ANTMINER S3+ Discussion and Support Thread

[ToQcHista]

i should will use paperclip  

i dont think you are understanding where to put the paperclip.. it is described on the first page how to set these up.. you dont have to make it so long tho.. i just used about 10mm of wire.


c.  As for the PC power supply, you will need 1 jumper wire to short the Power-Supply- On (active low) pin-Green to GND-Black to switch power supply ON, disconnect from GND to switch OFF.

why is better just wire than paperclip ?


most easy is paperclip

No one cares what you use
It's low current
Roll up a piece of foil if you want. Stop over complicating this and trying to be a pain in the ass.
For that matter you can cut the wire's off the back of the plug and twist them together.

 i don't should will touch paperclip .

[adaseb]

so one of my blades dropped off.. i went to checked and found this


the s3 wasnt overclocked, im running 2 on a Seasonic X850 with 4 PCI-E cables with 2 ends each so all 8 are filled.. the main ones are at the front and the 2nd ones at the back..so im not running 2 blades on 1 cord.

i switched it out to one of those molex adapters for now.. and i put them on 200M.
but i dunno what caused that to melt.. it shouldnt have been overloaded.

i am fortunate because this psu has 6 8-pin pci-e spots and 5 6-pin molex.
so the melted one can be avoided.

I don't think you overloaded the PSU. I think there is something wrong with your blade, and it caused a short. If you reconnect that blade, it might melt your PSU again.

[aarons6]

I don't think you overloaded the PSU. I think there is something wrong with your blade, and it caused a short. If you reconnect that blade, it might melt your PSU again.

doing some more research i think the PCI-E port was overloaded..
i have 4 pci-e cables with 2 ends each, i had one on each blade connected to both ports... so all 8 were plugged in.
i had the s3s at default speeds. so they should be around 175 watts per blade.

PCI-E specs is 75 watts per 6 pin plug (150 for the pair), or 150 watts if 8 pin plugs are used (300 for the pair) on the same cord..

i think bitmain should have used the 2 extra grounds and used the 8 port pci-e cords instead of 6 port ones.

since i had each blade connected to 1 pci-e cord with both ends in the ports i think they were being over loaded.. or at least very close to spec.


or maybe something got on the contacts and it didnt have a good connection.. i got the psu used so that plug could have been damaged, who knows..
i checked the other plugs and they show no signs of heating up.



i went ahead and ordered 2 more of those 2 molex to pci-e adapters and im going to run 4 pci-e cords and 4 molex to pci-e adapters on separate plugs on the psu.

that is probably the best way to keep from melting wires.

[soy]

so one of my blades dropped off.. i went to checked and found this


the s3 wasnt overclocked, im running 2 on a Seasonic X850 with 4 PCI-E cables with 2 ends each so all 8 are filled.. the main ones are at the front and the 2nd ones at the back..so im not running 2 blades on 1 cord.

i switched it out to one of those molex adapters for now.. and i put them on 200M.
but i dunno what caused that to melt.. it shouldnt have been overloaded.

i am fortunate because this psu has 6 8-pin pci-e spots and 5 6-pin molex.
so the melted one can be avoided.

I don't think you overloaded the PSU. I think there is something wrong with your blade, and it caused a short. If you reconnect that blade, it might melt your PSU again.

Pull the ribbon cables from the controller board and measure the resistance in both directions on each PCIE connection.  If the one that burned the connector shows different, it's suspect.

[soy]

S3#2 on which I re-flowed solder at the PCIE connectors and on the inside top of hashboard2/chain2, has been coming up with an x at Chain2#8.  That was one, among others, that had been coming up x before the work.  Now the S3, after running a day, will settle out at 426.xx with that ASIC x'd out.  This has happened a couple of times.  That difference from 441 is just about 1/32nd or what 1 ASIC puts out.

Someone said they had luck putting thermal conductive tape on all ASIC tops.  Noting his work, I took conductive tape from a TO-220 and cut it down, put it on the single ASIC showing an x, and have just put it back up to run.  It remains to be seen if the tape on that one ASIC, giving it a little more pressure than the other ASICs, pressing its underside harder onto the groundplane (if not well soldered under there) and tighter to the heatsink, will keep that ASIC from x'ing out.

