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

[not.you]

I have some of the Sinewave Line Interactive models.  I don't use them for always on though, miners go on the NCL bank and get turned off after just a few minutes if the power goes out and stays out.  Servers are on the main bank.  But what is super cool about having them is you can remotely power cycle the NCL bank as long as you have access to a computer on the same network.  So if I am away and a miner acts up and I remote in and still can't make it cooperate I can power cycle it which pretty much always brings it back to life.

[1713994824]

1713994824

[1713994824]

1713994824

[1713994824]

1713994824

[1713994824]

1713994824

[1713994824]

1713994824

[1713994824]

1713994824

[InvestexMarkets]

I have some of the Sinewave Line Interactive models.  I don't use them for always on though, miners go on the NCL bank and get turned off after just a few minutes if the power goes out and stays out.  Servers are on the main bank.  But what is super cool about having them is you can remotely power cycle the NCL bank as long as you have access to a computer on the same network.  So if I am away and a miner acts up and I remote in and still can't make it cooperate I can power cycle it which pretty much always brings it back to life.

Thats interesting. I have similar setup I've noticed sometimes when the power cycles thought It sometimes won't reconnect to the network, Had issues with this for a while now. seems intermitten sometimes it works. sometimes it hangs. nothing.

[NotFuzzyWarm]

I should look into that for mine. At home I have a 11kw permanent genset that kick's in after ~30sec so the on-battery time is minimal but I DO like that remote power cycling. Neva thought of that.

[ps_jb]

Weighs a freakin' ton though.

Sure - suppose to be a pure lead acid battery

[Ethey]

i would maybe buy one of theese after the halving and if it gets a little cheaper
so far its to heavy for my Moneyback

[notlist3d]

I have been doing a little testing. I don't have one in person so no high res pictures to share of the unit.   But I can say it seems to run very well right around 14T just as expected.  It reported very close to the 14T with very low hardware errors.  So again all signs here point to good things.

I used Antpool to get a 24 hour baseline which was at 13.72 TH/s.  Again this is a early test and the slightly lower hashrate could be due to things such as network variance, pool performance, unit itself, etc.  I do plan on testing other pools for more baselines. On the picture below is also an S7 just as reference. The S9 does tower above it on it's hashing speed.

[kodokbuduk]

i would maybe buy one of theese after the halving and if it gets a little cheaper
so far its to heavy for my Moneyback

good idea

[citronick]

I have been doing a little testing. I don't have one in person so no high res pictures to share of the unit.   But I can say it seems to run very well right around 14T just as expected.  It reported very close to the 14T with very low hardware errors.  So again all signs here point to good things.

I used Antpool to get a 24 hour baseline which was at 13.72 TH/s.  Again this is a early test and the slightly lower hashrate could be due to things such as network variance, pool performance, unit itself, etc.  I do plan on testing other pools for more baselines. On the picture below is also an S7 just as reference. The S9 does tower above it on it's hashing speed.

great info.

any comments on the temps and fans - did you hear any "woo-woo" sound like the S7s?

[philipma1957]

my 2 are on the truck.  can not wait to set them up.

[notlist3d]

I have been doing a little testing. I don't have one in person so no high res pictures to share of the unit.   But I can say it seems to run very well right around 14T just as expected.  It reported very close to the 14T with very low hardware errors.  So again all signs here point to good things.

(picture above)

I used Antpool to get a 24 hour baseline which was at 13.72 TH/s.  Again this is a early test and the slightly lower hashrate could be due to things such as network variance, pool performance, unit itself, etc.  I do plan on testing other pools for more baselines. On the picture below is also an S7 just as reference. The S9 does tower above it on it's hashing speed.

(picture above)

great info.

any comments on the temps and fans - did you hear any "woo-woo" sound like the S7s?

I don't have the S9 in person this is being accessed remotely.  So I can't say far as noise but fan's one is currently running at 4200 RPM and other at 3800 RM.  On temps they look good but I don't have another another one so I only have one to base it off of and can't compare it with being accessed remotely.

I can say currently around 8 hours into my 24 hour test on F2 and seems to be running great so far:

With not having it in person I can't say on somethings I know everyone want's to know/see.  But I can't wait to start seeing others on forum receive and see some pictures and comments about it in person.

Also on pools I have not been able to give it a 24 hour run yet but did test Kano for around 2 hour's and no issues.   So again all signs point to good.   I am testing pools for 24 hour runs based on pool hashrate currently as kinda a general path to follow.  So far very positive going off hashrate and performance.

[Xircom]

I have been doing a little testing. I don't have one in person so no high res pictures to share of the unit.   But I can say it seems to run very well right around 14T just as expected.  It reported very close to the 14T with very low hardware errors.  So again all signs here point to good things.

I used Antpool to get a 24 hour baseline which was at 13.72 TH/s.  Again this is a early test and the slightly lower hashrate could be due to things such as network variance, pool performance, unit itself, etc.  I do plan on testing other pools for more baselines. On the picture below is also an S7 just as reference. The S9 does tower above it on it's hashing speed.

