T17/S17 malfunction: cases, solutions, remedies, RMA history

[mikeywith]

I wouldn't say the fixture is lying, it is doing its job, which can't be perfect all the time, I suppose the solder balls you found around the chip came from "overdosing" by bitmain staff/robots, after being put under heat that extra solder has to go somewhere when it goes outside of where it has to be it will short the hash board, it probably does something to the component that reads the temp-sensor and that could explain why both the fixture tool and the miner kernel log mention the temp-sensor when sensors are actually good, and since temp-sensors are essentials the miner won't start thinking that all 4 sensors are bad, these are some terrible quality gears.

I am glad you managed to fix some of the hash boards, it would have been a lot better if you could take some photos, also share how did you manage to remove that solder without damaging anything else.

[1714051907]

1714051907

[CryptoLLC]

I will share all the info soon. I have a youtube channel. I will post a link soon.

[mikxxy]

An AntminerT17+ been running perfect for some months suddenly and after some power cuts, did not restart. Flashed it with Awesom FW and getting this in kernel log, is all boards dead or just the hashboard? Thanks Mike.

Code:

[2020/11/25 14:09:25] INFO: Power ON
[2020/11/25 14:09:27] INFO: Starting FPGA queue
[2020/11/25 14:09:27] INFO: Initializing hash boards
[2020/11/25 14:09:27] INFO: chain[2] - Initializing
[2020/11/25 14:09:27] INFO: chain[1] - Initializing
[2020/11/25 14:09:27] INFO: chain[0] - Initializing
[2020/11/25 14:09:43] WARN: chain[2] - 44 of 65 chips detected
[2020/11/25 14:09:46] WARN: chain[1] - 44 of 65 chips detected
[2020/11/25 14:09:49] WARN: chain[0] - 44 of 65 chips detected
[2020/11/25 14:09:54] WARN: chain[2] - 44 of 65 chips detected
[2020/11/25 14:09:57] WARN: chain[1] - 44 of 65 chips detected
[2020/11/25 14:10:00] WARN: chain[0] - 44 of 65 chips detected
[2020/11/25 14:10:06] WARN: chain[2] - 44 of 65 chips detected
[2020/11/25 14:10:06] ERROR: driver-btm-chain.c:488 chain[2] - Failed to detect ASIC chips
[2020/11/25 14:10:06] INFO: chain[2] - Shutting down the chain
[2020/11/25 14:10:06] ERROR: driver-btm-base.c:356 chain[2] - Initialization failed
[2020/11/25 14:10:09] WARN: chain[1] - 44 of 65 chips detected
[2020/11/25 14:10:09] ERROR: driver-btm-chain.c:488 chain[1] - Failed to detect ASIC chips
[2020/11/25 14:10:09] INFO: chain[1] - Shutting down the chain
[2020/11/25 14:10:09] ERROR: driver-btm-base.c:356 chain[1] - Initialization failed
[2020/11/25 14:10:12] WARN: chain[0] - 44 of 65 chips detected
[2020/11/25 14:10:12] ERROR: driver-btm-chain.c:488 chain[0] - Failed to detect ASIC chips
[2020/11/25 14:10:12] INFO: chain[0] - Shutting down the chain
[2020/11/25 14:10:12] ERROR: driver-btm-base.c:356 chain[0] - Initialization failed
[2020/11/25 14:10:12] ERROR: driver-btm-base.c:2154 Failed to initialize hash boards
[2020/11/25 14:10:12] INFO: Shutting down the miner
[2020/11/25 14:10:12] INFO: Stopping FPGA queue
[2020/11/25 14:10:12] INFO: chain[0] - Shutting down the chain
[2020/11/25 14:10:12] INFO: chain[1] - Shutting down the chain
[2020/11/25 14:10:12] INFO: chain[2] - Shutting down the chain
[2020/11/25 14:10:12] INFO: Power OFF

[favebook]

This is a new one, I've never seen any unit drop all hashboards at exact same chip (44).

Have you tried custom firmware or newest firmware from Bitmain? Could you try changing PSU? Could you try disconnecting one of the boards and running 1 or 2 hashboards only?

Please use [code/code] for logs.

[mikeywith]

Please use the code function to post kernel logs, it's really annoying and hurts the eyes. , use something like this

Code:

[2020/11/25 14:09:43] WARN: chain[2] - 44 of 65 chips detected
[2020/11/25 14:09:46] WARN: chain[1] - 44 of 65 chips detected
[2020/11/25 14:09:49] WARN: chain[0] - 44 of 65 chips detected

What the kernel log suggests is that all 3 hashboards have only 44 asics, which means the signal at chip 45 gets interrupted, it is unlikely that all 3 hash boards will have the exact same problem, so my guess is your PSU got a problem, can you test the miner with another working PSU?

