Bitmain's Released Antminer S9, World's First 16nm Miner Ready to Order

[not.you]

Anyone tried to overclock this one above 650? I have mine running very cool so was thinking if it would stand a little overclocking. What do you think?

If you have a 650 batch then you probably could. I have a 600 batch miner running at 606 without a problem. Just keep in mind that the S9's in general are known for a high failure rate so you are taking a bit of a risk.

[numnutz2009]

Wow, my A/C is set at 78F. At 65 I would probably spend something crazy here in TX (we were still solidly above 90F a couple of days ago).

Apart from this, it is unlikely that bumping the temp a few degrees would make much of the difference considering that boards are typically run in 50ies and 60ies.

that would be true if i paid for power and if i were using a typical home ac unit. the units i have in each room are industrial roof mounted chillers. they do bio research here so the labs have to be very cold so since they are running all of the time the building manager diverts some of the cold air into each of the rooms i use ontop of the dedicated ac unit for each of the rooms so i get double the cooling and also if one were to have issues the other is there to pick up the slack until the first is repaired. better safe than sorry i suppose.


its getting cold here at nights though so i started running some of the s7's i upgraded at lower clock speeds with fan filters (home mining it dusty. the ac units in the offices do all the air filtration so those miners dont need filters on them because they dont get dusty ever) an use them as space heaters for my bedroom and the basement and so on. they work quite well for keeping rooms warm and will offset my electricity bill for sure. i used to run a 1500w IR space heater in my bedroom which would run alot and wasted tons of power. now i run my s7 and use 700w and i make some btc on the side while doing it.

as for overclocking the s9's....if you have an older version of the s9 you can but like others said its risky and not really worth it. you may be able to squeeze a few extra gh/s out of the miner but is it really worth risking your miner for a few gigs?? i sure dont think so lol. with the new version of s9 it auto clocks itself so you cant set the speed of anything manually like you could before. one thing that is interesting is that by clocking itself to whatever speeds you can see which boards are the top performing ones and if your into this sorta thing you can take the good boards and combine them into 1 single miner that runs very well instead of 1 miner having 1-2 good boards and another miner having 1 good board and so on but it might be best to not consolidate them.

[ZymurBits]

Looks like another day, another build.  Antminer-S9-all-201610250334-autofreq-user.tar.gz

Any idea what's in this one?

[alh]

Looks like another day, another build.  Antminer-S9-all-201610250334-autofreq-user.tar.gz

Any idea what's in this one?

You might want to "open" up the file and see what's inside. I don't mean try and install it, but within a file named like <blah>.tar.gz there can actually be multiple files. While it might be a stretch to hope for, many companies will include a TXT with info about what's changed and possible installation info.

You can't count on Bitmain to do that, but it won't hurt to look inside.

If you are a Windows centric person, 7-Zip can help out.

[not.you]

I am curious if any of these newer autotune firmwares have actually worked to autotune the older model miners.  So far from all of the posts I have seen, installing the new firmwares on older miners does not actually get them to autotune at all.

[Sierra8561]

I am curious if any of these newer autotune firmwares have actually worked to autotune the older model miners.  So far from all of the posts I have seen, installing the new firmwares on older miners does not actually get them to autotune at all.

No autotune for me on old releases. I am currently testing the new release, I'll let you know if anything changes.

[citronick]

what causes this "socket connect failed..."?

Is there a firmware to get rid of this?

[Sierra8561]

what causes this "socket connect failed..."?

Is there a firmware to get rid of this?

I've wondered the same thing. All firmware I have tried suffered from this problem.

[NotFuzzyWarm]

what causes this "socket connect failed..."?

Is there a firmware to get rid of this?

I've wondered the same thing. All firmware I have tried suffered from this problem.

AFAIK, the socket connection error refers to a Virtual data port/connection setup between Linux/CG or BM miner/ and the hash boards themselves. Because the hash boards are busy and not taking calls right now, the full connection to the GUI cannot be setup.

The OS/miner firmware will keep trying to talk to the boards. Before the auto0tune update it normally took maybe 30-45 sec for first (warmup) connection, followed by a restart at normal settings.

After auto-tune... 'whole new ballgame as to when the connection can be made. It (a good Socket connection) comes and goes each time a board is tested before things settle down...

[psyscope]

i had the strangest reply from bitmain support today. one of my new s9's has been running 1.5th/s slower than it should be since the day i got it which is out of spec by more than double the numbers posted by bitmain. anyway i was talking to a very nice person names "barbara" which im sure is a fake name like when you call comcast support and get someone in india or south america....but anyway she was working with me then randomly this person with the username "overseas" takes over my ticket and asks me to do everything that i already did and when i replied back telling him no i already did all of this he came back and said basically my miners are cooled too well. he said i need to raise the temps in the rooms i run my miners in because my miners are running too cold. the funny thing is all of my s9's are in one room while my s7's and titans are in another room in the building next to it. each room has its own dedicated ac unit pumpin nice cool 65/70 degree air into the room. the room with all s9's runs perfectly (same with the other room of course but this is about the s9 miner not the other miners) except for that 1 s9 that runs out of spec.

