcome for finding the spot so you can now volt/clock per asic. I would like to know the file format for the data in chips_freq.config, i was considering trying to enable it to read the volt or freq thru the file...I had it as high as 1200mV so far running in 75-90c range. I have all my miners in an air-conditioned 12x12 building with all the units vented out the side. So the heat may be a problem for some until freq and voltage are added back to the config file for some control. I am in the deep south where it has been in the high 80's and humid.

Anything below 870mV you lose speed

I am going to get it to 156k for free then after that paids users....I have a plan that will be fair and work for everyone's budget. The details will be laid out on the first post when it is all ready

Eh, the chips_freq.config is stale code. I futzed with it some.

I reverse engineered the voltage control communications to the PIC and ISL this evening and by passed the factory stuff in favor of my own initialization code for frequency and voltage on a per-ASIC (not just per hashboard) basis (I already have the frequency stuff done from my previous work).

I am still not happy with the predictability of results, unfortunately. While it is possible I am testing with a machine that just isn't up to par (which would change the equation if so), the more likely scenario is that holding a stable overclock across a broad range of systems is improbable.

Who knows...

-j

Look into the voltage throttling the speed back down. It seams when I hit an area there is when it stopped throttling back to around 142 and instead started going up.It is either there or the scan freq that happening behind the scenes. That is why mine is in the 2 minute loop when it is not in the scan loop it will throttle back down but as long as it is in the 2 minute loop it holds steady. Weird...maybe resetting the 142 limit which is coming from somewhere I dont know of.

I cant explain none of it all I can really tell you is that it is a mess. If you are ever kind throw me a factory image with the voltage --bitmain-voltage and the --bitmain-freq re-enabled. I would like to try a couple things there but am busy doing too many other things right now to enable it.

[efudd]

@chipless -  

9h40m16s   158.57   151.34

I give you props on bypassing the voltage limits you thought were in the PIC -- now we just gotta figure out what the voltage maximums are.

-j

Anyone having throttle back down problems should try to set the fan manually to 75% or higher. I run mine at 90%, many problems with the miners is the voltage chip gets warmer and cuts out. The only way to keep it cooler is to turn up the fan. This scenario can happen without any errors being set in the log. Bitmain needed small heatsinks on the voltage regulator.

I guess I have to say thanks....I guess...Coming from you thats odd but.. you are also welcome for finding the spot so you can now volt/clock per asic. I would like to know the file format for the data in chips_freq.config, i was considering trying to enable it to read the volt or freq thru the file...I had it as high as 1200mV so far running in 75-90c range. I have all my miners in an air-conditioned 12x12 building with all the units vented out the side. So the heat may be a problem for some until freq and voltage are added back to the config file for some control. I am in the deep south where it has been in the high 80's and humid.

Anything below 870mV you lose speed

I am going to get it to 156k for free then after that paids users....I have a plan that will be fair and work for everyone's budget. The details will be laid out on the first post when it is all ready

Eh, the chips_freq.config is stale code. I futzed with it some.

I reverse engineered the voltage control communications to the PIC and ISL this evening and by passed the factory stuff in favor of my own initialization code for frequency and voltage on a per-ASIC (not just per hashboard) basis (I already have the frequency stuff done from my previous work).

I am still not happy with the predictability of results, unfortunately. While it is possible I am testing with a machine that just isn't up to par (which would change the equation if so), the more likely scenario is that holding a stable overclock across a broad range of systems is improbable.

Who knows...

-j

Look into the voltage throttling the speed back down. It seams when I hit an area there is when it stopped throttling back to around 142 and instead started going up.It is either there or the scan freq that happening behind the scenes. That is why mine is in the 2 minute loop when it is not in the scan loop it will throttle back down but as long as it is in the 2 minute loop it holds steady. Weird...maybe resetting the 142 limit which is coming from somewhere I dont know of.

I cant explain none of it all I can really tell you is that it is a mess. If you are ever kind throw me a factory image with the voltage --bitmain-voltage and the --bitmain-freq re-enabled. I would like to try a couple things there but am busy doing too many other things right now to enable it.

