Topic: ANTMINER S3+ Discussion and Support Thread - page 164. (Read 710164 times)

I feel if anything this is good news for BitMain, which is why I posted it.

We run roughly 100 S3+s.

Strato

I appreciate the info.  Won't hijack the S3 thread to dwell on those (see a series of shift up characters from the number row here) Kansas frauds.  Suffice it to say they were not ethical in their dealings.

Is there any way to undervolt the S3?

And what is the point of posting this into the  Antminer S3 Thread ?

Butterfly Labs has been Shut Down by the FTC. Closed.

Just happened. That's one less ASIC Miner Company.

Strato

oooooooo oooooooo
oooooooo ooxooooo

Do you know how to map 'Asic Status' to the electronic chip left/right board and chip.

Is this possible to have the statistics chip by chip?

Having already repasted the heat sinks in the past today's effort was aimed at cleaning the ASIC pins and anything near components that might effect operation like perhaps compound on capacitors.  Since the inside heatsink paste is assumed good, I removed the cover, the fans, the outside heatsinks, the controller board.  I found the ASICs are numbered from 2 to 17 starting at the upper left then to the lower left then to the next lower ASIC then up then across to the next upper ASIC so it finishes with 17 at the upper right.  Putting the miner without fans or controller board upside down outside I used a can of brake cleaner with straw to clean the board and it's inside then used a second can to clean the other board and its inside with a little cleaner left over for the controller board.  Then firing up my air compressor I blew off the boards.  Hoping it left the controller board okay.  Experience showed me compressed air hitting an asicminer blade FPGA could cause it to lose programming.  Now to reassemble....

Note there's a void in the outside heatsink over the LM75A and a channel for the line of capacitors.

...has awakened with all ASICs o's.

Can't ask for better conditions for this to be running right.  63°F outside, 72°F in the room with miners and that's at 20%rh.

Considering right now if I were to do this again I'd put the spring screws back in place without the heatsink before cleaning the board with the brake cleaner so the cleaner doesn't get into the screwholes.  Worth mentioning clumps of thermal paste on the ASIC pins, e.g. if the heatsink had been slid sideways in removal, might be removed CAREFULLY with a wooden toothpick.  The spray brake cleaner gets used up pretty quickly.

At 30 minutes it's doing 440.82GH/s.  I suspect it will continue to be reasonable most of the day unless it warms then drop tonight again to the mid-430's.  Engines turning off in one of the ASICs for some reason and all the cleaning and reapplication of thermal paste won't fix it.  I could be wrong and maybe it will stay good but I doubt it.

...maybe

40 min 444.29(avg)
 1 hr  443.72
 2 hr  441.60
 3 hr  441.22
 4 hr  440.74
 5 hr  439.68
 6 hr  434.89
 rebooted due to low hashrate

If I were still in the game I'd look for an open.  The low fan speed is an indicator of something amiss as well.  A plated thru hole that fails would fit the bill.  Tough to find without a schematic but not impossible.

If it's a charge buildup then a complete shutdown would be more effective than a restart.

I had adjusted the cgminer-avg-monitor script to look for a hashrate of 432 rather than 439 and run the program only on the hour not every 30 minutes as there were too many restarts and changed the script to reboot instead of cgminer restart.  It triggered at 431.

I'll see how this rebooting works out.  There is a 60 second delay added before rebooting.  The loss, as someone has accurately pointed out, in not much in dollars and cents so troubleshooting it isn't high on my list of things to do.  If I do opt to troubleshoot, I have the other board with which to compare.

 18 m   438.35
 31 m  437.63
 1 hr   439.67     Temps dropping
 2 hr   440.24     Room Temp 76°
 2h59m  431.83  Room Temp 75°
 3h rebooted due to low hashrate

 30 m  437.51 75.0°
 1 hr   440.45  75.0°
 2 hr   435.66  73.7°
 3 hr   438.18  73.9°  Lower room temperature does slow the decline.  Utility down to 14.  ~Midnight.

So with the sleep I added to let it cool it takes 2 minutes before it's hashing again.  Booting at 431GH/s, when it's 10GH/s down, the 10GH/s*120s=1200GH's which is like the miner not running for 2.72 seconds.  So, triggering a reboot with delay that takes ~ 2 minutes loses 117 seconds of hashing.

