S5 board gone bad, options

[soy]

My S5 lost a board last night.  Chain 1 showed many "x"'s.  I changed supplies, no help, in fact chain 1 went completely dashes.  I swapped sides.  Chain 2 went all "x"'s.  I lowered the frequency to 200M.  Ten "0"'s and 20 "x"'s on the bad board!  I cleaned the board.  Now all "x"'s again.  Must admit the board had gotten dusty and the heatsink fins had accumulated dust at the intake.  My last try today was the newest firmware.  So, I believe the board has gone bad.  This is a Batch 1 S5 I bought in March, well beyond the 3 month warranty.  What kind of charges have others seen from Bitmain for this kind of board return and repair?  Thanks for insight.

[VirosaGITS]

My S5 lost a board last night.  Chain 1 showed many "x"'s.  I changed supplies, no help, in fact chain 1 went completely dashes.  I swapped sides.  Chain 2 went all "x"'s.  I lowered the frequency to 200M.  Ten "0"'s and 20 "x"'s on the bad board!  I cleaned the board.  Now all "x"'s again.  Must admit the board had gotten dusty and the heatsink fins had accumulated dust at the intake.  My last try today was the newest firmware.  So, I believe the board has gone bad.  This is a Batch 1 S5 I bought in March, well beyond the 3 month warranty.  What kind of charges have others seen from Bitmain for this kind of board return and repair?  Thanks for insight.

I had no luck getting replacement from Bitmain.

You could try messing around with, only running it with the controller. Check if a cap is dead, etc. Maybe you can post a pic of the board or check if there's anything burnt. If not you could sell the board, if you're in North America i could be interested.

I dont think dust would had killed it.

[unholycactus]

I doubt replacements from Bitmain for a S5 (considering you have to pay outbound shipping) will be profitable. Have you tried contacting Bitmain's support to confirm?

Another option would be to buy another hashing board from someone closer to you to replace the broken one or sell the working one.

[soy]

My experience with buying other than from Bitmain directly has been disappointing for the most part.  Most used Bitcoin miners of any efficiency sold by individuals are problem devices.  Profits are low and getting a buggy miner is continually frustrating.

[soy]

My S5 lost a board last night.  Chain 1 showed many "x"'s.  I changed supplies, no help, in fact chain 1 went completely dashes.  I swapped sides.  Chain 2 went all "x"'s.  I lowered the frequency to 200M.  Ten "0"'s and 20 "x"'s on the bad board!  I cleaned the board.  Now all "x"'s again.  Must admit the board had gotten dusty and the heatsink fins had accumulated dust at the intake.  My last try today was the newest firmware.  So, I believe the board has gone bad.  This is a Batch 1 S5 I bought in March, well beyond the 3 month warranty.  What kind of charges have others seen from Bitmain for this kind of board return and repair?  Thanks for insight.

I had no luck getting replacement from Bitmain.

You could try messing around with, only running it with the controller. Check if a cap is dead, etc. Maybe you can post a pic of the board or check if there's anything burnt. If not you could sell the board, if you're in North America i could be interested.

I dont think dust would had killed it.

The U19/U20 pair seem shorted to ground.  There's a 4.99 ohm 1% resistor across the caps mounted between each ASIC pair but across that cap C12 there appears to be a short.  A close inspection shows no debris on the ASIC pins.  

This board has a T09 round sticker in the upper right.  It's heatsink screws with springs were torqued less snug than the other board.  On the inside, the heat sink side, perhaps as is true on all the boards (will think to look one day), the positive rail comes thru to the inside to that ASIC pair in clusters of thru holes.  One of those thru holes is not coated for contact purposes when troubleshooting.

[soy]

I dont think dust would had killed it.

Dust impeded cooling - hotter ASIC became inefficient - inefficient ASIC drew more current - internal ASIC component or conductor couldn't stand up to the higher current, melts and shorts to ground.

[VirosaGITS]

I dont think dust would had killed it.

Dust impeded cooling - hotter ASIC became inefficient - inefficient ASIC drew more current - internal ASIC component or conductor couldn't stand up to the higher current, melts and shorts to ground.

Which would all mean something if the ASIC was allowed to run to 90c and 100c, which it cannot because the miner will stop the mining if it reach 80c. And a 2% in power draw isint enough to make a difference.

[philipma1957]

I dont think dust would had killed it.

Dust impeded cooling - hotter ASIC became inefficient - inefficient ASIC drew more current - internal ASIC component or conductor couldn't stand up to the higher current, melts and shorts to ground.

