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

[mikeywith]

2021-01-02 16:33:31 driver-btm-api.c:942:check_asic_number: Chain 1 only find 40 asic, will power off hash board 1

This is the problem, Chain 1 shows only 40 asic when it's supposed to signal 48 chips, so there seem to be an issue with the 40th or 41th chip, usually, with these 17 series mining gears it's a loose heatsink that needs soldering.

With that being said, where is your chain 0 and chain 2, does this miner run with only 1 hashboard?

[1714972489]

1714972489

[1714972489]

1714972489

[spyroslo]

Thank you for you reply. Do you believe that it is a hardware problem or software? The miner has install free hashboard.

[favebook]

This is the problem, Chain 1 shows only 40 asic when it's supposed to signal 48 chips, so there seem to be an issue with the 40th or 41th chip, usually, with these 17 series mining gears it's a loose heatsink that needs soldering.

With that being said, where is your chain 0 and chain 2, does this miner run with only 1 hashboard?

It does have three chains, it just skips them in beginning. Loose heatsinks (they fell off almost certainly) are almost 100% problem on those 2, while chain 1 is just loose heatsinks but still haven't fallen of fully.

Thank you for you reply. Do you believe that it is a hardware problem or software? The miner has install free hashboard.

Hardware 100%, check other posts here, you will find very similar or identical problems.

[mikeywith]

I am 99% positive that it is a hardware issue, however, i would still test the miner with custom firmware since sometimes it could fix the issue, try vnish (asic.to or awesomeminer), chances are low but you have nothing to lose.

[RRZP028]

I bought a tester from Zeusbtc two months ago, I can tell you that there is still no successful result regarding the repair of my miners, first the tester is of very poor quality, it can be seen that they looked to save on everything, asik 0 on almost all boards and cooperation with them grade 2

[wndsnb]

I bought a tester from Zeusbtc two months ago, I can tell you that there is still no successful result regarding the repair of my miners, first the tester is of very poor quality, it can be seen that they looked to save on everything, asik 0 on almost all boards and cooperation with them grade 2

Yeah, the tester is only a small part of being able to repair hashboards. The output of the test is only sometimes useful, most of the time you need the fixture just to operate the board while you use an oscilloscope and/or voltmeter to measure signals at the test points.

Just FYI, with the COVID situation, Bitmain now offers an online repair course for $500. I haven't taken it so I'm not sure how useful it is, but once you complete it you get access to all of their official repair manuals and downloads, and you can order Bitmain's test fixture and spare parts direct from them. Previously it was $1000 and you had to travel to one of their sites to take the 1-week long course.

All of the testers of that type are built similarly ... I agree that it is not the best. Just kind of hacked together. I've ordered a different tester from here https://tester.asic.repair/en. It was much more expensive and ships from Russia but looks like a much better product. Slow shipping from Russia to USA though ... ordered a month ago and the tracking estimated arrival is still 3 to 5 more weeks...

[RRZP028]

did you manage to fix a board, do you have problems with the Asic 0 error that occurs a lot with me and I can't solve it

[wndsnb]

Yes, so far I've repaired 3 S17 hashboards and one S17+ hashboard. Most were reporting 0 Asics at some point. I'd say more issues cause 0 asics than some number fewer than the total asics.

To solve it, you need to trace the signals through the board to try to locate the spot where they go bad. The repair manuals on Zuesbtc are a good place to start. It's not easy though, both the locating of issues and then the techniques to remove/replace heatskins and chips. I'm an Electrical Engineer with >25 years experience and it took me a month of nights and weekends to decode the manuals/reverse engineer the hashboards to get to the point where I could repair one. I'm sure it would have been a lot easier to just take the Bitmain repair course, but I guess I'm just a glutton for punishment...

[RRZP028]

Well, I'm a beginner in that, today I successfully replaced my first chip, I replaced it after a few attempts, chip No. 01 after that it gave an error on chip 34, I still haven't been able to solve it, can you tell me how to glue a refrigerator for chip and what temperature do you use then and what when should you glue the chip to the hash board.

When it shows Asic 0, I first control the first chip, and then the last. I look at which test point the parameters are not good and after that I start with the control of the chip, first I take off the cooler and check if there are soldering balls then I try to reattach the chip at a temperature of 400 degrees Celsius, if that doesn't help I take off the chip and first try to I put it back again because I clean it nicely, if it doesn't help you either, then I mount a new chip, after that I measure the resistance, if everything is OK, then I go to test and check the values at the checkpoints again.


