Topic: ANTMINER S7 is available at bitmaintech.com with 4.86TH/s, 0.25J/GH - page 411. (Read 527634 times)

My usual spreadsheet, people have said it's been useful in the past.

Price doesn't include shipping, import taxes if applicable or PSU's. Assumes static diff and BTC price, it's just to give a rough idea

Mining still dead in the UK at $0.23, and shipping to a host is probably going to be a year to breakeven when the inevitable diff rises happen so no new machines for me

They have shown us three points on the curve with :-

0.71V 0.263W/GH
0.66V 0.219W/GH (S7 / 18 Chip Chain with 12V supply)
0.60V 0.181W/GH (S7 / 18 Chip Chain with 10.8V supply)

The S5 with BM1384 bottoms out at 0.6V. I do not know if being full custom opens the way to a lower core voltage but my guess as both BM1384 & BM1385 are 28nM is that BM1385 will also bottom out at 0.6V?

Rich

yeah I think I will. but for now I stopped auto buy.

I need to focus on this s-7

So far, so good. Looking like I will get more BTC out than I put in at the completion of the contract.

do you think we'll make btc on pacmic 3?.
I bought 5TH on day one. So far so good.

good, but it doesn't actually answer my question re using 2 cables/board instead of three. EVGA 1300 has only 8 connectors and you need one of them for controller, so that leaves 6 left for three boards in S7-will it work? I guess that we will have to wait until someone does it first (my money is on Philip since he was always a fan of evga).

I'm more interested to see BM1385's voltage curve as we didn't get one this time. It doesn't feel like .25 is the limit and like the S3 to the S4 and the S5 to the S4+ we could see even high efficiencies on a theoretical S8.

Very tempted to purchase one of these units, but this round, I think I'm going to purchase $1,823 USD worth of BTC on Coinbase when these go on sale instead.

Have ~22THs wrapped up in PACMiC v3 anyway. I would be very interested to know the underclock potential of these if they were running @ ~1kW each.

Congrats on launching this, Bitmain ! Very impressive specs.

I'm using an EVGA G2 gold 1300 now with two S5s and it's running at roughly 93/94% efficiency.  Per my 'at the wall' meter.  all red VGA cables used, vertical strato setup with a box fan.  I just checked this.  Was running at like 97% efficiency when only was running one , but lot less load.

Only if you can run each board on two instead of three connectors, contrary to bitmains instructions.
However, they said to use two connectors per board for S5 and some people used just one successfully (with a good quality connector such that found in EVGA).

I'll be interested to see if coupons materialize. My bet is that they won't. I think Bitmaintech is a different animal than they were 12-18 months ago. Maybe coupons if the S7 doesn't run at spec, but I'll bet that's very unlikely, since I expect they already have plenty of experience running them in-house (e.g. Hashnest).

If anyone wants to run a few of these machines, but doesn't have the capacity, we at http://toom.im can take some. We can provide a server-grade PSU with 16 awg leads for these for $80 each.

Indeed. Some of the more common official peak ratings are in my PSU guide, although for 24/7 its still going to officially be the stated rating. That being said, I've ran many corsair PSUs past their limits in challenging conditions and they've done fine.

 No, there are quite a few PS that will handle a S7 from the COMMON 15A 110V socket. The real issue is the number of PCI-E connectors, not the power draw.


 1210 watts quoted "at the wall" - an EVGA 1300 would have about 10% power reserve, even my favorite Seasonic X1250s have some power in reserve.


 10. one for the controller board. I don't know of any PS that has 9 connectors though, 8 and 10 yes.

 Or you could do what I'm tenatively planning to do, run 2 S7 on 3 PS (though I might finally break down and buy some used server PS/breakout boards instead, since I figure I'll be running THESE units for quite a while).

 They already upgraded their farm to S7s or more likely "S7+".
 That's where all those "used S5" came FROM, outdated units being upgraded.

 I don't see PACMiC v3 being unprofitable - yet. Perhaps buying on in 2-3 months might be, but not right now.

 Unless bitcoin price goes up somewhat faster than difficulty, I'd say you're correct.




 Half right.
 I've been saying the S5+ was overpriced since it first showed up.
 PACMiC though still should be profitable for a while, just don't believe the 25% APR claim as that doesn't appear to take difficulty increase into account at ALL.

 A little over twice the TH I was anticipating, a bit LESS than twice the price. I just wish they didn't make the "per unit" such a HIGH bar.




 My usual ROI calculation shows this will be very close to ROI at the halfing, presuming aquired in time to have it hashing on 30 September - with 6.7 cent/KWH power.
 It will also be SOLIDLY profitable after the halfing for another year or so, presuming my estimate of average difficulty increase is on the high side or very close on the low side.
 I'm estimating a HIGHER rate of increase than the past year to date has shown, BTW.

 10 PCI-E connections is an issue for most if not ALL 1250/1300 watt ATX power supplies. IMO a BAD design call there, or a blatant attempt to get more sales of their own 1600 watt unit. On the other hand, I've been running "split" PS for a while anyway, just have to switch to running 3 PS per 2 units instead of 2 PS per 3 units. 8-O

 Nothing new about the power up order - the S5 wants the hash boards to be powered up before the controller, or at the same time, already - or the hash board that is powered up late just don't hash most of the time. Doesn't seem to be highly timing critical, there's at least a couple seconds of leeway in there as the controller board boots up.



 Coupons were promised - but at the large PRICE of one of these units, I hope the coupons have a LONG expiration date.



 At the given power specs, the entire unit pulls a bit under 94 Amps out of all the +12v combined. Works out to a hair less than 30 amps per hashboard, probably in fact a hair under 30A (controller will soak at LEAST 4 amps JUST for the fans). In theory, you SHOULD be able to get away with 2 connectors per hashboard (24 amps EACH per connector based on the connector specs) *IF THEY ARE GANGED* as each would only pull 15 Amps if you use a pair which is WELL within the power capasity of the connector and even meets the specs for wiring on the common 18AWG most PS use to PCI-E connections.
 If each connector feeds a seperate string, you need all 3 period - but they're only pulling about 10 amps each, so you could get away with running a hash board and the controller off the same wires with 2 connectors, and might be able to run 2 hashboards off a doubled connector per wire set.
 16 AWG, don't sweat the wires at all no matter how you connect a doubled connector set of wires.


 

You can use molex2pci-e adaptors, but it will be a must to connect 3 pci-e each hashboard... I don't think i can make roi with that...

It says if you use a separate PSU then power it on last. You don't have to use a separate PSU.

Having 10 PCIe connectors might be a problem for those who use ATX PSUs. I think the venerable EVGA 1300 has only 8.

Interesting a little more power then I thought it would have.  We knew it was coming though.

Looks like we will be having a lot of people watching the website now.