Which would all mean something if the ASIC was allowed to run to 90c and 100c, which it cannot because the miner will stop the mining if it reach 80c. And a 2% in power draw isint enough to make a difference.

Except the sensor for temp is dead center on each board so if the hottest asic is on a top corner of the board the sensor may not read that it is at 100c

[VirosaGITS]

I dont think dust would had killed it.

Dust impeded cooling - hotter ASIC became inefficient - inefficient ASIC drew more current - internal ASIC component or conductor couldn't stand up to the higher current, melts and shorts to ground.

Which would all mean something if the ASIC was allowed to run to 90c and 100c, which it cannot because the miner will stop the mining if it reach 80c. And a 2% in power draw isint enough to make a difference.

Except the sensor for temp is dead center on each board so if the hottest asic is on a top corner of the board the sensor may not read that it is at 100c

It is taken into account and sure the chips are indeed hotter at the end, but the chips are from from breaking down at that temperature.
We do put these in the oven after all. It would take a long time of overheating for the chips to get damaged and i don't think thats the issue here, even if OP forgot his S5 in a sauna for a month and did not notice the temp alarms during all that time.

We did see some S5 with a certain firmware overheat and break while losing internet connection however.

[soy]

Is there another good reason that the resistance across the diodes is measurable above 4 ohms except in the quadrant of ASICs 19 & 20 where the resistance measured is well less than 1 ohm?  There are a relatively small number of filter capacitors to ground, but my money's on one of the two ASICs.  The large diode is good.  Other diodes are good.

-----------

Sorry, long day working on car, 4.99 ohm resistors in parallel with capacitors but my meter reads resistance of between 4 and 5 ohms except for 19/20.

[RichBC]

Is there another good reason that the resistance across the diodes is measurable above 4 ohms except in the quadrant of ASICs 19 & 20 where the resistance measured is well less than 1 ohm?  There are a relatively small number of filter capacitors to ground, but my money's on one of the two ASICs.  The large diode is good.  Other diodes are good.

-----------

Sorry, long day working on car, 4.99 ohm resistors in parallel with capacitors but my meter reads resistance of between 4 and 5 ohms except for 19/20.

I usually see between 4.6 - 4.7 Ohms (Important to check the meter lead resistance first touching the leads of the meter together at these low values) Anything much less than this indicates a short in one or both of the ASIC's. Anything much higher can only indicate that the 4.99 Ohm resistor has gone high value for some reason?

Rich

[soy]

Is there another good reason that the resistance across the diodes is measurable above 4 ohms except in the quadrant of ASICs 19 & 20 where the resistance measured is well less than 1 ohm?  There are a relatively small number of filter capacitors to ground, but my money's on one of the two ASICs.  The large diode is good.  Other diodes are good.

-----------

Sorry, long day working on car, 4.99 ohm resistors in parallel with capacitors but my meter reads resistance of between 4 and 5 ohms except for 19/20.

I usually see between 4.6 - 4.7 Ohms (Important to check the meter lead resistance first touching the leads of the meter together at these low values) Anything much less than this indicates a short in one or both of the ASIC's. Anything much higher can only indicate that the 4.99 Ohm resistor has gone high value for some reason?

Rich

Thanks.  That's what I thought.  Any way to get the board running without those 2 ASICs?  Removing one or both of the ASICs should clear the short but I hear they run in a chained operation.  I see the R39/D92 - R45D98 could be jumped, say removing D92&R45.

[RichBC]

I have thought about but never tried removing a pair of chips & jumpering the data, it is far from straightforward.

First the chips are difficult to remove you need a decent rework station. Then there are 4 or 5 data signals that will need to be jumpered across the chip(s). Then the power either needs to be jumpered or a resistive load put in it's place equivalent to the effective resistance of the chips at the frequency the chips are hashing at, if this is not done and the power just shorted you would need to reduce the supply by 0.8V, and remove one set of the level shifting diodes.

Finally there is almost certainly something else I have not thought of? In theory it's possible, in practice probably not to be recommended.

Rich

[VirosaGITS]

I have thought about but never tried removing a pair of chips & jumpering the data, it is far from straightforward.

First the chips are difficult to remove you need a decent rework station. Then there are 4 or 5 data signals that will need to be jumpered across the chip(s). Then the power either needs to be jumpered or a resistive load put in it's place equivalent to the effective resistance of the chips at the frequency the chips are hashing at, if this is not done and the power just shorted you would need to reduce the supply by 0.8V, and remove one set of the level shifting diodes.