[aarons6]

so one of my blades dropped off.. i went to checked and found this


the s3 wasnt overclocked, im running 2 on a Seasonic X850 with 4 PCI-E cables with 2 ends each so all 8 are filled.. the main ones are at the front and the 2nd ones at the back..so im not running 2 blades on 1 cord.

i switched it out to one of those molex adapters for now.. and i put them on 200M.
but i dunno what caused that to melt.. it shouldnt have been overloaded.

i am fortunate because this psu has 6 8-pin pci-e spots and 5 6-pin molex.
so the melted one can be avoided.

I don't think you overloaded the PSU. I think there is something wrong with your blade, and it caused a short. If you reconnect that blade, it might melt your PSU again.

Pull the ribbon cables from the controller board and measure the resistance in both directions on each PCIE connection.  If the one that burned the connector shows different, it's suspect.

i just replaced my power supply with this
http://www.rosewill.com/products/1799/ProductDetail_Overview.htm
that PSU came with 4 dedicated PCI-E cables and 2 dual shared ones.. so i put 1 on each and used 2 molex adapters with their own molex cable..

so all 8 ports have their own wire coming from the psu.

the S3s are happy, i even put them back up to the default S3+ speed..

[ToQcHista]

what is ?

your see this post .

https://bitcointalk.org/index.php?topic=1021200.0

[soy]

S3#2 on which I re-flowed solder at the PCIE connectors and on the inside top of hashboard2/chain2, has been coming up with an x at Chain2#8.  That was one, among others, that had been coming up x before the work.  Now the S3, after running a day, will settle out at 426.xx with that ASIC x'd out.  This has happened a couple of times.  That difference from 441 is just about 1/32nd or what 1 ASIC puts out.

Someone said they had luck putting thermal conductive tape on all ASIC tops.  Noting his work, I took conductive tape from a TO-220 and cut it down, put it on the single ASIC showing an x, and have just put it back up to run.  It remains to be seen if the tape on that one ASIC, giving it a little more pressure than the other ASICs, pressing its underside harder onto the groundplane (if not well soldered under there) and tighter to the heatsink, will keep that ASIC from x'ing out.

This isn't working.  After 2days it was down to 428GH/s so I rebooted and 9½ hrs later it's down to 424GH/s.  What I notice now is that although Chain2#8 isn't x'd, it's a degree hotter than chain1 but the fan is running slower - chain1/2400/42, chain2/2160/43.  

So, I tried to add a variable speed control to the pull fan and dial in same RPM's.  Not only didn't it work but now I'm not getting the fan to run with a simple SPDT switch between 12v & standard out.  Either my wiring is wrong or I fried something in the fan.  The PWM controller has pos & neg for both IN and Mtr.  I had made neg common for both.  If the PWM pulls motor in below ground, e.g. intending the motor to be isolated from IN, then there may be a problem.

------------------

No, I had put a strain on the fan wires and the connector to the hashboard, under the cover, had partly come free.  Interestingly, when the S3 starts up the ground to the fan is disconnected.  I know this because while the red led is blinking I can flip the switch from 12v (from the supply) to the internal fan speed output, the fan doesn't turn in either position.  When the board starts hashing both fans start and the toggle from dictated speed to full speed works properly.

It may be that the PWM motor controller can work and that the internal plug had been partly separated already but I don't think so as the fan ran full blast ignoring my dialing down the rheostat on the PWM motor control.

After 8 minutes the chain2 fan was at 2220.  I flipped the switch and the fan is now, well it was 4200 but has dropped to 3900.

[aarons6]

S3#2 on which I re-flowed solder at the PCIE connectors and on the inside top of hashboard2/chain2, has been coming up with an x at Chain2#8.  That was one, among others, that had been coming up x before the work.  Now the S3, after running a day, will settle out at 426.xx with that ASIC x'd out.  This has happened a couple of times.  That difference from 441 is just about 1/32nd or what 1 ASIC puts out.

