Repairing Antminer S9j

[Artemis3]

I have seen people successfully using the 140mm 3k "industrial" Noctua with 3d printed adapters, but 200mm?

The 140mm is 158CFM, while the 200mm is just 88CFM. The factory fans are usually 210CFM...

[1714810384]

1714810384

[1714810384]

1714810384

[1714810384]

1714810384

[NotFuzzyWarm]

The 140mm is 158CFM, while the 200mm is just 88CFM. The factory fans are usually 210CFM..

And do remember that the air flow rate is not the whole story - it is the air flow against back pressure that matters the most.

Miner fans are chosen to work against a fair bit of restriction caused by the heatsinks. (that is why they run so fast and loud) Most regular PC case fans are only rated for free flow and a fairly low back pressure so against the restrictions inside of a miner the resulting flow will be MUCH lower.

[MrMik]

I have seen people successfully using the 140mm 3k "industrial" Noctua with 3d printed adapters, but 200mm?

The 140mm is 158CFM, while the 200mm is just 88CFM. The factory fans are usually 210CFM...

I don't recommend using the 200mm Noctua, but I had some lying around and also half-finished 3D print designs for adaptors, so it was worth a try.
I still think in cold environments this could work quite well, for example to keep a cellar frost free in winter without hearing the miner upstairs. The 200mm fans just turn down the speed of the stock fans by a few notches, but it does make a difference to the noise level.

[MrMik]

The 140mm is 158CFM, while the 200mm is just 88CFM. The factory fans are usually 210CFM..

And do remember that the air flow rate is not the whole story - it is the air flow against back pressure that matters the most.

Miner fans are chosen to work against a fair bit of restriction caused by the heatsinks. (that is why they run so fast and loud) Most regular PC case fans are only rated for free flow and a fairly low back pressure so against the restrictions the resulting flow will be MUCH lower.

One interesting result from testing the S9J with Phresh fan is that some of the high pitch whining noise is actually created in the miner, not in the fan. Once airflow gets high enough (or probably 'too high'), that wind starts to howl around some corners of the heat sinks.

[Lida93]

I bought a used S9j miner and it is not in good working order.

Now I am trying to decide if I should send it back or negotiate a price drop and then learn how to repair it. Any helpful hints would be much appreciated!

One of the three hashboards does not work. The red LED comes on sometimes, and sometimes a small number shows in the GH/S(RT) column, but it never heats up or produces significant hashes. Mostly it hashes zero.



The power supply has damage to multiple cables, they must have been chafing against sharp metal shelve edges or something like it. L


Some of the output capacitors in the PSU seem to be getting much hotter than others, and the miner (hashing on 2 boards only) stops hashing every few minutes and starts over. I don't know if the PSU might be too weak, have not investigated in any detail yet.

Questions:
1) Is it potentially possible to repair the hashboard? I have swapped over the cables between the control board and the hash board, the fault is in the hashboard not the cable or control board.

2) Can the capacitors be replaced? It looks to me like I can solder new ones in, but I have no idea if they are hard to find, and if I would need to replace all of them instead of the 2 that look hotter on an infrared photo.

3) Could a replacement hashboard be used, or do they need to be a 'matching set' or specific serial number?

4) I think I can de-solder the cables at the PSU board and then put shrink-wrap around the damaged parts and solder them back on. Or is that a particularly stupid idea?

Thanks for any helpful hints!

If I have to be frank with you mate you have got a lot of work and spending to do. From all indication going by your explanations I think before you end up making replacements and repairs of the said miner parts in addition to any negotiation amount the seller and you might agreed upon you must have end up expending a significant sum that could have afford you with an almost new miner in the market.  It would be better to send it back cause it's most likely after you must have made few repairs you may not make some profits enough to cover for all your expenses before another repairing needs may arise again.

[MrMik]

Bugger me, but the old PSU is suddenly working again!

After it has been sitting in time-out in a cardboard box for over a month, I decided to have a look why it is producing 10VDC instead of 12VDC, and it is again making 11.97VDC.

I wonder if there is some component that fails when used for some hours, and then slowly recovers while resting?

Any ideas what this 'self-repairing' part may be?

To find out, I have connected the old PSU to the miner with it's 2 functioning hashboards. It is now running BraiinsOS+ so I have much control over the power consumption.

The old PSU is at least several times louder than the entire miner with the 200mm Phresh fan cooling it! It is impossible to hear the miner because of the PSU noise.

I start testing the old PSU by running the miner on 900W setting, which will actually use about 657W because of the missing hashboard. This setting is the most efficient one I have found so far, using 74J/TH at the wall, including PSU, fan and WiFi extender.

I'll crank up the power until the PSU fails again and then see if anythings is obvious as the cause.

[hZti]

Sounds like the PSU overheats and triggers some kind of protection. You can check if it is full of dust or the fan runs to slow. The issue with the very loud PSU is a general issue with the standard PSU and can be solved with aftermarket Computer PSUs. If you open up the PSU be careful, it contains power even if not plugged in!

[mikeywith]

You can try to run the hashboard with a different (higher) frequency with Braiinos

Do you mean higher voltage? because usually that is the case, many hashboards will need lower frequency or/and higher voltage to run stable, higher frequency cases issues even to stable hashboards.

OP, since you like fixing stuff, the issue with that bad hashboard is kind of straight forward, the kernel log states

Code:

driver-btm-c5.c:13522:bitmain_c5_init: retry Chain[J7] has 15 asic

This means that the test signal stops at chip 15, it either reaches there and doesn't move to the next one, or simply just doesn't get there, so you are going to need to check chips 15 and 14, here is a good guide on how to proceed.

https://www.zeusbtc.com/manuals/Antminer-S9-Hash-Board-Repair-Guide.asp

[MrMik]

Sounds like the PSU overheats and triggers some kind of protection. You can check if it is full of dust or the fan runs to slow. The issue with the very loud PSU is a general issue with the standard PSU and can be solved with aftermarket Computer PSUs. If you open up the PSU be careful, it contains power even if not plugged in!

Thanks, I had already cleaned the PSU before it failed and it is still clean.

What ever made it fail (10VDC instead of 12vDC) did not revert by just cooling down again, it was still broken hours or days after the drop to 10V.
But then after weeks in the 'naughty box', when I decided to investigate, it was making 12VDC again.

I have since then powered the miner (with only 2 instead of 3 hashboards) for a day or so with the dodgy PSU and it has continued to function as it should.
Interestingly, I cannot detect a difference in efficiency (measured at the wall) between using the beat-up and noisy old PSU and my almost brand new 2050W PSU, except that the new one is practically silent.

The old PSU is back in the 'naughty box' for being too loud (but I am quite certain that the noise level is the same as when it was new. The noise would not matter if the miner was operating by using it's stock fans, but the miner is now very quiet due to the Phresh 8" 240V fan).

[MrMik]

OP, since you like fixing stuff, the issue with that bad hashboard is kind of straight forward, the kernel log states

Code:

driver-btm-c5.c:13522:bitmain_c5_init: retry Chain[J7] has 15 asic

This means that the test signal stops at chip 15, it either reaches there and doesn't move to the next one, or simply just doesn't get there, so you are going to need to check chips 15 and 14, here is a good guide on how to proceed.

https://www.zeusbtc.com/manuals/Antminer-S9-Hash-Board-Repair-Guide.asp

Thank you very much, I'll have a closer look at it when I take the miner apart the next time. At the moment the broken hashboard is in the running miner (not powered or connected to the control board, just for stopping air from finding an easier way past the 2 running hashboards).