Topic: New antminer S-5+ selling options (Read 4165 times)

I'll order a couple of them. If someone is living in Europe and needs a space for hosting them, let me know. Noise is no issue....

Do not say I did not warn you.

And new ?  Where have you been living...  It vas
PACMiC Cloud Mining Contract --    Hashnest 's newest PACMiC Cloud Mining Contract
March 09, 2015, 08:39:24 AM
https://bitcointalksearch.org/topic/hashnest-s-newest-pacmic-cloud-mining-contract-982890

Hashnest PACMiC v2  https://bitcoinnewsmagazine.com/hashnest-pacmic-v2-available-now-for-subscription/

and "NEW": Hashnest PACMiC V3


..........................................................................

This is the miner that should be called the Widow-Maker. xD

I need to get more S5's instead, they're not too bad with the fan control and i'll prolly change the fans for quieter and better cooling yet.

Nice!  The price is pretty good. Not perfect but pretty good.
Like you Phil I can get 1 because of the sound. 

Had a feeling that they were low power.

Well the beast is selling now on bitmaintech website

price is  2375 with 98 to ship to usa or 2473  not bad.

if you have a low power cost  setup  this could do well.


I went with bitmaintech's new cloud offer .  I purchased 10th for 6.66 btc   

I can not justify getting the 7.7th beast.  due to  the noise  my wife would kill me.

OK understood. On my test I just unplugged the PCIe connector to one of the hash boards. I did not  unplug the 18 pin connector to the same board, perhaps I should have?

Just repeated the test with both connectors unplugged. Revise Controller current sligtly up at 0.5A so 6W.

The 4-pin connector is just for power I believe, so I never disconnected that - if it were, then BB wouldn't have any power to do anything (or power fans, etc).  The only things I unplugged were the 18-pin blade connectors (used to control the blades) and the 6-pin PCIe connectors (used to power the blades).  I did disconnect the fan and hook it up to another power supply, just to remove one of the variables from the mix.

That is good looks like we are in the same ball park. I did not unplug the Controller Connector when doing my test. I unplugged the other blade. I sort of assumed that things would not run with it unplgged? Which begs the question what is it for? I had measured 9V on it and assumed it was for a purpose?

EDIT. Assuming we are talking about the 4pin connector between the hash board & the controller I tried unplugging and nothing worked

Rich

I think your methodology sounds correct, so I've stopped being lazy and just run a couple quick tests to compare.  I'm using a Watts Up Pro ES to monitor power at the wall, and I have the fan hooked up to a separate power source - so the only things that are being powered are the BB, controller, and two blades.  All settings are the factory default, PSU is the Corsair HX850i at 110v:

Bassline - 558w: I booted it up and had it mine for a couple minutes like normal, hashing is right within spec.  Keep in mind that the fan is NOT included in this.

One Blade - 275w: Physically unplugged both the controller connector and PCIe power plugs from one blade - software ran like normal, although the hash rate was obviously 1/2 of normal.

No Blades P1 - 14.4w: I left the PCIe power connected to only one blade, but unplugged the controller connect - so no blades are actually connected.  Everything booted up fine, but no hashing.

No Blades P2 - 17.6w: Similar to No Blades P1, but this time I hooked up PCIe power to both blades, but left the controller connectors unplugged on both.  Everything booted up same as with P1.

Zero RPM P1 - 68.3w: This is running with their latest firmware, one blade connected correctly, but without the fan connected so it won't do any hashing.

Zero RPM P2 - 127.1w: Same as Zero RPM P1 but both blades fully connected correctly, but without the fan connected so it won't do any hashing.

Bassline+Fan - 580w: Latest firmware, fan set in software to 100%, both blades connected and hashing correctly.

.... So, now some insights from the numbers - a couple things stand out:

#1, with a platinum PSU (at 110v no less), it's beating the spec and doing 0.502W/GH (at least this particular unit)
#2, the overhead of a blade is 3.2w, regardless of it doing anything
#3, the BB+controller is somewhere between 8w and 11.2w
#4, the overhead of a fan (full speed) is 22w
#5, with the new firmware, when it benches a blade, it still consumes 58.8w

And back to the original comments about system savings, being generous the S5+ saves 89.2w (6 CPU/controllers and one fan) over 7 S5's.

