Antminer S9 breakdown issues - How frequent?

[MachineZero]

I just wanted to get a sense of what percentage of these units break down.

I had one running for exactly 7 days and 2 boards went out.
Under warranty, but Bitmain says to expect 3-4 weeks turnaround on repairs.
Basically losing 5 weeks of hashing at the peak of it's diminishing value.

[1714007212]

1714007212

[1714007212]

1714007212

[ncsupanda]

Saw this in the main S9 board today.

I just unplug the bad boards and then it runs fine.  Out of the 12 S9's I have bought from Bitmain, 5 of them have needed repair.

They have never gotten over 72F.  I have 140 S7's and maybe 10 bad boards between them all.  Out of the 200+ S5's I had there were maybe 5 bad boards total and out of the 300+ S3's I had maybe 4 bad boards.

I really hope they fix the S9's.  The R4's seem to be running great.  I have had no problems with them.

So 7 bad boards out of 36 total S9 boards = 19.4% failure rate.  Ugh.

[takagari]

Your best bet would be to look at the main thread.

What is with everyone starting a new S9 thread lately every time they have a question and simply figure everyone will give them the answer? Lazyness?

[MachineZero]

Your best bet would be to look at the main thread.

What is with everyone starting a new S9 thread lately every time they have a question and simply figure everyone will give them the answer? Lazyness?

Wanted to open up discussions. I already read what was there.
It seems to be yet another hurdle to ROI if you need to figure out the downtime of repairing under warranty.

Also It would be nice to get perspective on repairs from bitmainwarranty in the US. I spent over $100USD to ship back to HongKong and it took 6 days. I'm wondering how much it would have cost to pay out of pocket in comparison to international shipping and lost weeks. bitmainwarranty would not quote me without the machine shipped to them.

[NotFuzzyWarm]

bitmainwarranty would not quote me without the machine shipped to them.

I'd rather expect this to happen to all repairs involving batch-16 and higher (auto-tune rigs) and sounds like same for the R4.

Chips/boards that perform near perfect are rare. Call it curse of 16/14nm. That said, the point of autotune is so Bitmain can pick and choose between 'meh' and better boards so the miner as a whole gives advertised hash rate. They don't want to use a better board than needed nor send out a weak sister one when it was a better board that croaked.

[MachineZero]

bitmainwarranty would not quote me without the machine shipped to them.

I'd rather expect this to happen to all repairs involving batch-16 and higher (auto-tune rigs) and sounds like same for the R4.

Chips/boards that perform near perfect are rare. Call it curse of 16/14nm. That said, the point of autotune is so Bitmain can pick and choose between 'meh' and better boards so the miner as a whole gives advertised hash rate. They don't want to use a better board than needed nor send out a weak sister one when it was a better board that croaked.

I did notice all three boards had a very wide variance of frequencies to yield 11/th. Guess I was better off if I chose the 14/th S9 as it must have performed optimally at the spec frequency. One board was always consistently hitting 97-107 deg (fastest clock speed), while the other two were consistently fluctuating between 79-87 deg.

The slowest frequency autotune board was the one to survive though.

Just got the update 5 calendar days after EMS reported delivered, that the ticket status was changed from "uncollected" to "collected". A follow up email stated that the S9 will be repaired under warranty and turned around in 3 business days.

[Exoskeleton]

I seem to get lucky on bitmain gear because my failure rate has been low for the few S5/S7/S9 that I have owned. I have a few S9s and the only ones I have had a problem with were the two 11Ths units I bought. They still work with all 3 boards but sometimes they only run at 7-8Ths. A quick hard reboot always does the trick. This low hashing problem only happens after they reset for some reason like loosing internet for a second. I may have just been lucky so far.

[lokomotive]

So how many batches they have made alredy 22+? And it still has bad boards and etc.