Finally there is almost certainly something else I have not thought of? In theory it's possible, in practice probably not to be recommended.

Rich

So through all the mumbo jumbo, i understand OP has two bad chips? Assuming there is nothing else wrong with the board... Best thing might be to replace the 2 bad chips from 2 working chips from another bad board.

[RichBC]

I have thought about but never tried removing a pair of chips & jumpering the data, it is far from straightforward.

First the chips are difficult to remove you need a decent rework station. Then there are 4 or 5 data signals that will need to be jumpered across the chip(s). Then the power either needs to be jumpered or a resistive load put in it's place equivalent to the effective resistance of the chips at the frequency the chips are hashing at, if this is not done and the power just shorted you would need to reduce the supply by 0.8V, and remove one set of the level shifting diodes.

Finally there is almost certainly something else I have not thought of? In theory it's possible, in practice probably not to be recommended.

Rich

So through all the mumbo jumbo, i understand OP has two bad chips? Assuming there is nothing else wrong with the board... Best thing might be to replace the 2 bad chips from 2 working chips from another bad board.

Yes agreed   But the OP asked about running with chips removed, so I had a shot at an answer.. Best thing is, as you suggest, to sacrifice a board for repairs, only I guess that does not work if you only have a single failed board to play with.

I have not been that successful so far repairing the partially failed board I have. This has been partially because I only had a Hot Air Blower and also that I was never certain if the chip I was using as a replacement was good. Have now got an IR Rework Station so once I have learnt to drive it that should improve things.

Rich

[VirosaGITS]

I have thought about but never tried removing a pair of chips & jumpering the data, it is far from straightforward.

First the chips are difficult to remove you need a decent rework station. Then there are 4 or 5 data signals that will need to be jumpered across the chip(s). Then the power either needs to be jumpered or a resistive load put in it's place equivalent to the effective resistance of the chips at the frequency the chips are hashing at, if this is not done and the power just shorted you would need to reduce the supply by 0.8V, and remove one set of the level shifting diodes.

Finally there is almost certainly something else I have not thought of? In theory it's possible, in practice probably not to be recommended.

Rich

So through all the mumbo jumbo, i understand OP has two bad chips? Assuming there is nothing else wrong with the board... Best thing might be to replace the 2 bad chips from 2 working chips from another bad board.

Yes agreed   But the OP asked about running with chips removed, so I had a shot at an answer.. Best thing is, as you suggest, to sacrifice a board for repairs, only I guess that does not work if you only have a single failed board to play with.

I have not been that successful so far repairing the partially failed board I have. This has been partially because I only had a Hot Air Blower and also that I was never certain if the chip I was using as a replacement was good. Have now got an IR Rework Station so once I have learnt to drive it that should improve things.

Rich

Sound like you have a pretty nice setup. I'm guessing the first issue with repairing a board is finding out exactly what is wrong. If simply switching out chips, resistors or some other doodad works, you can probably get a dead S5 boards for 50$.

Anyways that was the trade-in value before the BTC price raise.

Maybe sidehack could simply sell 2 working chips if that is really all that is missing.

[soy]

I use to change ASICs on the thumb miners, sometimes the crystal.  These ASICs are of course even smaller, closer pins.  I would remove an ASIC if I thought there was hope but not try to replace one of these.

So, the board is sitting here.

Saw btc hit $500 today so tonight before dinner I took a C1 and got it up and running.  Tried to get my second up but in the process partly cut power to the first and stupido turned it on quickly perhaps losing the controller.  Tomorrow I'll try resetting.  Bother, my miners aren't on the .1.0 subnet.

[adaseb]

So since its on a chain, if the first ASIC goes bad all the ASICs after will go dead also?

If its the last ASIC then all would be good expect the last one ?

[RichBC]

So since its on a chain, if the first ASIC goes bad all the ASICs after will go dead also?

If its the last ASIC then all would be good expect the last one ?

Yes correct except that dependant on the fault with the ASIC the voltages to the rest of the ASIC's can be increased or decreased and that can also stop the entire chain from working.


Rich

[soy]

So since its on a chain, if the first ASIC goes bad all the ASICs after will go dead also?

If its the last ASIC then all would be good expect the last one ?

Yes correct except that dependant on the fault with the ASIC the voltages to the rest of the ASIC's can be increased or decreased and that can also stop the entire chain from working.


Rich

And notice the difference between an ASIC going x with the others o's but when the ASIC goes short to ground the whole chain goes bad.