Asic.to Custom Firmware - Optimize Performance & Efficiency S19/S21/T21

[thierry4wd]

Hi all, I am currently running antminer S9 - 13.5Th with AsicBoost and voltage control, I am 15.500Th / s for 1420W on the wall (with firmware bitmain), after a lot of work, test, brick of my S9 .. ..

My setup is 725Mhz / 895mv ===> 15.5Ths / 1420W
My room is at about 20 degrees and the chips are at 90 degrees, for a fan speed of 2650trs and 3150trs

Ps : ha yes, i have unlock TEMP CHIP 1 on miner statuts, for now say the pcb temp, but i share for all, but no people interessed lol

[sergw]

Well, I have tried with all of them and changing frequencies but nothing improve, always same problem with high temps. Answering to your question, yes, I tried flashing 650 fixed frequency firm first and from there just swapping bmminer file for your file. For example, I have tried with bmminer 880 and frecuency 650 onwards, like 656, 662, etc till 762. With all of them I get same problem. Thanks for your help !

i have 2 such s9 miners, chip temp jumps to 90-100C, but board temp stay intact, so i assume that old s9 have some difference with chip temp reading.

Hi all, I am currently running antminer S9 - 13.5Th with AsicBoost and voltage control, I am 15.500Th / s for 1420W on the wall (with firmware bitmain), after a lot of work, test, brick of my S9 .. ..

My setup is 725Mhz / 895mv ===> 15.5Ths / 1420W
My room is at about 20 degrees and the chips are at 90 degrees, for a fan speed of 2650trs and 3150trs

Ps : ha yes, i have unlock TEMP CHIP 1 on miner statuts, for now say the pcb temp, but i share for all, but no people interessed lol

what firmware with asic boost do you use?  I tested brainsOS with AB which allows freq and volt tuning, but i cannot achieve stable hashrate more than 14 Th with enabled LPW mode.

[thierry4wd]

i use my own codes

i working for it about 50 hours ....

[taserz]

seem like fan setting problem in bmminer.
with my S9 13.5 other firmware's plus bmminers and tricks all over the internet the custom fan speed not work my miner always follow auto fan. But custom fan speed only work with LPM firmware and the bmminer inside LPM.
my controller board is c41 v1.20

I will double check the compiled version. Possibility fan speed change not allowed but pretty sure I changed that. I know it is set so you can unplug the fans and submerge it too

[mercuryin]

After several says trying different firmwares I can confirm the following;

With original bitmain LPM firmware + original bmminer, just changing the frequency inside advance panel and modifying through css the fan speed control to 100%, right now I have my two antminers s9, one 13.0TH s9 running to 675 = 14.237TH and 73C and the other s9 14.00TH running 750 = 15.931TH and 80C and almost no hw errors and 1600w psu. So I´m not experiencing anymore the heat problems.

[Andartis]

After several says trying different firmwares I can confirm the following;

With original bitmain LPM firmware + original bmminer, just changing the frequency inside advance panel and modifying through css the fan speed control to 100%, right now I have my two antminers s9, one 13.0TH s9 running to 675 = 14.237TH and 73C and the other s9 14.00TH running 750 = 15.931TH and 80C and almost no hw errors and 1600w psu. So I´m not experiencing anymore the heat problems.

you mean the fixed frequency firmware, right? Or can wen now adjust frequency in the autotune firmware?

[mercuryin]

Yes, 650 fixed frequency firmware

[jtw1984]

I think the reason my firmware and braiins is heating up a bit more is because many of the newer releases from bitmain have not been made public so figuring out what they did to lower temps is difficult since the source code we have to work with is a few revisions old. I am trying to figure out the LPM but it is currently not being very friendly with me.

With the new S15 I might just try to work on that one if I can get my hands on 1 even though the tax is stupid high.

Came here to ask a similar question to this.

I'm showing Bitmain's last update being May 5, 2017.

How long do you believe it will take for them to release this new update?

[dutchlincoln]

After several says trying different firmwares I can confirm the following;

With original bitmain LPM firmware + original bmminer, just changing the frequency inside advance panel and modifying through css the fan speed control to 100%, right now I have my two antminers s9, one 13.0TH s9 running to 675 = 14.237TH and 73C and the other s9 14.00TH running 750 = 15.931TH and 80C and almost no hw errors and 1600w psu. So I´m not experiencing anymore the heat problems.

strange...
I have about 15 with 650FF firmware with the LPM fw on top, and highest i can clock them is about 668Mhz. Any higher and hashrate starts to drop.
Also i have 4 that are situated in a colder area, and they are so picky they wont even take the LPM mod; too cold, hashrate drops and they run stuck. Reboot and they go for some hours again. New flash with 650 (or autotune only) and they're fine...
So, temp. has a big influence also... Chip temp of 40 degree centigrade they don't like at all...

