@ Soros Shorts

I suggest a case grommet for your power wires.
The case is thin enough to cut insulation.

You could make one with about 2 inches of 3/8 or 1/4 ID tubing.
Split it lengthwise and install around hole in case.

YMMV

Exactly the plan I am implimenting now.
DPS-2000 to bus bars then wires direct from bus bar to miner.
No more quick disconnect on PSU for me, I'll need a screw driver.

YMMV

Quick update and rough numbers (very rough)

My control board shipped with neptune-rc6 FW
Still a bfgminer checkbox on the 'Mining' web page but it has no effect.
Guess I gotta try and hand roll bfgminer.
They removed the SPI clock and voltage options from the 'Advanced' page
The clock is likely 7-8x faster than default hoopiter cause of 1440 core per ASIC instead of 192
Still 4 die per ASIC, each die now has 360 cores instead of 48

There is an LCD on control board with IP address and hashrate
There is a BRIGHT backlite on the LCD.
(I keep thinking I left my LED flashlight on)
The power on white(really RGB LED) light is brighter still.
If ya lookin at the LCD waiting for it to boot when the white light comes on YIKES!


I tested each box seperately
All ASIC, all die, all VRM checked out.
No idea on individual core health yet.
There was a parade of cores disabling, no idea if it is more or less than 28nm yet
I didn't time the tests ~20-40 minutes per ASIC
I just waited till the numbers looked settled

380-400 Watts @ the wall per ASIC

280-290 Watts per ASIC as reported on 'Advanced' webpage
Roughly 0.6 Watt per GH/s @ the ASIC

I have  a plat PSU runnin @ the sweet spot near 50% of rated current,
should be >10% loss in the PSU the rest of difference is prolly VRMs effiency
(plus a few watts to keep the cable warm)

660-670 GH/s per ASIC with default 475MHz clock
~3.3T stock clock.
(I just added them up, not run 5 at once yet)

HW errors between 0.1% and 0.9% average under 0.5%
I predict improvement with better cooling and volt adjust
(If I had bfgminer I'd have a better idea with individual core statistics)

50-55 degree C ASIC temp with ambient @ 27 C

VRM temps between 70-90 C prolly 76 average

The PCIe cable gets a bit warm, I only tried one ASIC @ 500 MHz for a few minutes, it worked

The cardboard box is much improved, triple walled tough stuff
Inside the box is another triple walled origami nest below and above that is doubled toward the outside of box
3 layers of triple walled cardboard
Slight issue with the nest though, it is exact size of cube but fan grill is external and can (did) get bent in, causing it to sound like a bicycle with playing cards in the spokes for motor noise

I retract earlier statement about case strength
(open mouth, change feet)
The strength relies on fasteners and the cube screw holes strip if ya look at them
(might be a good asset for a stripper but not for a miner case)
IMneverHO
 
She ain't perfect but the ASIC seems solid enough!

YMMV

Can often go a few feet with ribbon cables.
You can probably raise the SPI voltage to compensate up to a point.
If you have longer cables just try it, I doubt it would damage.

YMMV

First impressions before power up.

The controller 'enclosure' is FRUGAL TO A FAULT.
All they did was slap a clear cover over the top.
The bottom is exposed PCB.
The only thing 'elevating' the bare PCB from perhaps a metal table is 4 snap in plastic legs that hold the clear top up.
The sides are open also.

The ASIC case is steel, claim of 'stronger' probably accurate.

Looking inside first cube can see the monkies cranked the heatsink down so tight that the brand new ASIC PCB is low in the center and high at the corners. Brand fuckin new PCBs warped by basic assembly.
I imagine it'll work OK the Oct units got the same abuse and 'most' survived.
The screws holding heatsink plate in place are glued or cross threaded.
I'll need to dig out my long t-handle T10 torx to get em out.

Only 5-7 mm clearance below ASIC PCB (5mm in center below ASIC, 6-7mm at the corners.
Only 8mm between top of heatpipes and inside of case.
This limits adding height (M3 male/M3 female hex standoff) to them, else cut a hole for the pipes to stick out the top.
Raising the PCB is a must do in my book. 6mm clearance starves the bottom of PCB for air.

5 out of 6 of the case screws on the first cube were stripped during assembly.
I'm using a small jewelers style torx handle with only my finger tips (not abused by me)
It looks as if the inside hole is too big for the self tapping screws they used on the thin steel.

About what I expected for fit and lack of finishing touches mechanically.
(I set the bar low)

Time for the smoke test.
If I don't return soon with twisted panties it prolly means ASICs passed basic testing.

It may be a while for more reporting from me as  I have NMI's this side of screen.

YMMV


EDIT # 1 ) The stripped case screws means there ARE metal shavings somewhere that used to be part of the case.

fan = 140mm by 25mm thick  12V 0.4A  4 pin
(4 pin usually means PWM)

YMMV

Each PSU is single  OR  multi rail.
Each rail needs to be capable 300 Watt minimum for each ASIC plugged into that rail.
(based on KnC statements)
For multi rail supplies you gotta dig into the literature for specifics on each rail.

For your 3 x 850 question, I'll assume single rail PSUs.
Use 2 ASIC per PSU and power controller cube from 3rd PSU with one ASIC.
If the 'suits' are to be believed this will load the 850's  to around 600 Watts.

If your three 850W PSUs are single rail and were in one box it would be a 'triple rail 2550W with 850W rails'.

Hope this helps.
Please if possible understand what is going on before applying power.
Internet advice can be faulty even with good intentions!!!

YMMV

LOL, I was'nt claiming issues don't exist, just that the link at the time I read it seemed to indicate PSU's.

