Careful WU includes all shares including those rejected.  You only get paid for valid shares.  In your case you probably are still coming out ahead just didn't want someone else to think higher WU is always better.  Higher WU with same or lower reject is always better.

It would be nice if cgminer had a WA (work accepted / min).

If some miners stop mining then difficulty goes down and those remaining miners get paid more.  The network needs someone to mine, it doesn't necessarily need you to mine.  If your electrical costs are too high, your rig too inefficient, or you simply overpay then you may never turn a profit.

Then don't buy ... and if enough people do that prices will go down.

Well it may be longer than April before that happens and it may turn out that someone who already owns a KNC miner will make more before the "break even electrical" point than someone else will even with a more efficient rig. 

For example an Avalon batch 1 rig has downright horrible efficiency (nearly 6x worse than KNC) but it hardly mattered as it has been a license to print money for the past six months.  I am just trying to point out that difficulty will have some relationship to economics and unless ASICs get significantly more efficient (i.e. 0.2 J/GH or better) dificulty growth will slow down at some point.

Of course this is coming from my point of view as a long term GPU miner.  I studied and built highly efficient GPU rigs (MH/$ and MH/J) and operated then with below average electrical cost.   I never once had to "mine at a loss" or idle my farm.   Sometimes margins were fat, sometimes they were thin but it was the marginal miner (high electrial rate and/or low efficiency) who got forced out first.

It won't do anything.

A little about how a power supply works.   It has to rectify AC power into DC power and it also has to step down the voltage to a level used by computers.  Namely that is 12V, 5V & 3.3V.  

At one time power supplies did rectification and step down to all three voltages

So if you will excuse my horrible ASCII art


120/240VAC --------[AC to 12VDC conversion] ----->  12VDC out
                           |
                           |--[AC to 5VDC conversion] ------>  5VDC out
                           |
                           |--[AC to 3.3VDC conversion] ---> 3.3VDC out

This worked fine when load on all three rails was relatively balanced but efficiency is generally low and when the rails are out of balance efficiency suffers even more.


Two things occurs in roughly that last decade.  The first is that 3.3VDC and 5VDC isn't really used for any significant load.  At one time they were used to directly power CPU, memory, motherboard logic, etc.  However the operating voltage of modern components (including ASICS) is <1 VDC.  The power requirement for a motherboard also increased significantly.   A high end CPU, memory, and motherboard can pull 200W+.  On multi-CPU servers with large banks of memory this can be 500W or more.  500W @ 5VDC = 100A.  You would need a cable the size that connects to your car battery to handle that safely.    So motherboards contain their own DC to DC regulators.   The powersupply pumps in 12VDC the motherboard converts it to whatever it needs.  

The second was the "80Plus" program and it became a big marketing campaign.   People began buying more efficient PSU and it became increasingly hard to design a PSU that
could deliver massive current on the 12V rail, operate with minimal or no load on 3.3V, and still remain very efficient.

Today almost all high end (500W+) PSU use a concept call direct DC conversion.   Since the majority of the load is on the 12V rail ALL the power is converted to 12VDC.  Then the small 3.3V and 5V loads are converted off the 12V rail.

120/240VAC ----->  [AC to 12VDC conversion] ------->  12VDC out
                                   |
                                   |-----[12VDC to 5VDC conversion] -------> 5VDC out
                                   |
                                   | ----[12VDC to 3.3VDC conversion] ------> 3.3VDC out

putting an extra load on 3.3VDC rail does nothing.  Absolutely nothing.

They won't be mining at a loss.  I never said network won't keep growing or that hardware vendors won't produce more hardware pleae don't put words into my mouth.

However say difficulty is now 20 billion and even a miner with a KNC device an $0.10 per kWh is break even.  Anyone with higher electrical cost of less efficient device would be mining at a loss.

With constant price it is highly unlikely difficulty will double again in a month.  Doubling difficulty in a month would mean deploying 147 PH/s of new hardware.  Actually it would mean deploying even more than that because some older tech will go offline due to economic.   