If I read this correctly, you said "scanning keeps things fast". The scan loop and the 'turbo' thread each just do some calculations and ultimately set a frequency on each asic, which takes a couple of clock cycles at most.

The scan takes a split second at most to run to completion for what it is worth.

-j

[chipless]

@chipless -  

9h40m16s   158.57   151.34

I give you props on bypassing the voltage limits you thought were in the PIC -- now we just gotta figure out what the voltage maximums are.

-j

Anyone having throttle back down problems should try to set the fan manually to 75% or higher. I run mine at 90%, many problems with the miners is the voltage chip gets warmer and cuts out. The only way to keep it cooler is to turn up the fan. This scenario can happen without any errors being set in the log. Bitmain needed small heatsinks on the voltage regulator.

I guess I have to say thanks....I guess...Coming from you thats odd but.. you are also welcome for finding the spot so you can now volt/clock per asic. I would like to know the file format for the data in chips_freq.config, i was considering trying to enable it to read the volt or freq thru the file...I had it as high as 1200mV so far running in 75-90c range. I have all my miners in an air-conditioned 12x12 building with all the units vented out the side. So the heat may be a problem for some until freq and voltage are added back to the config file for some control. I am in the deep south where it has been in the high 80's and humid.

Anything below 870mV you lose speed

I am going to get it to 156k for free then after that paids users....I have a plan that will be fair and work for everyone's budget. The details will be laid out on the first post when it is all ready

Eh, the chips_freq.config is stale code. I futzed with it some.

I reverse engineered the voltage control communications to the PIC and ISL this evening and by passed the factory stuff in favor of my own initialization code for frequency and voltage on a per-ASIC (not just per hashboard) basis (I already have the frequency stuff done from my previous work).

I am still not happy with the predictability of results, unfortunately. While it is possible I am testing with a machine that just isn't up to par (which would change the equation if so), the more likely scenario is that holding a stable overclock across a broad range of systems is improbable.

Who knows...

-j

Look into the voltage throttling the speed back down. It seams when I hit an area there is when it stopped throttling back to around 142 and instead started going up.It is either there or the scan freq that happening behind the scenes. That is why mine is in the 2 minute loop when it is not in the scan loop it will throttle back down but as long as it is in the 2 minute loop it holds steady. Weird...maybe resetting the 142 limit which is coming from somewhere I dont know of.

I cant explain none of it all I can really tell you is that it is a mess. If you are ever kind throw me a factory image with the voltage --bitmain-voltage and the --bitmain-freq re-enabled. I would like to try a couple things there but am busy doing too many other things right now to enable it.

If I read this correctly, you said "scanning keeps things fast". The scan loop and the 'turbo' thread each just do some calculations and ultimately set a frequency on each asic, which takes a couple of clock cycles at most.

The scan takes a split second at most to run to completion for what it is worth.

-j

Yes it was weird aparently the way I went into it the scan kept it higher then without. I am trying the full auto-tune currently have it running and  needs some tweaking for sure but have sustained good rates with it.

[ExEric3]

@chipless

Are you able make firmware with lower clocks? Is there hidden page minerAdvanced.cgi where you can choose turbo or balance mode but they doesnt do anything. Are you able add here for example clocks? 750MHz 700MHz 650MHz and 600MHz? Of course with volts mods also so miner can choose some value and finally I can replace fans.

Thanks.

[weremy73]

Beta 3 Update:

Luxor admin is telling me that my Z11 is submitting many duplicate shares, likely the cause of most of my rejected shares.

Elapsed           KSol/S(RT)   KSol/S(avg)
21h18m56s   145.63           146.41

Accepted shares: 7860
Shares rejected: 1738
Stale shares: 101

7860 / (7860+1738+101) = 81.01% share acceptance rate / efficiency

HW Errors: 14,586

Fan Speed: 100%

Chip Temps: 82c-86c
----------------------------------------------------------------------------------------------------------------------------------------
From the Luxor Discord:

NickH
Today at 2:14 PM

I see them on the server
that's the reject reason

2019/05/05 19:31:53 Error Message: {"id":9369,"version":"2.0","result":false,"error":[22,"Duplicate share from miner at address abcdefghijklmnopqrstuvwxyz",null]}
---------------------------------------------------------------------------------------------------------------------------------------

EDIT: I just swapped pools to see if I can replicate the issue.