[philipma1957]

So how many batches they have made alredy 22+? And it still has bad boards and etc.

it has been their worst machine in terms of breakdowns

[Gogreen]

the heat issue is a big problem, so many chip packed togather 

[lokomotive]

They should have just made a bigger box or something

[grumpyshamurai]

The first time i opened my miner I shook my head at the 100% lack of thought that went into the performance of cooling. Purely cost effective for them. If anyone has a dead s9 board they dont mind donating to the cause im more then happy to see if a water block is a possibility. Biggest concern is removing the dollar store heatsinks they seemed to have welded on.

[lightfoot]

Actually I have been wondering about Bitmain's devices. Would it be worth it to ship units to me for repair (not affiliated with Bitmain at all, if they wanted to send me schematics that would be nice) for a lower time to repair?

If so I'll look into their next unit and see what it would take to fix failures. But maybe the failure rate is low enough/the warranty is strong enough for this to not make sense?

C

[MachineZero]

So total turnaround from shipping via USPS from NYC and them sending back* was 2.5 weeks. However, I received a brand new 12/TH S9 to replace my defective 11/TH. Serial numbers, firmware corroborated.

The new 12/TH has been running flawlessly 24/7 for almost 3 weeks.

In retrospect it would have been worth it to pay for repairs at bitmainwarranty.com as the shipping out cost and lost mining time probably was about the same as what they would have charged for an out of warranty repair. But it appears that they replace units too according to other forum threads I've read. Must be Bitmain or have a very close association with them since their help desk software was exactly the same.

[philipma1957]

So total turnaround from shipping via USPS from NYC and them sending back* was 2.5 weeks. However, I received a brand new 12/TH S9 to replace my defective 11/TH. Serial numbers, firmware corroborated.

The new 12/TH has been running flawlessly 24/7 for almost 3 weeks.

In retrospect it would have been worth it to pay for repairs at bitmainwarranty.com as the shipping out cost and lost mining time probably was about the same as what they would have charged for an out of warranty repair. But it appears that they replace units too according to other forum threads I've read. Must be Bitmain or have a very close association with them since their help desk software was exactly the same.

Bitmainwarranty is in some kind of business partnership with bitmain .

They have two or three locations outside China to do repairs.

They are better  if you have out of warranty gear and live in the lower forty eight states.

[fanatic26]

Biggest concern is removing the dollar store heatsinks they seemed to have welded on.

Use a heat gun at about 400c and make sure you wiggle them back and forth slightly while heating, you will feel them loosen and that way you dont overheat them which can cause the entire chip to be pulled out with the heatsink. Also if removing heatsinks from both sides ALWAYS do the larger silver heatsinks first or you WILL pull the chip off when removing them from the other side. The trick is to roll them off and not try to pull straight up as they loosen. Then use a bit more heat and a razor to scrape off the remaining adhesive.

[lightfoot]

Bitmainwarranty is in some kind of business partnership with bitmain .

They have two or three locations outside China to do repairs.

They are better  if you have out of warranty gear and live in the lower forty eight states.

Ah, this explains why I haven't seen an antminer cross my repair desk. Which is fine, it's good to see a vendor actually support their product properly.

C

[fanatic26]

It has been said before but bitmainwarranty is not a bitmain company.


As far as the S9 reliability goes I have personally seen a 25%+ failure rate in the first 90 days of operation and I have heard a 30% failure rate number thrown around by other people with very large deployments (900+ units)

[lightfoot]

By support I mean work with an outside group to provide parts, spares, schematics, and whatnot to assist in fixing. For myself I get nothing, nada, and zilch which means I have to either reverse engineer or fuzz out every design the old fashioned way.

On the other hand it means I'm not bored. But I doubt I could compete effectively with someone who has the parts lists and/or schematics. For whatever reason miner builders don't like to share that.

C

[PeaMine]

3 S9, two failed boards total.
1 before autotune, batch 6, and one around batch 19 had a bad board after a couple months.
S7, 2 dead machines, 3 bad boards, out of 11.
S5, a few bad chips, all going within 10% of original performance and 100% uptime since day 1.  Good build and good reliability.
S4, many used, too many issues to count, terrible with heat.  Not sure due to second hand market.