Topic: S17 Pro no longer hashing. Bad ASICs ? (Read 1178 times)

Yes, the Asic Repair tester can burn the eeprom if you buy that option, but I don't think you can tell it what settings to write. So it just writes some default settings. If the hashboard doesn't work well at those settings then the stock firmware will work, but not very well. I haven't asked if there is a way to change the frequency and voltage setting that it writes, haven't needed it since the miners I've been repairing are pretty much all using vnish already.

Promising to hear there is some real success mixing boards under Vnish. I think the Asic Repair Center test jig also has the ability to work with the EEPROM...or just EEPROM flashing.

I have mixed boards a few times with S17s running vnish and have not had any issues.

To run stock firmware, I believe both the Pic program for all the boards need to be identical and the tuning data stored in the eeprom on the hashboard needs to be compatible. The pickit does not program the eeprom, that is a chip separate from the pic. I believe the tool from Zeus writes the eeprom, not the pic. It lets you read the eeprom settings from one board, and write them to another.

My understanding is that the different pic program revisions require different communications api, and at least in the stock firmware, it only supports one api at a time. That same limitation might also apply to vnish.

Along with what mikeywith has described in this thread I spoke to a knowledgeable person in mining on TG they say if you run Vnish, interchangeability between boards (50/53/56) shouldn't be an issue. I have my doubts, so I'd recommend buying a PicKit just in case you do find a pro board.

I've been seeing multiple sellers that advertise using a mixed collection boards for their s17s.  I was under the impression that this isn't possible.  I guess it is.  Any idea where i can find some boards to try and bring my system completely back?  Its currently running with 1 board at 17TH-20TH.

True, the recent batches are of better quality; also the ones that ship from Malaysia are worse according to what most people reported, however, the overall quality of all batches is combined is terrible, I bought those gears throughout the time of their release, as early as the second batch of the S17 (most of them for my clients), and nobody is really happy with any of them except for the s17 pro version which seems to be an expectation, not all great but a lot better than the rest.

On the other hand, we have some Whatsminers from the same bitmain 17 series era that my clients reported 0 failure rate on, some are over 2 years old, imagine for 2 years, sounds like S9 all over again.

With that being said, if you have a low failure rate, you should be thankful really as this is the unusual of how they are, you guys have decent luck I envy you.


Couldn't agree more, especially when you see Avalon and Whatsminer's owners be like "What does repair mean" , I know everyone has a different view on the subject, given what I think is a large enough sample size I find it very risky to buy any of the 17 series especially when other options are around, so far people are reporting good results with their S19s and T19s, some minor issues with their PSUs but nothing all that serious, my plan is to migrate to Whatsminer, selling my Bitmain gears for premium during this bull cycle and rebuying cheap M20s and M30s when things cool off little sounds like a great plan for the following years.

I wish I could keep all the S9s I have because nothing comes close to their stability, but these gears are getting too old, I just realized some of them are as old as 4 years.

Same here, running 24 x S17e and 12 x T17+ purchased around May / June 2020 and not had one single problem with them. I do have some older 2019 T17+ (that I bought used) and those are now dying though... I also recently purchased 20 x S17 Pro from a Chinese supplier who insists they were working 100% however 4 had bad hashboards on arrival, and now I think there are 12 dead hashboards in total :\

Also running 35 x S19 95t for around 7 months now and no problems at all.

I think it was certain batches, or maybe miners made at certain facilities. I have 7 S17 pros from Malaysia, and 2 out of 7 failed in less than a year. But I also have 10 S17s and 3 S17 pros from China, and I haven't had a single one go down.

Even if you do have the tools and skills to repair them, many times there are just too many issues to make it worth the effort. Many of these miners have flaky connections everywhere.. so you fix one issue and it runs for a few days until another issue springs up. Not worth the repair when it becomes a full time job to keep a single miner going...

I have been saying this for months or perhaps years on this forum, I kept mentioning how bad the 17 series mining gears are on every occasion I saw possible, I did that to the extreme that someone accused me of "fud", I can't blame them, it did sound like I own a Bitmain competitor company.