[DickyDck]

strange...
I have about 15 with 650FF firmware with the LPM fw on top, and highest i can clock them is about 668Mhz. Any higher and hashrate starts to drop.
Also i have 4 that are situated in a colder area, and they are so picky they wont even take the LPM mod; too cold, hashrate drops and they run stuck. Reboot and they go for some hours again. New flash with 650 (or autotune only) and they're fine...
So, temp. has a big influence also... Chip temp of 40 degree centigrade they don't like at all...

I agree, when too cold mine tend to be moody too, which makes no sense to me since heat is the enemy or so I thought. I think I could potentially get better results if I had the beefier PSU, I am running the 1600w one from bitmain on a dedicated 30amp 240v circuit, so I have plenty of power from the wall, but I think the PSU might be my limiting factor.

[AniceInovation]

i use my own codes

i working for it about 50 hours ....

Will you share it?

[taserz]

Will you share it?

Whenever someone brags with something with a and doesn't show any proof I call bs. Granted I can docker anything.

[Andartis]

So, interestigly on stock settings with autotune, my test miners hash at 14.1TH and draw 1180-1200W at the wall. Thats nice. But thats after Nicehash changed something to their servers so that they actually support asicboost (some time ago).

Any progress on the modded version?

[Bill48105]

[...]

Code:

set_Voltage_S9_plus_plus_BM1387_54 failed on Chain[1]!

set_Voltage_S9_plus_plus_BM1387_54 failed on Chain[1]!

set_Voltage_S9_plus_plus_BM1387_54 failed on Chain[1]!

Chain[J3] will use voltage=930 [1] to open core
set_Voltage_S9_plus_plus_BM1387_54 failed on Chain[2]!

set_Voltage_S9_plus_plus_BM1387_54 failed on Chain[2]!

set_Voltage_S9_plus_plus_BM1387_54 failed on Chain[2]!

i restore original t9+ firmware but what realy happened to my miner

what you think? i try to revert back to original firmware but fails

Anyone ever get their T9+ with this issue fixed? I've been searching without any luck..  My T9+ was working fine on the latest official firmware from bitmain's site but I made the mistake of downloading the source code which I saw was a year old but figured not a big deal since it specifically supports T9+.  I set miner type for T9+ like build instructions say (it defines T9_18 in the header file), CHANGED NOTHING ELSE, built it, uploaded to my T9+, killed the running stock bmminer, & ran my built copy & bam now I got these same "failed on chain"  errors.  I never installed the Dollemizer or taserz firmwares btw.

I had happened to be running the new build in screen session so I could watch what was happening & it was flipping out with those failed on Chain errors then "chain[1] PIC need restore ..." 3x each for each chain in succession but bitmain's dead miner detection kept kicking in mind reflash of the PIC's causing reboots so the process would start all over again.  The problem boards seem to show "Check chain[2] PIC fw version=0xff" vs the working boards seem to show "Check chain[3] PIC fw version=0x90" (ff vs 90). Logs show other things like

Code:

T9+ try get asic num for the 2 time Chain
[J2] has 0 asic
set_reset_allhashboard = 0x00000000

T9+ try get asic num for the 6 time
Chain[J2] has 0 asic
Chain[J4] has no freq in PIC, set default freq=200M
Chain[J4] has no core num in PIC
Chain[J13] has no freq in PIC, set default freq=200M
Chain[J13] has no core num in PIC
Chain[J14] has no freq in PIC, set default freq=200M
Chain[J14] has no core num in PIC

Pretty sure that's not a good sign.. lol  Btw I have them underclocked at 200mhz in conf file.

In between reboots (and hoping the newer factory original bmminer would repair the boards), I renamed the factory bmminer back & upon the next auto reboot the new one ran up but unfortunately as others have found some of my chains are dead/missing still.  Only chains 4 13 & 14 are working so my hash rate is awful. Whopping 1213 GH/s at 200mhz lol

So far I've tried factory reset & flashing the older autotune T9+ stock from bitmain with no luck fixing it. I ended up modding the source code I had downloaded to skip the PIC writes to see if the build at least worked & it does!  Other than Temp(Chip 1) which is normally just - (dash), now shows a negative value like -23 etc which is odd.. Temp(Chip2) shows what appears to be about right for temp. Since I defined T9_18 per the docs it's as if some parts of the code are ignoring that & building for the wrong model or something.. That would explain the PIC errors & corruption as well if that were the case.  

Unfortunately I can't find newer release of the source code to try building that and I can't find the older pre-autotune firmwares to try either so I'm kinda stuck.  I found this thread with multiple people with this error & hoped to find a definitive fix before I tried anything else. I read thru all the posts & didn't see any solutions so any insight on fixing this issue is appreciated.

Thanks,
Bill

[thierry4wd]

Whenever someone brags with something with a and doesn't show any proof I call bs. Granted I can docker anything.

