Topic: T17/S17 malfunction: cases, solutions, remedies, RMA history - page 17. (Read 6919 times)

Well after pushing and pushing at every known e-mail address at Bitmain I finally got in contact with somebody on a high rank at service dept. He confirmed what I have been suspecting about the quality of soldering is very likely the root of the problems on all S17 series, even on prior from S series as S9K and all the way thru S17. I got in touch with him via whatsapp here is the conversation: https://ibb.co/LdPzMyc

So, being that said, all my dear friends is useless and futile send you S17 to Bitmain even under warranty because is very likely that as soon as you receive them back other chips will come off, so learn how to solder using soldering paste and flux with hot air solder. Other than that, find yourself a local electronic tech near you to avoid and expose the miner to shipping "kick handling" even if shipped to Bitmain's Service Center in the USA.

Hope this clears out all theories about this equipment since the root cause of failure have been revealed.

Just for the record, I paid $263.00 for those crappy S9ks, they weren't at $60 like now, so really nearly $300 for that gear doesn't give bitmain the excuse to send them with fallen chips and heatsinks, in fact, the QS measure should be same on all gears, cheap gears cost less to manufacture, which is why they can afford to sell them for so cheap, but QS is a different story, around the same period of time I bought S9js for $276, and they are a trillion times better, those S9js were probably stored from previous years batches, and given that all the new gears S9 or not have the same issue of chips/heatsinks falling, it shows that something went wrong in their QS department, something which all the S19s buyers should keep in mind, although Jihan Wu promised that the S19s will be robust after he admitted the 17 series were crappy, he, of course, puts the blame on Micree Zhan.

Ironically I tend to believe Jihan this time, not because I think he is honest by any means, but the S19s are their last chance, so it's very likely that everything about S19 is going to be a lot better in terms of quality, I am just not willing to take the risk until it's confirmed.

Well... Considering their price, QC is probably not worth it for Bitmain to do on S9ks. It has been noticable over years that their QC is becoming worse (or even worse, it could be that it hasn't changed at all even though technology changed a lot and that can be a big problem too).

I believe it has little to do with the outside temp when the miner is not running, also bad handling of packages doesn't seem to be the reason why those chips on my S9ks fell, I had a lot of them shipped, some for me and a lot for clients, they all arrived at the same time with the same environmental conditions, a few had heatsinks fall right out of the box, some did mine for a few days, some for a few weeks, some are still good to this point but overall terrible, so my guess is that alongside with the terrible quality of their solder paste, those S9ks don't go through a quality process, so depending heavily on your luck, you might receive a bricked gear as I did.

Yeah. the box is well packed. the foam is good quality.  but the force generated with a 6 foot drop is quite a bit.

and if you drop the box flat on the bottom there will be no apparent damage to the box or the miner inside.

But every heat sink will be subject to a g-force from the drop then stop.

to make it more wrong is that if the gear is in a really cold spot ie during wintertime the  gear is allowed to travel in 0c or 32f or lower transport bays.

So now the glue/heatsink past is hard stiff and brittle so now drop it 6 feet and the box has no damage the  unit looks fine .  I could guarantee that multiple glue/paste/epoxy heatsink joints now have micro damage .  Same as solder joints.

Would love to demo that test.  As you know it happens.

---

edit To be clear 0c or 32f makes the heatsink to chip joint a bit more brittle and easier to crack then it would be at 20c or 68f

And of course  at -20c or -4f it would be a bit stiffer and even more brittle.

So if the fedex/dhl/ups truck drops off a unit on your porch in the winter time it is very likely the chip to heatsink seal if cracked was cracked right at your door step.

Here in NJ I have received hundreds of pieces of gear from China and sometimes the guys drop the gear off when I am not home.

Even if the guy drops it 3-4 feet to the step in the dead of winter it is not good for the heatsinks.

I will take a look at your S9 topic, if chips fall off during transport, that means that the reflow was to low in temp or the belt speed was to fast (if the temp profile in the oven was ok). POOR QA!

The best way to repair the faulty board (and produce the board) would be to use a vapor phase instead of the reflow. But that is to pricy for them and to slow. The heatsink is glued,  the S1 was a proper way to do a heatsink, but again, to pricy for BM.

This makes sense, but what I don't seem to understand is the reason why all gears made by Bitmain in 2019 onwards have the same issue, it's not just the 17 series, even the S9k had the same issue of heatsinks/chips falling off the hash boards, I am not sure if you saw my thread regarding the s9k, now I don't know enough about electrics but it seems like bitmain aren't using any type of solder but rather they inherit the postage stamp method of using a bit of saliva to stick them  , I would understand if the heatsinks fall due to over-heating them - that means the quality of the paste they use is bad, but for chips to fall off due to shipping and handling is terrible, I would like to get a an explanation on that.

Problem is that all the boards look ok, it would be like finding a needle in a haystack finding the bad chips. The log shows 0 chips detected on all 3 boards, so i have no idea witch ones need to be re-soldered.

I also tried blasting the boards with hot air gun, no results.

Not sure if reflow oven at higher temperature will work (already tried 10 min @ 200 deg c). Really have nothing to lose, i'm just experimenting at this point.

Well, this is not valid for all electronics.

Bitmain asics comes as qfn package not bga, that means that the coating of the chip is paladium. Everything in chip manufacture is oriented to lead-free now.
Depending on what kind of solder paste (activator chemicals) was used, a mix with other lead paste (acid chemicals as activator) may result in a good joint, but degrading of the influence of the different chemicals mixed now over the time. The heat of the board, the vibration from the fans, and a not homogen temp over the pcb result in warping the joints (microfracture).