[mikxxy]

Sorry about the annoyance with code format but im still a newbie , seems as a  plausible explanation, as the the power cut only involed one of the two powercords.  icould try with a psu from a T17 APW9 would that be sufficient for the  test of  the faulty T17+ APW9+? Upgraded to latest  Bitmain orginal FW, and now flashed the awesom SD 2.03.

[mikeywith]

Can you confirm that all 3 fans on the PSU spin just fine?

icould try with a psu from a T17 40 Th would that be sufficient for the  test of  the faulty T17+ 58th?

NO, don't do that, the T17+ uses APW9+ while the T17 uses APW9 EMC, I am not the right person to ask when it comes to electricity but my guess is that the APW9+ could damage your T17.

Upgraded to latest  Bitmain orginal FW, and now flashed the awesom SD 2.03.

Are you getting different results?

[mikxxy]

Jup, all 3 spinn when powerd by both cords. Regarding PSU its way around, APW9 instead of the plus version, for testing. the miner stopped rigth after i powered back one of the cords due to the powercut, red led came upp instead of the green in end of power cycle, so that seems indicate a total sudden loss of function, rather then a progressive problem?. Seems to be a new design of the top cover, no 4 screws holding top down, this one have screws on the sides and something looking as a "button" on backside, how do i unmount the topcover on this miner?

[CryptoLLC]

Watch this video on solder balls, 0 asic or when hash board only detects 44 chips or 22 chips and so on.
https://www.youtube.com/watch?v=mqd-AEpcJTw

[mikeywith]

Seems to be a new design of the top cover, no 4 screws holding top down, this one have screws on the sides and something looking as a "button" on backside, how do i unmount the topcover on this miner?

I don't have a miner with such a cover, but I think this video will be helpful > https://www.youtube.com/watch?v=czBf5E2lYDk&t=1s

Watch this video on solder balls, 0 asic or when hash board only detects 44 chips or 22 chips and so on.

That is a great video, I did hit the like on it, I have a few things to add

1- When there is a short-circuit caused by anything whether it's excessive solder or a heatsink that dropped and shorting other chips, 99% of the time you will get 0 asic and not 44, 22, and the other weird combinations.

2- You talked about buying the black glue from China, it's indeed a good product and extremely cheap, but that product has to be stored in the fridge, it can't be at room temperature for more than 10 days, and with the pandemic and the slow worldwide shipping it could take more than 10 days so make sure you check with the courier or get another alternative from the local store.

3- The problems you talked about are very common in all 17 series with no exception, so it's probably worth mentioning.

Also, I would change the video title to " Chain[]: find 0 asic Antminer" as this will most likely be the most common keyword people would use.

[mikxxy]

Thanks It was indeed a button to press for takning the top cover off, never seen this one. I dissassembled the miner and all the hashboards look fine no lose heatsinks. im leaning towards psu, as it happend simutaniusly.I conclude that  theres power to the boards and fans, but its not powering up hashboards. Should be a separate power circuit for the boards. I have another, a T17 that lost one (occasionly one more), but controler detects the faulty board, some data as voltage is shown, and also a asic count before powering it off
ill disassemble the psu to check if theres any visual damage, dont know anything about testpoints or so, would appreciate any suggestions for where to search.

[mikeywith]

Should be a separate power circuit for the boards.

There are indeed two different circuits and this is so confusing to most people, they see the LEDs turn on and they think it's a confirmation that the PSU is in perfect shape which sadly isn't really the case.

I have another, a T17 that lost one (occasionly one more), but controler detects the faulty board, some data as voltage is shown

If I understood you correctly, you have another working T17 PSU (APW9) that is powering a single hash board, and sometimes two, and we know you "most likely" have a dead PSU and 3 working hash boards, wouldn't it be wise to take the working PSU and instal it on the "potentially" working 3 hash boards?

ill disassemble the psu to check if theres any visual damage, dont know anything about testpoints or so, would appreciate any suggestions for where to search.

We milked this topic quite a bit, but long story short these PSUs were not really meant to be fixed, it's not that it's impossible to fix them, but once you open it you will get the point, to me, anything except for a simple fan change on those PSU isn't working looking into, and I doubt you will find many resources regarding the subject, so if you are not an expert in this field, chances are you should be looking for a PSU replacement.

[wndsnb]

So you put the thermal solder on the chip? Do I need to put anything on the heat sink such as flux or something else?