No problem I understand;) I certify that it's true, free to everyone to believe me or not, it's not only track that out to the public with dev fee, because miner is no longer profitable for me, so I try to survive in this world as I can;) there is absolutely nothing nasty, although I share some of my know how like eco mode for the T9 + for example;) but perssonne does not seem interested lol, friendly;)

[...]

Hi! actually in the source code there is an error that modifies the "firmware" of the PIC32, this peak precisely manages the tenssion of the chains ... even if you followed the program, this error was inevitable, I was also trapped ! just like dollemizer and the others;)

to my knowledge, only one solution (for the moment), I explain everything here: https://bitcointalk.org/index.php?topic=5032987.0

[Bill48105]

[...]

Hi! actually in the source code there is an error that modifies the "firmware" of the PIC32, this peak precisely manages the tenssion of the chains ... even if you followed the program, this error was inevitable, I was also trapped ! just like dollemizer and the others;)

to my knowledge, only one solution (for the moment), I explain everything here: https://bitcointalk.org/index.php?topic=5032987.0

Wow yeah that surely seems like the same issue.. Thanks for the heads up!  I'll read thru that thread but at 1st glace it seems pretty straight forward but stinks can't flash it with software..  I'll reply to your thread instead of there though. Thanks!

[thierry4wd]

with pleasure ! trust me I knew these minor very well (to repair and modify them) and the problem comes from the, without any doubt;)

[Bill48105]

I am going to take a look at this today.  I wrote / modified a program to help restore the pic firmware on an L3+ ( https://github.com/darval/bitmain-tools ) and I am going to see if I can modify it to work for a T9+.  The pic is a microcontroller on each hash board.  Usually when it (bmminer) tries to get the version and it fails (-1 or 0xff), it means the pic firmware got trashed.  My program has a mode to try to restore it.  It currently doesn't work on the T9+, so don't try it yet, but if I am successful, I will update it and post about it here.  Then you can try it out.  Like anything, this stuff is all use at your own risk.

Hi darval, did you ever get something working with your tool & the T9+? few posts back you can see link to where thierry4wd posted a how-to with PICKIT 3 & MPLAB X IPE along with a HEX dump from working card using that tool so hoped to be able to flash my broken card with software.. Save me ordering & waiting for a pickit.  I'm tempted to convert his hex file to the format of the txt file but I wasn't sure if the format of the text file caused the corruption to begin with.  (I doubt the antminer monitoring script auto-rebooting mid flash helped.. x(  I mean single-board-test & bmminer obviously can flash the boards to store settings & messed up the boards so would think it should be possible to flash them back correctly.  I've monkeyed around with the source of both of those & get the basic idea of what's going on (I disabled the auto-flashing in both so I could get my miner up & running on chains 4 13 & 14 at least.  The latest official firmware just auto-reboots over & over when it has heartbeat failures so made my T9+ worthless.)  So maybe I'll order a PICKIT 3 and while I wait try hacking away at the source to see if I can get anywhere in the meantime but hoped someone had a working software-only solution already
Thanks

[taserz]

Hi darval, did you ever get something working with your tool & the T9+? few posts back you can see link to where thierry4wd posted a how-to with PICKIT 3 & MPLAB X IPE along with a HEX dump from working card using that tool so hoped to be able to flash my broken card with software.. Save me ordering & waiting for a pickit.  I'm tempted to convert his hex file to the format of the txt file but I wasn't sure if the format of the text file caused the corruption to begin with.  (I doubt the antminer monitoring script auto-rebooting mid flash helped.. x(  I mean single-board-test & bmminer obviously can flash the boards to store settings & messed up the boards so would think it should be possible to flash them back correctly.  I've monkeyed around with the source of both of those & get the basic idea of what's going on (I disabled the auto-flashing in both so I could get my miner up & running on chains 4 13 & 14 at least.  The latest official firmware just auto-reboots over & over when it has heartbeat failures so made my T9+ worthless.)  So maybe I'll order a PICKIT 3 and while I wait try hacking away at the source to see if I can get anywhere in the meantime but hoped someone had a working software-only solution already
Thanks

As of right now I think that is the current solution.

[Bill48105]

As of right now I think that is the current solution.

Ok thanks. It'd be cool to get a hex file made that could be used to reflash cards that have this happenbut I ordered a pickit3 since bitmain support was hung up on warranty status of my unit rather than understanding what happened or providing much help but we'll see. Supposedly my issue was being escalated although not holding my breath. lol  It seems the new "low power" T9+ firmware is built to auto-reboot if it doesn't find heartbeat of all 3 cards though which is very annoying when trying to test single boards..  I ended up tweaking the old available source a bit & running that with 1 working card for now. It boots up & mines WAYYY faster than the official version even when the cards were working. I do need to measure power usage at wall to see how much I'm losing out on vs the new low power version but that one won't run without all 3 working cards installed so can't test that until/unless I get my other 2 cards repaired.