Could you try to connect it to Kano.is, since rumor is that Bitmain made a hack of cgminer, so most likely only their own Ant "Shit" pool works ??

[HerbPean]

my 2 are on the truck.  can not wait to set them up.

Yeah exciting time ! ;-)

Can't wait for the review

Have a great one Phil !

[bspurloc]

at 1275watts is it now ok to use something like a EVGA Gold G2 1300watt?
 Been Eyeballing 1600watts ones but 1300 is liek 150$ cheaper

you'd be pushing it pretty hard.

easy solution is to underclock the S9 slightly, such as to 13TH - which would reduce your PSU load by 100W

Under clock a bit more to 11.5th. The 1300  g2 will work fine..

Then figure your math. You save 150 on the psu. But lose 2000gh to 2500th.

At current diff that's about $3 per day before power costs. So probably 100+ days ROI on that difference alone... I'd say if one already has a 1300W PSU they should probably underclock. If buying new - I'd buy a server PSU anyway.

well the 1300 only has 6 atx connectors so lame.

first of all, eight (four singles and two doubles=8); second, I bought two PCIe split cables (for $4 ea) that increase total numbers of connectors to 10. has been running many B1-5 and 6 like this for 6-8 mo already on EVGA 1300. S7 b1,3,5 has similar 1293 wattage, B2,4,6 even less watts.
EDIT: exactly what @suchmoon said as well ^^^^

where did you get the PCIe split cables. I just want to make sure I get same gauge wire as you! random amazons are bad choices! I did that for my S3's and got like thinner than 22 gauge or something. almost fried them. rewired them with 18 gauge.
 so using a 1300watt GOLD how much under clocking? like a 4.7 S7 needs to run at work? 4.5? or something worse?
thanks

[philipma1957]

at 1275watts is it now ok to use something like a EVGA Gold G2 1300watt?
 Been Eyeballing 1600watts ones but 1300 is liek 150$ cheaper

you'd be pushing it pretty hard.

easy solution is to underclock the S9 slightly, such as to 13TH - which would reduce your PSU load by 100W

Under clock a bit more to 11.5th. The 1300  g2 will work fine..

Then figure your math. You save 150 on the psu. But lose 2000gh to 2500th.

At current diff that's about $3 per day before power costs. So probably 100+ days ROI on that difference alone... I'd say if one already has a 1300W PSU they should probably underclock. If buying new - I'd buy a server PSU anyway.

well the 1300 only has 6 atx connectors so lame.

first of all, eight (four singles and two doubles=8); second, I bought two PCIe split cables (for $4 ea) that increase total numbers of connectors to 10. has been running many B1-5 and 6 like this for 6-8 mo already on EVGA 1300. S7 b1,3,5 has similar 1293 wattage, B2,4,6 even less watts.
EDIT: exactly what @suchmoon said as well ^^^^

where did you get the PCIe split cables. I just want to make sure I get same gauge wire as you! random amazons are bad choices! I did that for my S3's and got like thinner than 22 gauge or something. almost fried them. rewired them with 18 gauge.
 so using a 1300watt GOLD how much under clocking? like a 4.7 S7 needs to run at work? 4.5? or something worse?
thanks

10 % downclock      so 4700 = 4230

or do a 15% downclock to 4000gh. 

I run my s-7 at 4000gh on a seasonic 1200 plat.

 batch 2 been running since october  at freq 512 vs 575

note batch 2 uses 162 chips  not 135

[bspurloc]

at 1275watts is it now ok to use something like a EVGA Gold G2 1300watt?
 Been Eyeballing 1600watts ones but 1300 is liek 150$ cheaper

you'd be pushing it pretty hard.

easy solution is to underclock the S9 slightly, such as to 13TH - which would reduce your PSU load by 100W

Under clock a bit more to 11.5th. The 1300  g2 will work fine..

Then figure your math. You save 150 on the psu. But lose 2000gh to 2500th.

At current diff that's about $3 per day before power costs. So probably 100+ days ROI on that difference alone... I'd say if one already has a 1300W PSU they should probably underclock. If buying new - I'd buy a server PSU anyway.

well the 1300 only has 6 atx connectors so lame.
  What server PSU? I only have 110volt outlets. I'm not paying the power bill so don't really care about those costs.
I assume with a server PSU you have to add the 10 atx connectors. any good place for those if so? the cheapos are made with wire too thin for these amp ratings.

If I recall correctly EVGA G2 1300W has 8 PCI-e, two of which are shared but sufficient gauge to do ~150W or so needed for an underclocked S7/S9. A high-quality aftermarket splitter should be ok for the 9th one and the controller can be powered off Molex. Attempt this ONLY if you know exactly what you're doing otherwise it's a fire hazard.