This video shows a method of cleaning up and applying solder to attach the heat sinks: https://youtu.be/F6L9hCQ4OhY

A bit of tacky flux can help to get the solder to flow and adhere to the chip surface. Flux plus solder wick is used to clean up excess solder. This is the solder paste, flux, and solder wick I use and have had good results with:

https://www.chipquik.com/store/product_info.php?products_id=310001
https://www.chipquik.com/store/product_info.php?products_id=470006
https://www.chipquik.com/store/product_info.php?products_id=640003

[mikxxy]

Now i changed the PSU took it from a working T17+. Same problem persists . Inotice this message in Awesom miner

Code:

[2020/11/27 07:56:00] INFO: Auto-tuning
[2020/11/27 07:55:58] INFO: eFUSE protection is active
[2020/11/27 07:54:49] INFO: Auto-tuning
[2020/11/27 07:54:47] INFO: eFUSE protection is active
[2020/11/27 07:53:24] INFO: Auto-tuning
[2020/11/27 07:53:17] INFO: eFUSE protection is active

JES, problem identified! The culprit is the control board, looks normal, there seems to be some problem in or out coming signals to the net,it does not identify itself on a IP scanner. changed inbetween the healthy one an d the bad one, so no problem with the hashboards. Awesom was inndicating 44 of 66, wich is wrong as there is only 44 chips on each board. Bitmain has produced several models of the T17+ this particular one has 58Th, so the Awesom FW got confused. Regardind the control board it may just be something with RJ45 connector, but i think its easier to just get a replacement board?

I open up another one, a T17 it had actually a heatsink off! there is some videos in Russian language, about fixing this, but i semm not to understand if the chip is replaced or resolderd? i should be burnt shortly after heatsink falling off, while in mining mode? Isnt there a way to just "short" the bad chip out and thus restoring integrety of the chain?

[mikeywith]

JES, problem identified! The culprit is the control board......but i think its easier to just get a replacement board?

That is good news, it's very unlikely for a control board to toast, which is also why it would make sense if you found some very cheap control board, bitmain store is out of stock, so you need to get it from elsewhere, are you located in the U.S?

I open up another one, a T17 it had actually a heatsink off! there is some videos in Russian language, about fixing this, but i semm not to understand if the chip is replaced or resolderd? i should be burnt shortly after heatsink falling off, while in mining mode?

If the chip is bad then it needs replacement, but most of the time if the chip is toasted the heatsink won't fall, if the heatsink is off that hashboard is likely going to stop mining before the chip is burned, so usually, a heatsink that is off just needs resoldering.

Isnt there a way to just "short" the bad chip out and thus restoring integrety of the chain?

None that I am aware off, but let's see what the other guys have to say although I doubt such a fix is possible.

[mikxxy]

i'm located within the EU, whats confusing as said is the T17+ and none with different amount of chips on each board, 44 and 64?, should they need a designatet FW depending on asic count on each hash bord? Awesom issued two different FW's 1.00 and 2.0.3? Ill try the first one on the faulty card, may be it will run?

Yes rigth, the T17 with a poped heatsink present it self at initial phase of power cycling, 000000 and then it wanish from the Gui, so i migth be alive, lets see what will happen as i glue i back then.

[mikeywith]

whats confusing as said is the T17+ and none with different amount of chips on each board, 44 and 64?, should they need a designatet FW depending on asic count on each hash bord? Awesom issued two different FW's 1.00 and 2.0.3? Ill try the first one on the faulty card, may be it will run?

That is strange, I don't think the firmware knows the number of chips i.e it isn't hard coded into it, it probably gets the information from the hash board itself, the reasoning behind my point is that using the same control board and the same firmware you can run different hashboards, for an example if you take S17 hash boards and put them in your S17 pro they will work just fine despite having different asic count.

But with that being said you should reach out to Patrike and explain the issue to him.

[mikxxy]

Yes im with you,. when i try run the bad controler in the second maschine it behaves exactly as in the previous setting, by not beeing able to detect correct asic valuse but in addition, now also giving error of not detecting any fans, even though the are spinning accordingly to pre setting of 30% (awesom FW), could it be affected by any virus in any of processors? A leds onboard are behaving as expected, boot sequenceas as it should, then it tries to initialize the hash bords iindefinitely. So what could by wrong with this controller?

Regarding the poped heat sink can it just be glued on and then reinforced by glueing the top of the heatsink to those in vicinity?