My impression is that every S3 has its own 'personality'.  This is probably related to the consistency (or lack of) in putting these together.  During removal of hashing boards for the S1 to S3 upgrade I noticed how the amount of thermal paste and its distribution is different among different units.  This is just one example where inconsistent workmanship would cause differences in performance.  There may be also external differences such as different power supplies or different air-flow circumstances to the different miners.

Cheers

Well if there is something I could do to fix this. I wouldn't mind the worry (:

You do know that the difference of 32GHs is currently 22 cents per day in earnings? Not really worth the worry. Just let them run.

Need help/info

I'm overclocking both my S3+ miners to 231.25GHz. One is working fine and after 3 days of mining non stop is stable at 465GH/s. But the other one quickly drops it hash rate to about ~433GH/s after 1 day of mining. I've rebooted this one before because "x"'s started showing up. But this time they're all "o". So no chip faults or shutting down.

Is this because the load is to high because of the overclocking or is it because the miner is just underperforming?

^^ Definitely!

My house wiring is solid and my GE breakers are 15A.  They can support up to 4 S3's, but five is too many.  Personally, I think three is the best choice.

Last month I tried using a second 15 amp circuit in my house with a 12 gauge extension cord. The plug on the cord got warm.  At the time I was running two RMA1000's, each powering two S3's.  I then got a Contractor Grade extension cord and used that for the second power supply.  Now the cable stayed cool, but the outlet it plugged into got warm. I did some further investigation.  This entire circuit was daisy chained through a bunch of outlets and light switch boxes.  I could see a 5 volt drop when the miners were running.  Also, my previous suspicious of a "funny smell" on occasion were now continuous.

I called an electrician and got two new 20 amp circuits direct from my panel to two new outlets in the room where my miners reside.  At present these are running 8 S3's through four RMA1000's.  I have checked every connection on the AC and DC side and all the wires.  They are all cool to the touch. I am getting 3.9 TH out of these machines.

Be careful. Lots of homes have marginal wiring that may be just fine running a few 100 watt light bulbs.  This does not mean it will be safe for what amounts to an industrial application.  I spent $150 for the new circuits and view this as a good investment.

The cgminer-avg-monitor restart cgminer at low hashrate is functioning but my S3 isn't cooperating.  I went out to shop and returned after a few hours.  It was restarting cgminer every half hour due to low hashrate until 5PM then passed at 5:30 and at 6PM but  then the hashrate started to drop.  It restarted again for low hashrate and woke up with two x's, one in each chain so no way it was going to pass the next restart. 

I think this might be memory that isn't cleared or chips that aren't reset rather than bad ASICs.  I've tried adding delay, e.g. sleep 60, but it still fails to restart properly.  I know a cold boot works.

What I have to do with my cable box and routers when there's an internet disconnect is use the br command & an X-10 transmitter on a linux machine to cycle cable modem and routers off and wake those up in order with the proper timeouts.  This works efficiently albeit the cycling is a bit noisy and I'd wish X-10 would move up to solid state relays.

Since a cold boot works almost every time, I'd like to use the same X-10 idea with an S3 but there is no ftp and although /etc/services lists ftp, connections are refused.  I don't know if I can just add any repository to whatever angstrom program lists repositories, this without fouling up the system with incorrect version programs.  I'm not even sure there is a br command for angstrom.

This all may be moot as I've purchased additional thermal compound and cleaner so I'm going to address the problem S3 once again tomorrow.  None the less I'd like to know how to get in and out of the S3 via ftp.  Assistance would be appreciated.

Bull.  That would be like storing rubbing alcohol in a vodka bottle.

I received my two S3+'s today. One works, the other makes a ticking noise and I can't connect via LAN. I just sent an e-mail to bitmain.

Anyone had a similar problem?

a lot of new homes are wired with 14 gauge wire and only rated for 15 amps. Don't forget to take in anything else on that same circuit as the lights etc.. I have 12 gauge wire and 20 amp circuits but still only run about 1500 watts. Not worth burning the house down. Also get some watt wall meters and you can see for sure just what your pulling at the wall.

+1. 220V can run safely using 120V outlets/plugs and is often more cost-effective to use than the $20-100 alternative of specialty NEMA 6 equipment and adapters.

The real risk is absent-mindedly plugging something 120V in and overloading it. A coloured outlet with a warning label in the mining farm should be pretty safe from this concern

To each their own.  For me, since I'm the only one that uses that part of the basement, a simple label and or color code is just fine.  Not to mention that everything in that section can work off 220v.  If I were to sell the house, I'd most likely revert the breaker or remove the circuit altogether.