[chipless]

@chipless

Are you able make firmware with lower clocks? Is there hidden page minerAdvanced.cgi where you can choose turbo or balance mode but they doesnt do anything. Are you able add here for example clocks? 750MHz 700MHz 650MHz and 600MHz? Of course with volts mods also so miner can choose some value and finally I can replace fans.

Thanks.

I am currently working on the autotuning which should help some. Eventually clocking can will be able to be done from the web gui like other miners. There will be more updates as time goes on. I currently have the clock set at 775 the next release should be a little higher until I get the rest together. This week updates may be slow I have a lot to do and may not get much time to tweak the firmware some more.

My biggest goal was to get something out there that the majority of users can use right away, now I can work on the tweaks.


As far as duplicate shares or rejected shares that may be a pool problem, I have a low reject rate from flypool but will add it to the list of things to work with. I have heard that even with the stock fw the reject rate was higher then other miners.

[cokewithanr]

Chip,

    I have been following both yours and Jason's threads and I am certain he will read it here as well. I am running several z9 mini's and saw a chip tear down post that shows the memory is built into the chip.. (My Dad said the only dumb question was the one not asked..so..here it goes.) Does the freq change you guys are working on also effect the memory that is built in to the asic? or is this something that is a given and I am just not catching it? ...on another note is there a way to unlock the controller for the 4th hash board, and a gui update to show the 4th board? I had several v9's that I flashed, made some mod's to the cases and am running the mini boards, but would like to combine them in a custom case with more than 3 boards to a controller.

If I overlooked something and this has been covered I am sorry, Just thought this would not be a bad place for my first post.

Thank you both for the hard work.

[chipless]

Chip,

    I have been following both yours and Jason's threads and I am certain he will read it here as well. I am running several z9 mini's and saw a chip tear down post that shows the memory is built into the chip.. (My Dad said the only dumb question was the one not asked..so..here it goes.) Does the freq change you guys are working on also effect the memory that is built in to the asic? or is this something that is a given and I am just not catching it? ...on another note is there a way to unlock the controller for the 4th hash board, and a gui update to show the 4th board? I had several v9's that I flashed, made some mod's to the cases and am running the mini boards, but would like to combine them in a custom case with more than 3 boards to a controller.

If I overlooked something and this has been covered I am sorry, Just thought this would not be a bad place for my first post.

Thank you both for the hard work.

The V9 or z9 mini should show 4th board if you use the web gui /www/pages/cgi-bin/status.html  from the Z9 full. the 4th board should show in the overall speed just not the web gui but it is working on the v9.....the z9 full web gui will show all the boards on a z9 or v9 with the 4th brd.

[cokewithanr]

Chipless,

    Thanks for the reply, I will try some of the thoughts shared from both threads as soon as I get a chance and see what I can cobble together.

My mind wanders, but eventually it comes back...
KC

[cokewithanr]

I was going to try your firmware last night, but it was later than I thought when I sent the PM. I will try it tonight if I can make time. I am not sure what happened, but right after I read your reply my ISP dropped out, had to reply from my phone. As with Jason I will also send you a donation today because it has been fun and informative following the threads. I am still learning and that is the important part.

Thanks and keep it up,
Keith

[chipless]

Version 1.1 is now released with some improvements. The speed was increased a little and the main image was updated to Bitmains 4/23/2019 version which corrected some issues, please leave me your feedback. The link on the first post has been updated. Please read all updated info on the first post there is important information regarding future releases.

[Roguepanther]

Version 1.1 is now released with some improvements. The speed was increased a little and the main image was updated to Bitmains 4/23/2019 version which corrected some issues, please leave me your feedback. The link on the first post has been updated. Please read all updated info on the first post there is important information regarding future releases.

Hey Chipless, are you only using one power supply for the boards? I just updated to your latest firmware and the Z11 shuts down momentarily and then starts back up but doesnt hash, just stays at 0.

cheers mate.