The problem of the miner is to look overall, so the pcb design and the material used is also an issue. Given the condition of in wich the pcb has to operate, this is not an easy (quick) task. But time to market is the main concern at bitmain. Anyway we have to live with it, as we got only the option not to buy.

PS: If bitmain can offer you a new S9 for 63$ now and still make some money, I do not think that they are realy designing the miners up to quality or a long life.

All you do is a reflow, make sense if the board is new, if the equipment is old a reballing is needed.

The lead-free solders melt at higher temperatures of about 217°C/422°F compared to 183°C/361°F for the lead-based option, if you can use lead paste/ball is better.

The problem with Lead-free solders is that make micro-fracture occurs in the solder joints, this is a know problem with the consoles (XBOX and PS), reballing with lead balls make then working perfectly, no more cold joints and micro-fracture.

Lead free solder is an abomination for circuits, help the environment but kill circuits especially the ones that work with high temperature.

Just to let you know, the greyish finish comes basicly from lead free solder. All Manufactors are forced to use leadfree from 1996 on for consumer products. You may test it using a handheld analyzer (in production used to check material prior production).

That you expirience cold joints is a old problem also on S5 and S3. To repair that, I use no clean flux paste around the affected chip then a heatplate on the underside of the board (regulatet to 100° c) and warm the pcb up for arround 3 min. before I am using a regulated hotair (360°c) on top to melt the joint and reestablish connection. 99% this method works. 1% the chip was realy dead before that action.

In conclusion I can say that bitmain did not got their reflow process in control, and no qa off corse.

Sorry to hear your story, although by now I am not surprised to read such stories, in fact, I'd be surprised hearing the opposite about these crappy gears Bitmain makes.

Did the support bots at bitmain give you any troubleshooting aside from the

"Sir please reset the mine and flash the latest firmware" or "Sir this is the recovery files put on sd and flash, use this guide"?

I am sure that members here provide a million times better troubleshooting than bitmain's support, did they tell you about the cheap soldering they use? I can bet the house that the 2 boards that failed had a loose heatsink/chip on them, just like 99% of other issues happen, so another month or so of delay, only to get the same unit you ordered a long time ago, by the time you start hashing - all those S19s/M30s have been plugged in, well good luck with that.

Recently purchased 2 x S17+ 70TH, April batch but delivered in May due to delays at Customs. Both machines worked fine initially, then dropped a hashboard. One machine after just 1 hour and the 2nd unit after approx. 2.5 hours, very disappointing Spent several days back and forth with Bitmain support doing everything they asked but to no avail and now both units have been shipped back for repair.

favebook brought up a very good point, just for everybody's information, the heatsinks that usually fall are the once that are directly on top of the chips ( the bigger ones), if you take out 2 heatsinks placed in the opposite direction of one another, one will have the asic chip below it, another will be "empty", however, without totally negating the effect of positioning the miner in a horizontal position, it's safe to assume that unless you can do it easily [1] it's really not worth the effort, there isn't much you can do about these heatsinks until they fall, sadly.

[1]- if you use spilt-room design as I do and you have already made a whole to fit the miner vertically, going horizontal is a lot of effort, which IMO not worth it given the little advantages you get from this, but if it's a matter of simply repositioning the miner in 30-60 seconds, go ahead, you lose nothing.



Had to go from f to c (apparently what 99% of human beings use  ) and it turns out to be 196 c, please DON"T do that, the chips and most other components will most likely rest in peace when they are put in that temp for 10 mins, your method could have probably fixed the soldering problem but created a far greater issue.

 I suggest to simply apply some pressure on the heatsinks see which one falls, try to soldier it or do as The_binary_Observer suggested.

Jbillk, don't throw them to the garbage, find someone near you that have soldering skills using hot air guns, buy yourself a good soldering paste and bring one hashing board and ask the tech resold the chips and most of the components on the hashing board.

here is a top notch soldering paste

https://www.digikey.com/products/en/soldering-desoldering-rework-products/flux-flux-remover/266

That seems smart. Will test too!

I understand what you are talking about, you are referring to the chips that are on the other side of the hashing boards, the small ones with small heatsink. However, what I am doing is putting upwards the chips with the long heatsink which are the ones that creates most of the heat and thus are prone to fall off the board due to this high temperature. The ones on the opposite side have a small heatsink which tells me they develop less heat. I am testing this position on several miners and see what happens.

All S17 T17 series need a change of FANS, those that come are the worst wining shitty fans i haver use, all bitmain equipment this days come with those fans, they move almost nothing air, if you have DELTA or other good brand (2.6 amp) just change the 2 back FANS.

I have a S17 with 3 dead hashboards that gradually dropped out. No warranty, no good-will. For sure its the solder breaking on the chips, tried everything.

This is not recommended, but as a last ditch effort tried to re-flow the hash boards by heating to 385 Deg f for 10 min. Haven't had success. I've heard of people bringing motherboards and GPUs back to life using this technique. Theory is the heat melts and reconnects the broken solder joint.

Maybe something to try before you throw your board in the trash. Not sure if anybody else has tried this with any degree of success.

Are you sure about the horizontal positioning? By some laws of physics and my basic understanding of it, while miner is in vertical position, all heatsinks have the same sag and gravity is pulling them downwards at angle of 90°. And that may not be perfect, but I think it is best to stay in that position since if you put the miner horizontally, half of heatsinks are in perfect position (gravity is pulling them towards chips and that would make them almost impossible to fall off), but half of them are in worst position as they are pulled downwards (away from chips) at angle of 180° and that would make them (by some raw and stupid calculation) 50% more likely to fall off than in vertical position.

Please, do correct me if I am wrong in any way!


Yeah, I understand you... I hate those answers too.