Topic: BITMAIN Antminer S3 support and OverClocking thread - page 32. (Read 158144 times)

  if you do it correctly yes.


 if you do it wrong short dead gear

Anyone know if I can power one AntMiner from two different PSUs? I have two AntMiner s3 connected with 4 PCI cables each and over clocked.. I still have room inbetween my two cx750s
Is it possible?

the evga 1300 has 6 cables 4 singles and 2  doubles.  it runs 3 at  freq 237 and 475 g-hash pulls about 1100 watts 0 issues.  

today I decided to try freq 250  I am getting 505 gh and pulling 380 watts each,   but I am doing 2 per unit for now.

I have some more s-3's on order my batch 4 were outstanding .

 my batch 1's were meh.

 I am hoping my batch 6 are as good as my batch 4.

I'm assuming the 1300 has the same setup or better than the 1000 G1. There are 6 pcie cables but 4 of them have single pci connectors on the ends and 2 of them actually have double pcie connectors on the end, effectively giving you 8 pcie per power supply. I have 16 Ants running off of 8 EVGA 1000 G1's, all with all 4 connectors plugged in and they've all been doing great since they arrived.

Even if overclocked to 250Mhz, 1 PCI power cable per blade is just enough.
But its strongly recommended having 2 PCI power cable per blade.
The system will be more relaxed and the wire temperature will be much better.
Cooler wires will give less power loss, even though its not much large.

I'd say the main sense is spreading the power distribution over two pcb tracks instead of just one.

Ok, see this seems to be going against earlier theories in the thread, but this at least makes sense. With a power draw of 400w or less, I would think we could use the same power setup we had for overclocked s1's. I.E. only TWO pcie cables. But other people have STRONGLY recommended using 4 pcie cables when overclocking the s3's. The reason they give is decreasing the power draw on the pcie cables. Makes sense, but this implies we have been near melting our pcie cables with the s1's for the past couple of months. What gives ?

1) yes - but generally it would mean using a server power supply which is louder and less friendly to use in a small setup. THE EVGA 1300 is a great unit at a good price
2) yes, even at ~380W each you are left with over 150W of overhead capacity on the 12V rail
3) either get a good quality splitter or just use 2 connections per unit. AFAIK using all 4 connectors doesnt provide much realworld benefit but might be slightly safer for the pcb traces
4) closer to 380-400W

I'm thinking of buying 3 or 4 Antminer S3, but right now I'm puzzled about getting proper power supplies for these.

• Such EVGA 1300 G2 is quite pricey, are there any cheaper alternatives?
• Can 3 Antminer S3 be powered by a single EVGA 1300 G2 without problem?
• When overclocking 3 or 4 Antminer S3, how do you solve the power supply issue with 4 power connections each S3 (since a EVGA 1300 G2 only comes with 6 cables)?
• What's the power draw of an Antminer S3 when running at 250 MHz?
  

Disable DHCP and set the IPs manually is the solution.
If not working Reset with the small button in the front side of your antminer S3.

I setup 4 S3s in my office room for testing and configuration. They were all hashing away fine after the test so unplugged them to relocate to more permanent location in the adjacent room on the same network and wired router. My router obviously reassigned each one a new ip, but i just looked a the DHCP list to find them and typed in the new ips and three of the four logged in just fine. My problem is the fourth will not log in. The password screen comes up and i type "root" and then it goes to this  screen:

/usr/lib/lua/luci/dispatcher.lua:448: Failed to execute function dispatcher target for entry '/'.
The called action terminated with an exception:
/usr/lib/lua/luci/sauth.lua:87: Session data invalid!
stack traceback:
   [C]: in function 'assert'
   /usr/lib/lua/luci/dispatcher.lua:448: in function 'dispatch'
   /usr/lib/lua/luci/dispatcher.lua:195: in function



And also now when i type 190.168.1.99 for the setup after the password screen it just goes to Page not Found . The fourth one is still hashing as normal as reported by my pool. Im afraid to reboot again since everything is hashing normally, but would like to be able to access all miners for future OC. Can anyone offer some suggestions?  Or tell me what other info you may need me to provide.

I have already purchased one, the warranty is 7 years. I'll try it on monday, I hope it works well.

Ummmm....thanks?.  You made my brains hurt!  Is that even....I don't know.  Apparently I've got some studying to do.  I better just read the manual from the beginning.  I'll see what I can figure out.