If you don't care about the power costs I'm guessing you won't care about installing a 240V circuit either so perhaps the server PSU route is not for you

  I didnt say i didnt care, I said I wasnt paying the power. all my 220 outlets are in use. so server ones people use are all 220v?
There is no secret 1400+ watt 110v server power supplies out there?
 there is always something out there man!
When I was mining USB miners I found a dude selling old old old brand new USB 1.0 MAC 6 port hubs and used cisco router 5v 7amp adapters for them for dirt cheap while everyone else was buying super expensive hubs and power supplies. The dude was flabbergasted why I bought every singe one he had. When no one had bought any for over a year. lol. who would want a USB 1.0 hub lol. in MAC Fancy colors too lol.
  Each HUB came with a 5v 3amp adapter which I promptly sold so all the hubs cost me like $2 each. lol.
gotta be some work around!
  can I use 2 750 watt supplies to run s7/s9? I will be shutting down some S3's opening up 750 watt supplies. still need some PCIe split cables though since the 750's only have 4 PCIe cables on them.
  If I did that how about CPU to PCIe cables? they have like 2 CPU cables each.

[bspurloc]

at 1275watts is it now ok to use something like a EVGA Gold G2 1300watt?
 Been Eyeballing 1600watts ones but 1300 is liek 150$ cheaper

you'd be pushing it pretty hard.

easy solution is to underclock the S9 slightly, such as to 13TH - which would reduce your PSU load by 100W

Under clock a bit more to 11.5th. The 1300  g2 will work fine..

Then figure your math. You save 150 on the psu. But lose 2000gh to 2500th.

At current diff that's about $3 per day before power costs. So probably 100+ days ROI on that difference alone... I'd say if one already has a 1300W PSU they should probably underclock. If buying new - I'd buy a server PSU anyway.

well the 1300 only has 6 atx connectors so lame.

first of all, eight (four singles and two doubles=8); second, I bought two PCIe split cables (for $4 ea) that increase total numbers of connectors to 10. has been running many B1-5 and 6 like this for 6-8 mo already on EVGA 1300. S7 b1,3,5 has similar 1293 wattage, B2,4,6 even less watts.
EDIT: exactly what @suchmoon said as well ^^^^

where did you get the PCIe split cables. I just want to make sure I get same gauge wire as you! random amazons are bad choices! I did that for my S3's and got like thinner than 22 gauge or something. almost fried them. rewired them with 18 gauge.
 so using a 1300watt GOLD how much under clocking? like a 4.7 S7 needs to run at work? 4.5? or something worse?
thanks

10 % downclock      so 4700 = 4230

or do a 15% downclock to 4000gh. 

I run my s-7 at 4000gh on a seasonic 1200 plat.

 batch 2 been running since october  at freq 512 vs 575

note batch 2 uses 162 chips  not 135

  Ugh my stupid brain looks at that and thinks. ok I will be losing 700 so leave an s3 running to make up for it. but I want to get rid of all the s3's lol.
Ugh... why do I have to think things out! hrmm... I could run a 1300 watt gold and use a leftover 750 watt and keep it at 4.7 couldn't I!

[MyRig]

@Xircom,

S9 works on any pool to include P2pool!!! 

https://bitcointalk.org/index.php?topic=1493601.msg15147339#msg15147339

Currently testing an S9 on Eligius if anyone is interested... Will let it go for about a half hour to get some data.

http://eligius.st/~wizkid057/newstats/userstats.php/1NastyFRkeUTmMdbMmzggDVTQA6r3ibUoX


UPDATE:
30-Minute Test Results:








Also testing on the NastyPool p2pool node.  You can see the live chart below (click it if it doesn't update).

[sidehack]

If you want to run it on two PSUs, there's a risk of the controller PSU shutting down and the boards on the other PSU overheating. If I was running a 750 and a 1300, I'd hook the controller and two boards off the 1300 and the third board off the 750. I would then wire one of the fans to run off the 750, except you need to keep the blue and yellow wires from the fan tied into the controller so it still has speed control and RPM readings. But what'll happen is, worst case, your 1300 shuts off and you still have one board making heat but you still have one fan to cool it.

[bspurloc]

If you want to run it on two PSUs, there's a risk of the controller PSU shutting down and the boards on the other PSU overheating. If I was running a 750 and a 1300, I'd hook the controller and two boards off the 1300 and the third board off the 750. I would then wire one of the fans to run off the 750, except you need to keep the blue and yellow wires from the fan tied into the controller so it still has speed control and RPM readings. But what'll happen is, worst case, your 1300 shuts off and you still have one board making heat but you still have one fan to cool it.

  Interesting the blade keeps mining even though the controller board is not communicating with it anymore? are you sure? I thought cgminer issued each blade the work to do and when cgminer ceases to run the boards go idle.
   I was looking at some things, so each board needs 480 watts. so a 1000 watt should be enough to power 2 boards and a 750 to power the other board and controller.

[sidehack]

If you're looking at DC power, each blade needs around 400W and the controller and fans burn another 60 or so.

The chips cache work and cgminer updates it, so if cgminer stops sending work the chips will keep hashing on the last thing they were given. Even if they don't, there's still some power being burned in there while the chips idle (maybe on the order of 1W per chip), and if it's not allowed to escape things will get toasty. One guy on here has reported losing a couple blades due to overheats from that very problem.