Has anyone completed the BItmain course that wants to help a little.

[wndsnb]

I use a heatgun set at 315 degrees C set at low airflow to remove/attach heatsinks and chips.

Attaching heatsinks on S17 boards is pretty tricky. For S17 and S17 pro, I've been removing most residual solder from the heatsink and adding what I think is the correct amount of solder to the top of the ASIC. I put some tacky flux on the top of the chip, then heat the heatsink with my heatgun for ~20 seconds and set it down on the chip. There is then a small window of time where you can adjust the placement of the heatsink. If you are not quick enough, the solder holding the chip to the board can melt and then you can shift the chip.

On the S17+, that technique didn't work for chips with the smaller heatsinks. So on those, I prepare the chip and heatsink the same way, but then just place the heatsink on the chip and heat it directly until the solder flows.

For attaching chips, I remove the solder from the board with solder wick, apply tacky flux to the board and place the chip in position, then apply heat for ~45 seconds. For new chips, I use a stencil to apply solder paste to the chip and melt the solder before placing it on the board. Nearly every time I do this, all connections are not made, so I add solder paste with a small syringe to the edge of the chip, reflow, add more flux, reflow again, and clean up any excess solder with an exacto knife while the solder is still liquid. Pretty tedious process.

Flux I use: https://www.chipquik.com/store/product_info.php?products_id=310003
solder paste I use: https://www.chipquik.com/store/product_info.php?products_id=470006
solder wick: https://www.chipquik.com/store/product_info.php?products_id=640002

Heat gun I use: https://www.masterappliance.com/proheat-1500-lcd-programmable-professional-heat-gun-kit/


Started working on another S17+, this one is a real mess. One hashboard had 9 heatsinks fall off, some with chips still attached.

The heatsinks that fell off look like they came off with the film from the top of the chip. Without the film on the chip, it is impossible to get solder to flow evenly on the top of the chip, so impossible to get a good contact with the heatsink. I saw the same issue on some chips on another S17+. Looks like a manufacturing issue with the chip packages. Or maybe it just breaks down if it gets too hot for too long.

To save the chips, I'd have to use a thermally conductive adhesive instead of solder to attach the heatsinks. Not too excited about that idea. A lot of the chips got hot enough to melt the solder holding them to the board and slid down towards the bottom of the miner. Pretty sure some/most of them would be toast anyway so might just replace them all.

Pretty sure this board is going in the "maybe later" pile.



These chips would normally be covered in solder, or would be an amber color if a heatsink was never attached.



Most of the film from the chip is still stuck to the heatsink.

[mikeywith]

Started working on another S17+, this one is a real mess. One hashboard had 9 heatsinks fall off, some with chips still attached.

I had the same issue with my S17+, these gears are terrible compared to the S17 pro, they are less terrible than T17/T17e but still, the failure rate on my S17+ is 50% easily, and most of the heatsink fell along with the chips unlike the T17s, I also noticed this on the S17, I wish I had the skills, time and tools to fix these chips like you and the other folks, I had some success with the S9s chips but my overall performance was very poor, I had someone else fix them for me after that, and that person is no longer available, so all I can do is watch those gears sit there do nothing.

[wndsnb]

I had someone else fix them for me after that, and that person is no longer available, so all I can do is watch those gears sit there do nothing.

If you want to sell any of your dead gear instead of watching it do nothing, shoot me a PM. Shipping might be too expensive to make it worth it, but maybe not.

Not sure how commercial places are making money repairing these. Hmtech charges $40 diagnosis + $150/hashboard for "basic" repair, $300/hashboard for "advanced" repair. I'm sure their techs are way more skilled than me, but I wonder if they would repair the board from my last post for $300 or if they'd just say it's not repairable. It's possible it could need $100 worth of replacement chips.

So far, every board I've fixed has taken more than 8 hours of work. The S17+ that I repaired took ~24 hours of work, although that was the first S17+ for me so there was some learning curve on that one.

I also found on the S17+ that chips will come up with the heatsink when trying to just remove the heatsink. So just pulling heatsinks to get access to repair other chips can cause other issues.

[favebook]

I have 2 units that are sometimes working, sometimes not. S17 Pro and S17+. One has dead hashboard (exactly one loose heatsink) while all 5 other hashboards sometimes work. Where are you from and would you try to repair them or want to buy 'em out?

