Bitmain S21 Gen. Asic reliability?

[Auslander]

Hey, was wondering if anyone had insight into the reliability of bitmain’s newest generation of BTC ASICS (S21, S21Pro, S21XP etc). Common issues with units, hashboard reliability/build quality, power supply issues or anything else y’all have had come up. Anyone managing a large farm or those that have a large sample size of machines would be much appreciated. Curious in general if you even have one unit and have had issues what those have been. Thanks in advance!

[mikeywith]

I have investigated the subject a while ago and I can tell you for sure that there is no single correct answer to this question, reading from multiple sources that I know and trust, the failuare rate on the S21s series is anywhere between 2% a year to 40% in a few months.

You may be asking how come? Bitmain has this habit of inconsistency with each batch, you are going to have to count on your luck, I know the guy who reported 2% failure was legit, same for the guy that reported 5% and the same for the 40%.

I would say however that you should be prepared for at least 5% failure rate for the first year, and probably 2x or 3x going forward.

Most issues with all Antminers would be the hash boards, so it's something beynod repair for the average folks, they have slightly better PSUs than MicroBT but I would rather have 5% failure rate on PSUs I can buy and plug in myself than having to ship a hashnoard and hopping to get it fixed.

[Auslander]

Thanks mikey! Inconsistency with the batches makes sense but 40% is staggering , I hope they were able to warranty with bitmain at that rate. For sure hash boards would be the main point of failure, psu’s are much cheaper and easier to replace. Do you know how many units they had at those failure rates? Also do you know if they’re considering the % of failure per unit or per hashboard. For example 2% of all units had a board go bad vs 2% of all hashboards went bad. Thanks!

[Teejani]

Unfortunately for me, it is looking closer to 40+% and they have only been running since July.

At the moment I have only diagnosed 10% with PSU and/or Fan issue which at least is not an expensive fix however 2% looks like a hashboard issue. I am yet to find out what is happening with the rest.

Warranty with Bitmain is six months from shipping date. The S21 were delivered in April and the farm was set up in July, problems started to arise in the fourth month.

[Auslander]

Unfortunately for me, it is looking closer to 40+% and they have only been running since July.

At the moment I have only diagnosed 10% with PSU and/or Fan issue which at least is not an expensive fix however 2% looks like a hashboard issue. I am yet to find out what is happening with the rest.

Warranty with Bitmain is six months from shipping date. The S21 were delivered in April and the farm was set up in July, problems started to arise in the fourth month.

That sucks to hear, I hope they start trying to address and fix some of these build issues with future batches but knowing bitmain it’s probably a slim chance. Fans make sense, they go out all the time but do you mean the fan bearing/sensor is going out or has it been issues with the control board connections to the fans? I would love to get some S21’s but I’m definitely gonna be waiting for the bear market when they’re cheap and hope future batches become more reliable. For now I’m going steady with my S19XP’s.

[CryptoPig333]

I have had not the best experience with the S21s. My fleet right now is 10x Antminer S21 (200T) which has been running since March 2024. My hosting provider has had to manage 6x warranty repairs since then for me. Over $1,200 on shipping! Not happy. Glad to hear I am not the only one.

[mikeywith]

Do you know how many units they had at those failure rates?

The guy who reported 40% probably had a few hundreds of them running.

Also do you know if they’re considering the % of failure per unit or per hashboard. For example 2% of all units had a board go bad vs 2% of all hashboards went bad. Thanks!

Failure is a failure, it doesn't really matter because you would have to ship the entire unit to Bitmain, so even if it's just a single board that went bad (which is usually the case), you will still lose that unit for a good while until they repair it and send it back to you, it's extremely unlikely for an entire gear to go bad at once, it will always be one part of the many parts it consist of.

Depending on the market conditions, you will always have to make specific decisions when your gear fails, sometimes voiding the warranty by trying to fix it yourself or running it with 2 hash boards is better than sending the whole unit for god knows how long, sometimes fixing the hash board at a local repair shop is cheaper given that you need to pay for shipping, say you are sending 5 units to another country/state, have to pay $200 for shipping and to miss out on 1 month of mining, but the guy next door fixed them for $500 in 3 days and you don't pay for shipping, I still remember when Bitmain sent their technicians to some large farms in China to have them fix the 60% something failure on the 17 series, it would probably cheaper for Bitmain than having to ship back everything.

[philipma1957]

Do you know how many units they had at those failure rates?

The guy who reported 40% probably had a few hundreds of them running.

Also do you know if they’re considering the % of failure per unit or per hashboard. For example 2% of all units had a board go bad vs 2% of all hashboards went bad. Thanks!

Failure is a failure, it doesn't really matter because you would have to ship the entire unit to Bitmain, so even if it's just a single board that went bad (which is usually the case), you will still lose that unit for a good while until they repair it and send it back to you, it's extremely unlikely for an entire gear to go bad at once, it will always be one part of the many parts it consist of.