Just some friendly pushback from someone who has been known to 'open mouth change feet'
It's all good

Either way I'll know in a few hours if they sealed the smoke in there real good.
My box is in the Chicago area already so scheduled delivery seems likely today.

YMMV

The link you posted seems to indicate an issue with power supplies not ASIC's.
Your statement that they claim 'faulty rigs' is not supported by your link.
Although the issues are related.

YMMV

Sounds like you already isolated the issue.

About all a consumer can do is inspect bad unit visually.
Subdivide and conquor!

Is the ribbon cable installed correctly? (it can be hard to tell on the ASIC end of the ribbon)
Is the heatsink fan spinning?
Is the heatsink still firmly attached to the ASIC?
Has the heatsink been bumped/twisted hard?
The copper heatpipes can become un-attached from the aluminum block near ASIC.
Are the power wires OK? (visually inspect both ends of cable if modular power supply)
REMOVE BLUE EXTENSION CABLES!!!!!!!!!!!!!
(each extension adds 12 possible points of failure)

You could also try swapping ribbon cables since you have a known good one on the working ASIC.
Also try different connectors on the controller board.
 

Hope it helps!

YMMV

@ Phoenix

Most residential AC has 2 phases/legs installed.

All you need to 'make' 220 is a doublewide breaker.
(the phases are staggered and every other breaker is on same leg)
Exception is 3 phase industrial panels where it is every third breaker.
In fact most double wide breakers are 2 single breakers screwed together with the switch levers connected.

If you have a voltmeter and long extension cord you can prove it to yourself.
Just plug in the cord and measure from cord hot to outlet hot.
(hot is small prong/hole in US outlets)
Same phase, hot 2 hot = 0 volts.
Different phase, hot 2 hot = >220 volts.

The main REASON a miner would insist on 220 is efficiency.
A >220 PSU is often 10% more efficient at making 12VDC (depends on model)

I'd guess you have 220 and don't know it.
I don't claim to know what service is installed in Volcano, HI. though.


YMMV

Good idea!
Just watch out for mechanical strain on added connectors as they may only be held in place by solder.

I plan to solder ALL power wires directly to the PCB.
Connectors are a potential point of failure.
12 connectors per PCIe power cable on a modular PSU
(modular sux for performance and safety on power hungry miners)


I have 2 consumer PSU's that have smoked connectors from my Oct. adventures.
This is what happens when ANY one of the pins of a connector fails, the neighbors have to take up the slack.
The 18 gauge wire on consumer PSUs is not up to the task.

YMMV

I took the bait (while you wait) and spit it out.
My 15Kilogram box is in Germany now (headed to Chicago area).

THANK YOU FOR THE HEADS UP IN THIS THREAD!!!!!!

Had I not spit it out I would have traded 1 neptune for under 24 hours of hash while you wait.
It shipped same day as my HWYW would have started.

They make it easy to dislike/distrust them.

Perhaps I neet to get fitted for a hater hat.

On another note!!!
FrozenCPU now takes bitcoin!
They sell water cooling equip.

It's too bad RAM waterblocks are so expensive, 10-15 of them would tame a neptune's VRMs.
(I'll try making my own out of copper tubing)

I waited till shipping notice before installing new electrical load center.
Around USD $150 in parts if you DIY!    


YMMV

I have no idea how far into 'tweaking' you tolerate but,,,

On previous boards if you overtighten the heatsink it causes the board to bow (curve)
You might put a screw driver on the holddown screws and try adjusting while observing the VRM gap.

One must assume that KnC uses unskilled drunken illeagal aliens for assembly labor and plan accordingly.
If the plate is thick enough to hold a bend, it might be bent to match the curve of the board.

Optimally use seperate hold down for main heatsink and contact VRM heatsink directly on VRM's.
(IE: buy standard heatsink holddown for Intel LGA115x)
(Assumes the hole pattern is still 75mm)

Could also glue copper washers to top of VRM as spacers and remove or thin foam spacers.
(J.B. Weld is common thermally conductive high temp epoxy, else Artic Cool makes a thermal adhesive)
Foam is a shitty heat conductor.

Another easy tweak is to add M3  male x M3 female hex standoff to raise board for better bottom of board cooling.

A plug and pray nightmere, a HW tweakers eye twinkle

YMMV


P.S do not glue VRM directly to heatsink as FrankenHoopiter customers have posted pictures showing the covers are not mechanical stress tolerant except in the down direction.

Sam Cole?

Considering that the "n" in KnC stands for FRUGAL TO A FAULT,
my bets would be on one controller and serial connections. It need not be ribbon cable though.

YMMV

I was in error.

I had not seen the 'press release'
Oddly, I rarely go there. (I have noted previously I don't get out much)

For reference
http://forum.kncminer.com/forum/news-and-anouncments/46242-the-first-20-nanometer-chip-devoted-to-bitcoin

5 boxes it is.

I wonder what happenes to the 600x400x300 dimensions that are on the actual order form.

YMMV

No: KnC puts more than 1 PCB in each box.

The exception was the Mercury.

Saturn, 2 ASIC 1 box
Hoopiter, 4 ASIC 1 box
Frank, 6 ASIC and lots of loose parts 1 box

YMMV

Mark the calendar, I agree with Phoenix.

For those who actually have skin in the game.
1 box 5 chips

For the haters
OMG !! I heard it's 6 boxes now and they are each the size of a house
sky is falling
buried in the wreckage of your future


One constant in this thread is haters declaring victory prematurely.
12 days til it's 'late'

I'm not implying there will not be victories for the haters.
Just don't count ya Satoshies till ya have >1 confirmation

YMMV

unproven until seen in wild but,,,

15 days for packaging, not a month, not 2 months etc...

YMMV