Topic: ANTMINER S3+ Discussion and Support Thread - page 135. (Read 710164 times)

It does have sense, but I'm not good enough in english to explain.
1/ Cold air and warm air don't use same space because of it's thermal expansion.
2/ the S3 is not full close, air can flow outside by many ways, not only by the pull fan.

Because of those two points, input air flow may be different from output air flow.

Doesn't make sense in a push/pull config to have one fan running at a different speed than the other.

I updated to the latest firmware on one of my S3's that is on leaserigs and now I can't lease it out.  Even after editing the files and allowing the api access and whatnot....all that happens is cgminer just restarts over and over again.  Went back to previous firmware (s3+ support) and all is fine. 

This is also the underperforming unit that barely holds 420ghs at stock speeds.

Anyone else experience this? 

Sometimes I can't communicate with one of my S3+ either through a browser and the GUI or through SSH. Even though I can't access it, it continues hashing according to the pool stats.

Has anyone seem this problem before?
Could it be a defective controller card?
It's still under warranty and I could try switching cards.
If so, will Bitmain send a new card or want the whole unit back?

I'm getting the following error on the config page but the unit is hashing...

This page contains the following errors:
error on line 348 at column 34: EntityRef: expecting ';'
Below is a rendering of the page up to the first error.


Not sure what to make of it.

Anyone is paying attention to unknown IPs by the antminer in addition to IP of pools setup on the miner.
I am not sure if this is related to this firmware upgrade but in my case there are one or two unknown ips connected at times that I cannot figure out.

Will this PSU be able to handle 2 S3+ ?

I just started S3 mining and went though similar PSU choice issues, in the end i went with 1:1      1 psu per 1 S3.

To give me some flexibility I went with a FSP Raider 750 Watt 80 plus Silver PSU, it has 4 6/8 pin pcie connectors so I can fill all the S3 power connectors,
apparently this allows you to overclock and keeps the unit slightly cooler per given freq.

Secondly, it has a common 12 V rail giving 60 amps which is more than enough for an S3 and allows the PSU to run at around 50% capacity which has an efficiency over 90%.

One thing I have noticed with some Gold modular psu's is they have multiple 12 V rails and may add up to 700-800 watts but only deliver 18 amps (per rail),
 if you power you S3 with one connector per side from a rail like this you are going to have trouble as an S3 needs 24-30 amps depending on who you read.

With lots of PSU's it does chew up more space and power points but you don't have a single point of failure and your not running your PSU at 90-100 % capacity.

Good luck with your PSU hunting.

I am still testing it as the chip sometimes takes too much time to die. Is there a way to disable it completely? It dies after some time at any clock so it's definitely not a good chip.
I was using a PC Power & Cooling Silencer Mk III 1200W and just changed it with Corsair RM850. No difference at all. Both are single rail PSUs.
Anyways.. Seems like I got one of the not so good units.

If you are located within North American Regioin, Please return this S3 to the following address with a tracking number and PM your shipping address, phone number and the e-mail address in the original shipping box (if possible)

Bitmain Warranty
3700 Quebec St. #100239
Denver, Colorado 80207

Phone: 844-248-6246

Thank you!!

Anyone is paying attention to unknown IPs by the antminer in addition to IP of pools setup on the miner.
I am not sure if this is related to this firmware upgrade but in my case there are one or two unknown ips connected at times that I cannot figure out.

So I did a little experiment with one of my Antminer S3+ units (One of the better performing ones that is able to hold any frequency setting for long durations without any 'Xs.' I essentially tested out each of the frequency settings and compared them to the wattage numbers used at that setting with 110v power, then tested again at 230v.  The results of my testing are below:



In essence, what I found was somewhat obvious...that you get better Joules/Gh at 230V than you do at 110V, regardless of the frequency the miner is set to.  I also found that the 'sweet spot' of energy used per GH/s is around the 200 MHz mark (200 MHz @ 230v and 206.25 MHz @ 110v.)  I noted that in green, which incidently was the only reading that matched the 'published' efficiency numbers.  A few notes about my testing:

1. I left each of them at the specific MHz frequency for exactly 3 hours, then recorded the Wattage and Average hashrate at that time. I was off by a few minutes sometimes, but suffice it to say I left them for at least three hours. I did note that this number is more accurate the longer you wait...however, I didn't have so much time to play with this, so there may be some variation that I couldn't account for or the numbers may change more significantly over time.

2. I used the same Antminer for every setting, the same PS (Corsair CX750 with all four PCI-E cables plugged in) and the same Kil-a-watt unit for each test.  I was using the 8/11 firmware on the unit with the updated cgminer version (4.6.0-140908) posted by ckolivas.

3. The one major variable I couldn't control for was ambient air temperature, which was generally about 5C warmer in the 230V location than in the 110V location, plus during the day there were fluctuations. This likely would affect efficiency, though probably not by too great of a deal.

4. I used three variables which you don't see here...the price per kilowatt/hour ($0.10,) the current price of BTC in USD ($351.59,) and the amount of BTC/Gh/day based on difficulty (0.00001398 currently)

5. I included the power usage of the unit above when it is not hashing (noted as 0) just to note the overhead that the system takes up if it is on but not running.

Anyway, hope you can get some use out of this...at a minimum it taught me that it's not just a myth that 230V is more cost-efficient than 110V, but when it comes down to it, it's only more efficient by 1 or 2 cents a day...