Wow. It's beatiful.  
Should be OK, but it looks like chock-a-block PSU. You better try to find 1400+ Wt PSU, just to be safe. What about warranty period for that Lightning?

I started with the bit layout for freq_value and from there you can calculate the chip_freq.

Bit layout for the freq_value is as follow:



OD can have values between 0 and 3
R can have values between 0 and 31
F can have values between 0 and 127
BS can have only 0 or 1

From page 6 of BM1382 spec (https://www.bitmaintech.com/files/download/BM1382_Datasheet_v3.0.pdf) you get:

NR = R + 1
NF = 2 * (F + 1)
NO value from OD value (see Table 4 from same spec page)

Fref = Fin / NR
Fvco = Fout * NO
Fout = Fin * NF / (NR*NO)

So Fout is "chip_freq" and can be calculated based on valued from above ("OD", "R" and "F"). BS select "high-band" or "low-band" (see page 7 from BM1382 spec for valid freq values for each band).

chip_freq = 50 * ( F + 1 ) / ( ( R + 1 ) * POWER( 2 , OD ) )

For timeout I used following formula:
timeout = 4000 / chip_freq

I suggest to calculate all 3 Fref, Fvco and Fout and make sure all 3 value are in range described on page 7 for the band specified by "BS" value.

EDIT: Here is an example:

freq_value = 1306

OD = 2
R = 1
F = 38

You can compute the rest as following:

NR = 2
NF = 78
NO = 4

Fref = 12.5
Fvco = 975
Fout = 243.75

BS = 0 and all 3 values for Fref, Fvco and Fout are in range for Low-Band.

i'll trade you for mine that doesn't go above ~400GH   

but if you're worried about 10GH just restart cgminer.

Please help, can I use a Rosewill LIGHTNING 1300W for three overclocked antminer s3s?

I'm sorry if this was already covered...so Ill keep it brief.  I have an AntMiner S3 that's a rockstar.  I overclocked it to 250 MHz, and it was hashing right along at 505-506 GH/s, even after 2-3 hours.  Temps were great 39C-41C, absolutely no X's on my chips.  This morning, I wake up...take a quick peek before heading off to work and it was sitting at 490GH/s, and it was still sitting at 490GH/s (rock steady) when I got home from work 10 hours later.  After work I calculated the HW errors and i'm sitting at 0.21%(pretty good unless I missed the boat).  Now I know that's nothing to complain about...and I'm not trying to.  I'm just wondering if anyone else has experienced a similar situation and has any kind of explanation or trick to get it back above 500.  Thanks.

PSUs are less efficient and will burn out sooner if you run them at full load.  My advise is to get yourself a gold or platinum certified PSU around 1000-1200w.  Overclocked at full tilt (350 MHz) you will pull apx 400w from the wall per miner...so your looking at ~800w for the pair. The conversion from 120V AC to 12V DC isn't perfectly efficient, so some energy is lost as heat(why PSUs have fans).  If you have a decent gold/plat PSU your looking at 87-92% efficient...so if your pulling 800w from the wall your miners are only consuming 696-736w...the other 104-64w are lost as heat. The less efficient PSUs (uncertified, bronze, silver) put out more heat, as the conversion is less efficient.  You also want to have some headroom on the PSU so your not pushing it at the max 24/7( = burnout & replacement costs).  I only use gold/platinum certified Corsair PSUs (wish I got $ for the free advertising) because they are pretty much tanks.  I'm running 4 Corsair RM1000 PSUs (gold cert.) and they have a 5 year warranty if I remember correctly.  They worked great on the S1s (I had 6 S1s overclocked on 3 PSUs...2 each), and the S3s.  You can do your own research, and this isn't exact ,but 80% of the rated load is normally approaching the point where efficiency starts to decline because of the heavy load/heat buildup.  You could also run an additional single blade(half an S3) if your not overclocking any of them...overclocked that would be using 1000w and would be incredably hard on any PSU.  My advise is pay the extra for a reliable strong PSU.  You also pay a little more for fully modular, but if you have alot of gear, it keeps clutter to a minimum....may not make a difference to you if your only going to ever have 2 miners, then modular doesn't matter so much.  Just remember...additional cables in the way reduces airflow and increases temps...maybe not to the point of making an impact...the rule holds true to life (ie...computer case cable management).  Good luck and happy mining.