Anyway, hopefully other people find these numbers helpful, or at least interesting...

I measured the controller current by making one measurement with 2 Hash boards connected and then with just 1 hash board connected. The difference is the current that a single hash board is taking. Multiply by 2 and take it from the number you first thought of  & the have the Controller current.

So the controller takes just over 0.4A, so controller Power is only 5W. So unless there is a flaw in my methodology not much of a power saving from sharing the controller.

Rich

Ok that is good. Does anyone know what the additional 2 pins are for, or the additional push button on the 18 Pin controller?

You can use new S5 Hash PCB  ( 18 pin connector)   with S5 Control PCB  (16 pin connectors) Need only a 16 pin cable.
Picture of my water cooling S5 project    3 x old blade + 1x new 18 pin blade.



Click on the image

It has got the 18 pin blade connector so you will need later hash boards (V1.91) with the 18 as opposed to 16 pin connector. I do not know what the additional 2 pins do, or if they are anything to do with the additional button on the controller board, or if a 16 pin blade can be used with a 18 pin controller? Perhaps someone does???

Rich

Excellent points - I'm planning on building one quad-blade machine just to kind of prove things out, plus I have a water cooling kit I haven't built yet, so it's a good excuse. 

If I SSH in and kill CGMiner, then it runs ~17w pretty consistently - it's once CGMiner runs and that it's not hashing that it maintains that pretty constant 50w for several seconds regardless of clock rate, then it kicks into gear and starts really hammering power.  But to your point, it could also be some characteristic of the ASIC's starting up, and not additional load of the FPGA starting up.

I also emailed Bitmain to see whether or not the new controller is compatible with the existing S5 blades - the connectors look the same, so if is, then it might be possible to cannibalize 4 S5's into a single mini S5+ - maybe call it an S5- perhaps?    It really depends on whether the controller ASIC just detects and adapts to the blades, or if it's got specific firmware on it for the newer S5+ blades.  At worse, even if it's not compatible, they'll hopefully sell me the cables which look to be much longer, so no need to hack up my existing ones.

you flux between 17 and 50 with no hashing.  It could stay stable at 17 when hashing.

 But for arguments sake  running 2 s-5's and using 1 controller.
then running 2 s-5's and using both controllers will give the true controller power draw.

17 is meh 50 is not.  

If I run 6 s-5's on 3 controllers I would save 150 watts. which is  about 110 kwatts per month.

But I read a post saying voltage was dropped to the chips on the s-5+. So I suspect your 17 to 50 with 0 hash while accurate has something to do with real numbers while hashing. along with the voltage setting dropped.

If you could run 2 s-5's machine on 1 controller and check watts  say 1150 is the watts.  then run 2 s-5's with  both controllers  and get 1200 watts  then you are correct.

I think you may get 1150 then 1170  my guess is the 17 watts number is more accurate not the 50 when hashing.

no I mean like we get a few guys interested to split 1 or more s5+ miners and the trusted member ships the boards to the people that paid accordingly but trying to get a good price point would make it hard to do. I'm more interested to see if they sell replacement boards for this. I have seen about 3 different listings for the s5+ on ebay and they are in the US.

I looked over some power monitoring data I've been doing on a S5 I have - and when it's not hashing, and when the fan is on another power supply, it looks like it consumes about 50W (it periodically dips down to 17W, but for the most part 50W seems to be when it's doing something with the FPGA, but hashing hasn't begun).

If that's the overhead, then for 7 S5's, it would be 350W total for BB/controller - and without the BB/controller as part of the power calculation the S5 drops from 0.51W/GH to 0.46W/GH (pretty close to 0.44W).  Now 7 S5's is obviously 7 fans and the S5+ only has 6 - the fan at full power consumes 0.20 amps (~24w), so the system savings would be 324W over the same S5 configuration.  Not bad, more than 1/2 the wattage of another S5 by just changing the configuration around a bit..

From this I would say that they've done almost nothing in hardware design to increase efficiency of the S5+ (as opposed to the work Sidehack has done), and instead have made their gains by optimizing the system.

Center section only.   Can't see myself wasting more on + models.  The wait has been long but will need to wait a little longer (biting lip).