i usually back bitmain to a certain point unless im wronged because i cant go by anyone elses experiences except my own but who would have thought i would get a reply back telling me to run all of my miner hotter and harder in order to fix this one miner running outta spec? all of the other s9's run perfectly at the temps selected in the room so what makes this particular miner less willing to run at spec? it doesnt get anymore extra cooling that the other s9's dont already get. its plugged up to the same 230v apc 5000w pdu as 2 other s9's are but somehow these speeds are my fault and not bitmains fault....lol not likely. just thought i would throw this info out there because i found it amusing.

if your miner using there new auto freq firmware
i lost 3 TH/s from mine and after a week with them on a ticket i told barbara to move my ticket to some one more experiences and he did
after that (Customer Service-Overseas) take my ticket and they asked my kernel log more than once, every time this issue happen
after all they told me that (There is a bug in our latest firmware. We have fixed it.)
my miner now working for 1d18h31m41s without losing hash rate but i will upgrade it now for there latest

Typical Chinese customer service, they find a new bug that is in the firmware, and then will work on it. The next release, the same bug will still be there, and they will work to fix it.... and on and on. Now who thinks that they have an internal firmware for their machines to run better. 

[Vasiliy]

Antminer-S9-all-201610260255-autofreq-user.tar.gz
https://shop.bitmain.com/files/download/Antminer-S9-all-201610260255-autofreq-user.tar.gz

[HagssFIN]

They keep releasing new firmware almost every day?  
I tried the October 25 one in my S9, batch 2 because it has problems reaching the nominal 12.93Th/s hash rate.
The update was no good for my miner.
I usually need dosens of restarts with the miner to reach the nominal hash rate.

[numnutz2009]

the socket connection error will show up until the miner starts hashing away. if you havent upgraded your firmware the auto tune feature can take the better part of an hour to complete so you might see that error for a long while....if its been more than an hour something is wrong.

[edonkey]

Does the autotune function run every time you restart mining?

Is it actually hashing during the tuning time?

Seems like the firmware could cache the results of the last tuning in the local file system, then use it the next time instead of tuning all over again. To detect environment changes, it could flush the cache if there's a big temperature change.

Some people have scripts set to restart the miner software. If it takes an hour hit every time the miner restarts, that could be bad.

My S9's are hashing fine with the old firmware. I'm going to leave them alone.

[AriesIV10]

Does the autotune function run every time you restart mining?

Is it actually hashing during the tuning time?

Seems like the firmware could cache the results of the last tuning in the local file system, then use it the next time instead of tuning all over again. To detect environment changes, it could flush the cache if there's a big temperature change.

Some people have scripts set to restart the miner software. If it takes an hour hit every time the miner restarts, that could be bad.

My S9's are hashing fine with the old firmware. I'm going to leave them alone.

Don't change it if it is working.  On the other side, I have had a couple of problem miners that would decrease hashrate every few hours.  With the installation of the new firmware, they became stable!  I am really a fan of the new firmware.

[d57heinz]

A simple fix for these would have been to use the temperature code they already have implemented to set the throttling. It should be user selectable the freq then the miner needs to run a check every 30 min to see if it has reached 80c or 110 chip temp.  If so downclock 2 or three freq and restart mining.  Then same can be made if it stays below a set temp for a certain amount of time then it ups the freq one notch and resumes mining.  Of course all this would be able to be toggled on or off not automatic. Just a thought. Also I don't understand why they aren't saving the values so that it doesn't have to run the check over again thus losing an hour of mining per reset.  Then have a reset button within the code to allow us to rerun the initial tests.

Best Regards
d57heinz

[mamiakimo]

my miner B9 600M working now on there new firmware and one board on 104C chip temp
we are in the beginning of winter and the Ambient temperature less than 30C
What will happen to the temp next summer ?
do you think this firmware will lower the freq by self or what ?

[BITMAIN]

Dear Bitcoiners,

We are finally launching two more batches of the Antminer S9 at 2300 hours (GMT+8, Beijing time) today (27 October 2016).

Batch 18 and batch 19 can deliver a hashrate of 11TH/s and 13.5TH/s, respectively.

We advise those who were unable to order before the quick sell-out of the previous two batches to place their order immediately after they are available at https://goo.gl/pFtSev (batch 18) or https://goo.gl/fNKU4T (batch 19).

Please feel free to reach out to us at support.bitmain.com fo rany queries or tech support.

Happy mining,

the Bitmain team

[not.you]

my miner B9 600M working now on there new firmware and one board on 104C chip temp
we are in the beginning of winter and the Ambient temperature less than 30C
What will happen to the temp next summer ?
do you think this firmware will lower the freq by self or what ?

Did it set different freq for different boards?

[wolfen]

Just received a new 13.5th s9.
Had oct 21 software on it.
Saw bitmain was up to oct 27
Flashed to oct 27. Running ok.
No fan control, no freq control.
When running ok green light on 8 seconds, off 8 seconds.
Seems to take an eternity to boot up. Thought it was dead. Need to be patient.

Fans are running 2500 (slow)
two boards auto clocked at 640, one at 590
hardware errors at 0.0001% (nice)