Depending on the market conditions, you will always have to make specific decisions when your gear fails, sometimes voiding the warranty by trying to fix it yourself or running it with 2 hash boards is better than sending the whole unit for god knows how long, sometimes fixing the hash board at a local repair shop is cheaper given that you need to pay for shipping, say you are sending 5 units to another country/state, have to pay $200 for shipping and to miss out on 1 month of mining, but the guy next door fixed them for $500 in 3 days and you don't pay for shipping, I still remember when Bitmain sent their technicians to some large farms in China to have them fix the 60% something failure on the 17 series, it would probably cheaper for Bitmain than having to ship back everything.

40% is too high if you have 100 pieces and 40 die it could be you not bitmain

if you have 5 and 2 die that could be bitmain.

I have found aftermarket firmware and or boards can help kept gear up and running.

Sending an entire machine in to repair and losing 200th when only 1 board is dead can be avoided with an epic controller as it lets you run 1 2 or 3 boards.

I rather give up a board then a whole machine as many times bitmain takes weeks to send back a miner and the fix does not last at all.

[Auslander]

Failure is a failure, it doesn't really matter because you would have to ship the entire unit to Bitmain, so even if it's just a single board that went bad (which is usually the case), you will still lose that unit for a good while until they repair it and send it back to you, it's extremely unlikely for an entire gear to go bad at once, it will always be one part of the many parts it consist of.

Yea I get that, I was just trying to get a better idea of the ratio of hashboards that have issues per unit. Definitely depending on the circumstances it can be cheaper and easier to just find someone locally to do repairs and just keep running with two boards to minimize the down time. Especially with cases I’ve seen where units get held hostage for months or get “scrapped” over something that could’ve been fixed. Thanks for your input!

[Auslander]

40% is too high if you have 100 pieces and 40 die it could be you not bitmain

if you have 5 and 2 die that could be bitmain.

I have found aftermarket firmware and or boards can help kept gear up and running.

Sending an entire machine in to repair and losing 200th when only 1 board is dead can be avoided with an epic controller as it lets you run 1 2 or 3 boards.

I rather give up a board then a whole machine as many times bitmain takes weeks to send back a miner and the fix does not last at all.

I was thinking the same thing 40% seems way too high. I can definitely see that if they’re running at too high of ambient temp or with unfiltered air. I prefer to use aftermarket firmware too so bitmain warranty is out the window anyways, and like you said I can run one or two boards with the firmware and keep stacking sats while waiting on repairs.

[mikeywith]

The guy who reported 40% probably had a few hundreds of them running.

The guy runs a huge mine, so I am pretty certain it's not bad cooling or anything of that nature, it's probably a bad batch, I had what 100% failure rate on my S17/T17s while your was a lot lower, the first large farm in China reported 40% as well IIRC, the 17 series gears I had run at very good conditions, I abused MBTs and S9s a lot harder and the failure rate was nothing close, the 17 series were failing like flies regardless how cool and clean I kept them, I am not saying the 21 series is as bad as the 17 series, but with Bitmain -- anything is possible.

Yea I get that, I was just trying to get a better idea of the ratio of hashboards that have issues per unit. Definitely depending on the circumstances it can be cheaper and easier to just find someone locally to do repairs and just keep running with two boards to minimize the down time. Especially with cases I’ve seen where units get held hostage for months or get “scrapped” over something that could’ve been fixed. Thanks for your input!

Can't speak about those S21s although I am sure they were mainly a single hashboard gone bad, however, judging by the 17 series that I wasted a few Bitcoins on, you first get an unstable hashboard that causes the miner to reboot, it eventually gives you the chain x found 0 chips, you unplug that board, run two, they go on for a while and the thing repeats with a second board, eventually, you get 1 board running on this miner while the next miner has 2 runnings waiting for the other one to fall, eventually you end up with dozen miners each running 1 or 2 boards at most, you get an aftermarket firmware which allows you mix and match because obviously, you wouldn't want to maintain 50 miners that operate at the capacity of 20 miners.

[Kryss191]

I don't know I think the S21 series is a solid pick for the performance it delivers, but if you’re buying a bunch,man there are so many things that can go wrong like the power supply, heat, and making sure firmware updates go smoothly

[Teejani]

I have had not the best experience with the S21s. My fleet right now is 10x Antminer S21 (200T) which has been running since March 2024. My hosting provider has had to manage 6x warranty repairs since then for me. Over $1,200 on shipping! Not happy. Glad to hear I am not the only one.

Has your hosting provider informed you of what the warranty claims were?

After some troubleshooting at my new farm, the issues seem to be power related (spike in voltage from the grid during rain/thunderstorm) which fried the PSU over time.


For anyone else experiencing a similar situation, the solution for my site seems to be reduce the voltage limit to the site and replace the PSU on the miners.

[BitJannik]

.From what I've gathered, the failure rate of these machines can vary depending on the production batch and usage environment. Issues like thermal paste application and hash board durability seem to be common with some reports like overheating problems that could cause a higher than expected failure rate.

Preparing for around 5% rate in the first year and potentially higher subsequent years seems reasonable.

You can check more on this article https://www.futurescope.co/bitmain-antminer-s21-review/