Someone said they had luck putting thermal conductive tape on all ASIC tops.  Noting his work, I took conductive tape from a TO-220 and cut it down, put it on the single ASIC showing an x, and have just put it back up to run.  It remains to be seen if the tape on that one ASIC, giving it a little more pressure than the other ASICs, pressing its underside harder onto the groundplane (if not well soldered under there) and tighter to the heatsink, will keep that ASIC from x'ing out.

This isn't working.  After 2days it was down to 428GH/s so I rebooted and 9½ hrs later it's down to 424GH/s.  What I notice now is that although Chain2#8 isn't x'd, it's a degree hotter than chain1 but the fan is running slower - chain1/2400/42, chain2/2160/43.  

So, I tried to add a variable speed control to the pull fan and dial in same RPM's.  Not only didn't it work but now I'm not getting the fan to run with a simple SPDT switch between 12v & standard out.  Either my wiring is wrong or I fried something in the fan.  The PWM controller has pos & neg for both IN and Mtr.  I had made neg common for both.  If the PWM pulls motor in below ground, e.g. intending the motor to be isolated from IN, then there may be a problem.

------------------

No, I had put a strain on the fan wires and the connector to the hashboard, under the cover, had partly come free.  Interestingly, when the S3 starts up the ground to the fan is disconnected.  I know this because while the red led is blinking I can flip the switch from 12v (from the supply) to the internal fan speed output, the fan doesn't turn in either position.  When the board starts hashing both fans start and the toggle from dictated speed to full speed works properly.

It may be that the PWM motor controller can work and that the internal plug had been partly separated already but I don't think so as the fan ran full blast ignoring my dialing down the rheostat on the PWM motor control.

After 8 minutes the chain2 fan was at 2220.  I flipped the switch and the fan is now, well it was 4200 but has dropped to 3900.

1 deg is normal, as well as the front fan spinning faster then the rear..
my s3s usually sit around 40/38 with the fans at 1740/1440

bitmain says right on the first page that there is a 5% variance on the s3s..
so 440-5% is 418GHS.. you are above that so you are OK..
really tho the difference is just pennies per month.. you are losing more BTC playing around with it.. and possibly burning out the PWM circuitry.

[ToQcHista]

i will buy it pSU for s3  > http://www.fsp-europe.com/professional/fsp700_50arn_88plus.php

[ToQcHista]

i will from where should guess that s3 working ? Cooler to keep running?


else. what time interval turn on GREEN led ?

[soy]

S3#2 on which I re-flowed solder at the PCIE connectors and on the inside top of hashboard2/chain2, has been coming up with an x at Chain2#8.  That was one, among others, that had been coming up x before the work.  Now the S3, after running a day, will settle out at 426.xx with that ASIC x'd out.  This has happened a couple of times.  That difference from 441 is just about 1/32nd or what 1 ASIC puts out.

Someone said they had luck putting thermal conductive tape on all ASIC tops.  Noting his work, I took conductive tape from a TO-220 and cut it down, put it on the single ASIC showing an x, and have just put it back up to run.  It remains to be seen if the tape on that one ASIC, giving it a little more pressure than the other ASICs, pressing its underside harder onto the groundplane (if not well soldered under there) and tighter to the heatsink, will keep that ASIC from x'ing out.

This isn't working.  After 2days it was down to 428GH/s so I rebooted and 9½ hrs later it's down to 424GH/s.  What I notice now is that although Chain2#8 isn't x'd, it's a degree hotter than chain1 but the fan is running slower - chain1/2400/42, chain2/2160/43.  

So, I tried to add a variable speed control to the pull fan and dial in same RPM's.  Not only didn't it work but now I'm not getting the fan to run with a simple SPDT switch between 12v & standard out.  Either my wiring is wrong or I fried something in the fan.  The PWM controller has pos & neg for both IN and Mtr.  I had made neg common for both.  If the PWM pulls motor in below ground, e.g. intending the motor to be isolated from IN, then there may be a problem.

------------------

No, I had put a strain on the fan wires and the connector to the hashboard, under the cover, had partly come free.  Interestingly, when the S3 starts up the ground to the fan is disconnected.  I know this because while the red led is blinking I can flip the switch from 12v (from the supply) to the internal fan speed output, the fan doesn't turn in either position.  When the board starts hashing both fans start and the toggle from dictated speed to full speed works properly.