[wndsnb]

I have 2 units that are sometimes working, sometimes not. S17 Pro and S17+. One has dead hashboard (exactly one loose heatsink) while all 5 other hashboards sometimes work. Where are you from and would you try to repair them or want to buy 'em out?

I'm in the northeast USA.

I'd consider buying them, but if they are partially working it may be better for you to limp along with them while profitability is as high as it is now. Guess that depends on how often the 5 sometimes working hashboards go down.

[RRZP028]

Every day I spend up to 10 hours on the same hash board and new problems are constantly born, problem Chain 36 after a while an error appears chine 21 then chine 6 then Asic 0 and so constantly, a lot of trouble for a beginner, and those from Zeusbtc will not really cooperate,

Every chip I want to glue, I zoom in ten times to see if everything is in place and again I can't set it up nicely, and I'm also interested in whether you test the desoldering plate right while it's warm or it has to cool down really well.

I ordered all the equipment from Zusbtc https://www.zeusbtc.com/.

Can the solder be removed from the chip with a soldering iron, how safe is it for the chip, and is it wise to clean all the solder from the board before installing a new chip?

[wndsnb]

I use a soldering iron set at 315 deg C to remove solder from chips. As long as you don't let the chip get too hot for too long, it won't get damaged. The key is time + temperature. Most components and the PCB have materials that will start to break down above 150 to 200 degC. Both have specs for how long it can safely be above a certain temperature so they can be reflow soldered in a reflow oven.

As for removing solder from the board before installing a new chip, I do that because I find it very difficult to place the chip when there is solder on both the board and the chip. Since the solder on the board and the chip create rounded bumps, the bumps on the chip nearly always will fall off the bumps on the board.

[mikeywith]

If you want to sell any of your dead gear instead of watching it do nothing, shoot me a PM. Shipping might be too expensive to make it worth it, but maybe not.

Shipping from where I live to the U.S will sure cost a ton of money, the same thing to China, it's why I throw away most of the gears that die on me, but depending on your offer we might work it out especially if you only need the hash boards and not whole miners.

I'll confirm what do I have left untouched and send you a PM accordingly, I know I have hash boards of these models (S17 pro, S17, T17, T17e, S17+) I just need to confirm the numbers.

[mtnrotary]

First two miners are finding 0 chips on one hashboard at startup. Not something that is generally fixable without getting the hashboard on a test fixture and tracking down the issue. As mikeywith said, your best option is probably to try to just run with 2 hashboards each. If you can find a repair shop reasonably close you could consider sending them to get repaired, but by the time you pay shipping both ways and repair of 2 hashboards it would probably be > $500. Sending them back to Bitmain is not a good option in my opinion. I RMAed a S17pro under warranty to Bitmain's repair facility in California in May and still don't have it back.

As for the 3rd one, it doesn't see any chips on bootup from any hashboard. Not likely that all 3 hashboards decided to die at the same time. So, like mikeywith suggested, you should see if your host will swap a PSU to see if it runs with a good one. If that doesn't do it, you might see if they'll swap the control board out.

The miners are located in Eastern Canada. Do you know of any repair sites in that part of the world?

[RRZP028]

today I ordered another 50 chips, although I haven't even learned how to install them yet, I'm from Serbia, so sending from here is difficult, both to China and the USA,
I ask about the temperature because these with Zeusbtc advise 450 degrees Celsius when it is removed and when the chip is glued, so I assumed that it might not be much and whether that temperature will damage the chip
and how do I know that I haven't burned the chip from overheating and I don't know if I should put pressure on it during the gluing process,
for soldering I use two solder pastes 138celsius and 183 celsius, which is your opinion which is better

[favebook]

I'm in the northeast USA.

I'd consider buying them, but if they are partially working it may be better for you to limp along with them while profitability is as high as it is now. Guess that depends on how often the 5 sometimes working hashboards go down.

Let's say that two out of five hashboards work every 10th try if miner is preheated and other three hashboards work every 3rd try if miner is preheated. If they are not preheated, there is maybe like 5% that 2 of them will work rest will not work at all if atmosphere temperature is under 25C.

Sadly as RRZP028 noted, shipping from Serbia to China or USA is way too expensive for anything that is heavy.

I am from Serbia as well which means I might be in need of your recently acquired skills RRZP028.