Who would be deploying more hardware enough to double difficulty even with older less efficient tech going offline?   Someone who wants to mine at a loss from day 1?  This is different than someone buying a rig today understimating future difficulty growth.  We are talking about a scenario where for most miners buying on day 1 with anything worse than 1 J/GH means instantly suffering a loss.  Even if the hardware is free you would suffer a loss.  Who is going to convert $1 in electricity into $0.50 in BTC?

So I stand by the claim that the network won't keep doubling forever.  Please don't read into that statement.  It doesn't mean difficulty won't go much much higher, it doesn't mean those with high power costs should mine, it doesn't mean miners can't suffer a loss.  It simply means that difficulty won't keep doubling every month forever and thus saying a rig 2x as efficient as KNC will only be profitable for a month longer is incorrect.

Ultimately for a miners to be profitable LONG TERM in the ASIC world isn't much different than being profitable in the GPU world.  A miner needs:
a) a competitive price on hardware ($/GH or BTC/GH)
b) an efficient design (J/GH)
c) below average electrical costs ($ per kWh or BTC per kWh)

The choice of having the boards run at 5V is horribly bad.  I can't believe it was even considered.

The problem is that most ATX PSU today only produce 12V "natively.  To meet the ATX spec they then use very tiny DC to DC converters to convert some of the 12V into 5V or 3.3V.  There are two problems with this.  The first is to save cost the max load on those rails are usually tiny something like 100W or less.  Also they often don't increase with PSU size.  So a 500W PSU might be 100W max on 5V rail and a 1200W PSU from the same company is still only 500W on the 5V rail.  The second is that efficiency at 5V is going to be lower due to the double conversion.   At first this might seem like a bad choice but the reality is that on a modern computer 90% to 100% of the load is 12V.  So making the 12V rail as powerful and efficient is the best route.

The problem with dedicated 5V devices like the meanwell is that they tend to have bad efficiency.  The unit you listed is 77% avg efficiency.   A cheapo 80Plus Bronze PSU will be 82% to 85%.  The other problem with 5V is it means higher current for the same amount of power.   125W @ 5V is 25A.  125W @ 12V is only 11A.   Higher current means larger (or more) wires, more voltage drop, more power lost in the wires.

There is a third option and that is a high efficiency 12V to 5V converter.   This could be put close to the ASIC module(s) to limit the length of the 5V runs.   I need to run to a wine festival but I will see if I can find an example this evening.

If at all possible I would they would reconsider the input power choice.  It isn't like the ASICs are running at 5V o they are already converting from 5V to ~1V anyways.  Swapping that component out for one which converts from 12V to 1V will make it a better device.

It doesn't matter what it costs them it matters what the market price of 1 GH in chips is worth.  At $1 per GH market prices 1,600 GH/s is $1,600.   HF has indicated they will have third party board designers.  They might screw over their customers but that would be insanely stupid.  Future versions likely won't have an MPP and if they are as efficient as they say they are they can go very low on prices over the next year.  Not saying they won't "scam" their customers but it would be making a few quick bucks over a cut of the entire market for the near future.

Most high end modern PSU no longer need any load on the 5V or 3.3V rails.   They rectify the AC and produce all the output on the 12V rail.  If their is some 5V or 3.3V load it isn't produced by AC power they just use low cost solid state DC to DC conversion to convert some of the 12V power into 5V or 3.3V power.

Honestly there is absolutely no reason that ATX isn't updated to be 12V only.  Nothing runs at 5V or 3.3V anyways.   It sure would make the connectors and PSU design simpler and more efficient. 

Or a thousand months longer. 

The network won't keep doubling forever.   At some point we will reach an equilibrium between energy consumption and mining reward.   For a machine which is twice as efficient to only remain profitable for 1 month longer would assume that difficulty doubles yet again.  However how would it double if their is no chance of a positive ROI% from day 1.