By joggling the controller back and forth, i started to catch on Awesom, detecting asic chains but the first one, also it warns if try to raise freq some steps above 500Mhz

Code:

N: chain[0] - Failed to reset pic (attempt = 2), resp: 0x90 0x90
[2020/11/27 23:45:34] WARN: chain[0] - Failed to reset pic (attempt = 3), resp: 0x90 0x90
[2020/11/27 23:45:35] WARN: chain[0] - Failed to start pic app (attempt = 1), resp: 0x01 0x01
[2020/11/27 23:45:36] WARN: chain[0] - Failed to start pic app (attempt = 2), resp: 0x01 0x01
[2020/11/27 23:45:36] WARN: chain[0] - Failed to start pic app (attempt = 3), resp: 0x01 0x01
[2020/11/27 23:45:38] WARN: chain[0] - Failed to reset pic (attempt = 1), resp: 0x01 0x01
[2020/11/27 23:45:39] WARN: chain[0] - Failed to reset pic (attempt = 2), resp: 0x01 0x01
[2020/11/27 23:45:41] INFO: chain[2] - 44 chips detected
[2020/11/27 23:45:41] WARN: chain[0] - Failed to reset pic (attempt = 3), resp: 0x01 0x01
[2020/11/27 23:45:41] WARN: chain[0] - Failed to start pic app (attempt = 1), resp: 0x01 0x01
[2020/11/27 23:45:42] WARN: chain[0] - Failed to start pic app (attempt = 2), resp: 0x01 0x01
[2020/11/27 23:45:42] WARN: chain[0] - Failed to start pic app (attempt = 3), resp: 0x01 0x01
[2020/11/27 23:45:43] INFO: chain[1] - 44 chips detected
[2020/11/27 23:45:44] WARN: chain[0] - Failed to reset pic (attempt = 1), resp: 0x01 0x01
[2020/11/27 23:45:46] WARN: chain[0] - Failed to reset pic (attempt = 2), resp: 0x01 0x01
[2020/11/27 23:45:47] WARN: chain[0] - Failed to reset pic (attempt = 3), resp: 0x01 0x01
[2020/11/27 23:45:48] WARN: chain[0] - Failed to start pic app (attempt = 1), resp: 0x01 0x01

Code:

[2020/11/27 23:39:24] INFO: Initializing
[2020/11/27 23:39:22] INFO: eFUSE protection is active
[2020/11/27 23:39:00] INFO: Initializing
[2020/11/27 23:38:58] INFO: eFUSE protection is active
[2020/11/27 23:38:34] INFO: Initializing
[2020/11/27 23:38:34] INFO: Pool connection FAILED. Check INTERNET, POOL URL or WORKER
[2020/11/27 23:38:32] INFO: eFUSE protection is active
[2020/11/27 23:38:00] INFO: Mining
[2020/11/27 23:36:38] INFO: Initializing
[2020/11/27 23:36:36] INFO: eFUSE protection is active
[2020/11/27 23:36:30] INFO: Initializing
[2020/11/27 23:36:28] INFO: eFUSE protection is active
[2020/11/27 23:35:59] INFO: Initializing
[2020/11/27 23:35:59] INFO: Pool connection FAILED. Check INTERNET,

[mikeywith]

So what could by wrong with this controller?

When a control board is bad, you would expect all kind of weird problems, and honestly, things now are becoming confusing and hard to follow  , can you just not test the bad control board so we all can focus on the good one?

i'm located within the EU...

That is unfortunate, I don't know of any trusted suppliers there, but should you need one I think you should contact one of these guys (in no particular order)

- favebook
- thierry4wd
- o_solo_miner

There might be others, but those are off the top of my hand for a list of EU based miners that can be trusted.

Regarding the poped heat sink can it just be glued on and then reinforced by glueing the top of the heatsink to those in vicinity?

More explanation is needed, I don't seem to understand why exactly are you planning on doing.

[2020/11/27 23:38:58] INFO: eFUSE protection is active

This line is very strange, I did text Marc (Asic.to) on telegram, let's see what he knows, it seems like a power-related issue, are you sure the PSU you using is APW9+ and the input voltage is 200-240v?

Also, it shows this

Pool connection FAILED. Check INTERNET, POOL URL or WORKER

Are you sure your pool settings are in place? is Asicboost is activated make sure the pool is AB compatible if this log comes from the bad control board and all this info can be misleading.

[mikxxy]

Alrigth i see what mean.. i hava tendency to go down into the Rabbit holes when something from my point of view ocours, yes and things may get out of contex and thus causing confusion. So im sorry for any incoinvience caused.

i have no outher means to test of the bad card by any other means than trail and error. with the apprieciated help from you guys. All the equipment mentioned, has been running att smal miningpark for at least 6 mounths , and all the settings are also the origional against Poolin. So all the odd behavior described herein emmenate from known and reliable mining conditions. The Antminer T17+ is out of stock and those few contrilers that find their way to Ebay, are very expensive. //Mixy----------------------------------------------------------