It may be that the PWM motor controller can work and that the internal plug had been partly separated already but I don't think so as the fan ran full blast ignoring my dialing down the rheostat on the PWM motor control.

After 8 minutes the chain2 fan was at 2220.  I flipped the switch and the fan is now, well it was 4200 but has dropped to 3900.

1 deg is normal, as well as the front fan spinning faster then the rear..
my s3s usually sit around 40/38 with the fans at 1740/1440

bitmain says right on the first page that there is a 5% variance on the s3s..
so 440-5% is 418GHS.. you are above that so you are OK..
really tho the difference is just pennies per month.. you are losing more BTC playing around with it.. and possibly burning out the PWM circuitry.

I try to ignore the 'don't bother it loses more by not mining' advice I've gotten not infrequently but then I spent years in a proto electronics labs and probably view it differently as a result.

I've sometimes considered trying the cook it in an oven electronics reflow solution.  I have used that approach on a laptop.

Now I see this promising scheme:  http://martindrapeau.tumblr.com/post/116344427765/how-to-fix-iphone-4s-wi-fi-grayed-out-with-a  

Not going to try it right away tho.

[sloopy]

In your opinion what is an acceptable amount of HW error on an S3 or S3+?
I understand there are many variables, so I think it is fair to also ask people to include the time, so .001 in 24 hours is fine, but would .01 in 24 hours also be fine if it added the same amount of errors every 24 hours?

What is your acceptable amount of HW errors in what time?

Edit:
And in searching I see comparisons between discard rate versus accepted.
What do you run in that department?

[aarons6]

In your opinion what is an acceptable amount of HW error on an S3 or S3+?
I understand there are many variables, so I think it is fair to also ask people to include the time, so .001 in 24 hours is fine, but would .01 in 24 hours also be fine if it added the same amount of errors every 24 hours?

What is your acceptable amount of HW errors in what time?

Edit:
And in searching I see comparisons between discard rate versus accepted.
What do you run in that department?

it was stated awhile ago the discarded section is not a valid number nor should it be a concern.
my s3s usually have under a .0005% HWE. i dont overclock them tho.. a lot of times they start out with a few HWE, then go days without getting any more.

[Mikestang]

In your opinion what is an acceptable amount of HW error on an S3 or S3+?
I understand there are many variables, so I think it is fair to also ask people to include the time, so .001 in 24 hours is fine, but would .01 in 24 hours also be fine if it added the same amount of errors every 24 hours?

What is your acceptable amount of HW errors in what time?

Edit:
And in searching I see comparisons between discard rate versus accepted.
What do you run in that department?

Anything less than 1% is generally acceptable.  My S3+ at 250/0750 error rate varies from 0.004% to about 0.1%.

[Real-Duke]

Are you going to release a new firmware for S3 & S3+ like S4 and S5 yesterday ?

[luthermarcus]

Ok I have a 1200 gold psu and when i  connect the fourth pin to the top of my S3+ it wont start up. I want to mess around with the overclock on the ant miner but cant get the up and running with the 4 6 pins. It will run with two in doesn't matter which config.  . Anyone else with this issue?

[daddyfatsax]

Ok I have a 1200 gold psu and when i  connect the fourth pin to the top of my S3+ it wont start up. I want to mess around with the overclock on the ant miner but cant get the up and running with the 4 6 pins. It will run with two in doesn't matter which config.  . Anyone else with this issue?

You only need 2 PCIe for running on normal or OC.

[luthermarcus]

Ok I have a 1200 gold psu and when i  connect the fourth pin to the top of my S3+ it wont start up. I want to mess around with the overclock on the ant miner but cant get the up and running with the 4 6 pins. It will run with two in doesn't matter which config.  . Anyone else with this issue?

You only need 2 PCIe for running on normal or OC.

I do have it overclocked right now but i was thinking that all four would push me to 500 currently getting 480 avg @ 237.5M which from what i read has a potential of 500 with depending on your unit. But i dont understand why i can't plug in the fourth pin. It's just baffling. Thanks for the quick reply i was hoping this thread wasn't dead.

[Mikestang]

You should get 500 if you bump the freq to 250 and set the voltage to 0750.  480 sounds about right for your freq setting.