Phill is probably the only one I know who didn't have a high failure rate, the vast majority of people I know had well over 30% failure rate, some are above 50% in less than a year, this is so unfortunate given that most of us bought those gears when BTC price was at or below 10k.

It's a lesson learned, I hope more people become aware of that, if you don't have the tools and skills needed to fix these gears DO NOT buy them.

Agilmore, x17 could be considered the worst in terms of quality and depend on a lot of luck. Would you entertain selling your working hashboard? Wndsnb mentioned he'd also be interested in the dead hashboards, PSU, case, and control board left over. Appreciate the consideration.

They basically told me that these were the worse units that bitmain produced and has had a ton of issues.  The solder or whatever compound thats used to attach the chips to the board basically melted away.

It's more than likely going to do the job just fine, but the price is more than double, the Pickit 4 has more features than the 3 version which you probably will never use, I'd say stick to Pickit3.

The Pickit seems like the smartest choice here for a small scale operation and rare use situations. Also, thierry4wd's post provides a good reference to get things off the ground along with information in this thread. Would you recommend the newer Pickit4?

Extracting the pic firmware from a hashboard with the pickit it pretty easy. I've done it and can verify it works. No need to pay extra for the zeus tool if you have the time to spend 20 minutes to do an internet search and play with the pickit software.

The repair manuals on the zeus website describes how to program the pic with the pickit, all you need to do is find the option to read the firmware in the pickit software.

Well you are going to need some special device for that, you are going to write on a microchip somewhere on those hash boards, the easiest way would be using zuesbtc tool > https://www.zeusbtc.com/ASIC-Miner-Repair/Parts-Tools-Details.asp?ID=236, it's plug and play and easy to use, however, it's kind of expensive.

The other alternative which I believe will achieve the same results would be using pickit 3, thierry4wd explained the use of that tool in great detail, and despite the fact that his topic talks only about fixing T9+ in an unrelated subject to the code/id difference, I believe by extracting the hex from one board and writing it on the other one will make the latter have the same product id and thus will be able to function on the same control board.

Keep in mind while the second option is A LOT cheaper but it has yet to be tested, the first option has been tested and works perfectly, but it's a matter of 250$++ difference.

You mentioned that if you plan to mix and match boards they will need some work and seems like one would have to undergo some editing so that the control board recognizes the correct product id's. I'm interested to know if this work can be performed without specialty equipment.

Sometimes it's no brain and sometimes it's no keyboard, depends on the day

A tutorial on what exactly? if we are talking about mix and match of hash boards I believe what I posted earlier on is enough of a "tutorial", if you are talking about something else, please explain.

By the way, do you really not have a keyboard?

wdsnb, I agree, they probably have a massive backlog as well. With 6 actual technicians servicing the US I shouldn't be surprised it's taken since mid December to get this diagnosis and repair. Totally time-consuming I can just imagine the gaps you have to fill in but you know they work and your not under the corporate-type time crunch. Would you entertain repairs and perhaps complete overhauls that improve the factory build quality to boards? I tried to contact agilmore but he's not accepting messages from newbs. Perhaps you would have better luck.  

This is great information and clears up a lot of confusion mikeywith. I think this would be helpful seeing as replacement parts for the 17 series are becoming rarer and working hash boards are few and far between now. It's what I was afraid of and far beyond my skill set. Maybe a tutorial soon?  

NKBTW

I have reasons to believe that this isn't the case, just to stay on topic I rather not mention them here, but I can tell you one thing, it's probably the product ID difference, The 17 pro version has like 3 different product id's, not sure what the reference but it seems like it does reference a certain format in the hex stored on some microchip on each of those hash boards. So if you mix two boards with different product id's the control board will refuse to turn on both of them, you could check the kernel log and you will find the product ID represented in something like so


Or/And hexadecimal values on the PIC fw version:


If the hash boards have any different values, then they need some work done before they can mine using the same control board.

I guess it isn't very surprising given the rates they are charging. On average they probably can't spend more than 2 or 3 hours on a hashboard to break even.

I think the Bitmain training is probably pretty specific, focusing on the most likely issues. If the problem doesn't fall into the scope of their training, there isn't really any information out there on it. I've been going off the Bitmain repair manuals and then filling in the missing information by reverse-engineering the boards. Pretty time consuming ...