That being said no miner who currently has gear needs to be worried about energy cost.  Profit or loss will be sealed long before that becomes a factor. However hardware costs will continue to decline.   Imagine a future 6 board Sierra at $1 per GH/s.  2.4 TH/s at 1920W.   Hardware cost is ~$2,400.  Electrical consumption annually at $0.10 per kWh is $1681.  High efficiency (and low per kWh prices) can be the difference between a moderate profit and a crushing loss.

100% sure.  HF has said multiple times the MPP will be chips only.   Period.  It isn't ambiguous at all.


Sorry about that.  I read to quickly.  I would assume it will be pretty simple no more complex than installing a second GPU in a computer.  Upthread HF indicated it only requires a screwdriver.

I already point it out thread.

First of all it isn't pin 4, 5, or 6 (pins #1 to #12 is the first row and pins #13 to #24 are on the second row).  The first row is opposite the "clip" and the second row is the side with the clip.  The PS-ON pin is #16.  On 99% of PSU it will be the only green wire.  When PS_ON (#16) is connected to ground (any ground) it will "turn on"* the PSU. 




Any ground pin will work equally fine. As you can see in the pinout above there are multiple ground pins in the ATX-24 connector and they all are exactly the same (the all connect to the ground plane of the PSU).  You are just completing a circuit from the positive voltage on pin #16 to any ground.  What you call #5 & #6 are #17 & #18 and they are both ground pins, either one is fine.   For the record pins #3, #5, #15, #17, #18, #19, and #24 are all grounds, connecting pin #16 to any one of those will complete the power on circuit.  Hell you could even connect a wire from PS_ON to a bare screw on the power supply and it would work (UL listed PSU means the metal chassis is also grounded).



* I use word "turn on" but technically the PSU is always on unless unplugged or it has a physical power disconnect switch.  Internally it is energized and it is sending power to the 5VSB "standby" pin at all times.  Connecting the PS_ON pin to ground pulls the voltage low and the PSU is "looking" for that event to energize the output rails.  Sometimes you can here a click if you jumper PS_ON after PSU is powered on (don't do this with equipment connected).  That click is the power supply switching a relay which turns the output rails on or off.

The ASIC isn't on a board already.  The MPP is raw chips.  IF due to ROI% the MPP entitles you to 2 chips you will receive literally 2 chips. It will require manufacturing the rest of the system including the board the chip is mounted to.  It won't be done by you at home.  Either you will pay (an as of yet not announced) price for HF to build an ASIC board from your free chip or you will use a third party to build a board for you (like third parties build boards using Avalon, Bitfury and even BFL chips).  Another option would be to simply sell the chips. 

According to KNC only Corsair makes booby trapped power supplies.


Then again Corsair doesn't actually make power supplies Channel Well and SeaSonic do and between the two of them they make a good portion of the power supplies on the market from just about every brand.
http://www.tomshardware.com/reviews/psu-manufacturer-oem,2729-5.html

Good thing no Corsair PSU has ever reset on any computer system anywhere in the world prior to today.  If it had it would have blown the caps in motherboards, GPUs and other components. Very lucky for Corsair not a single power supply in the field (out of millions) has tripped before yesterday.

Yes.  Even a CPU.  In a hotswap capable high availability server

I think he was more asking is the PSU ok.  The fact that someone wanted to energize a power supply and THEN connect wires well I mean that is just asinine.  Some people should just buy Apple products and treat them like magical toasters.  Plug it in, use it, throw it away.

You have to be fucking with me right.  They are automated assembly machines.  We are talking a half a second (conservatively).  It isn't like these are hand soldered.  The assembly house is going to crank out thousands of boards a day.  If KNC had a big enough run a good assembly house could do tens of thousands of units per day.  The automated board construction isn't the bottleneck, it is the rest of the mundane manual stuff like bolting down boards, connecting wires, and mounting heatsinks which takes a "long" time.

KNC is claiming that Corsair (one of the most trusted enthusiast brands) "ramps up current too quickly" and that is why a mining board (powered by Corsair PSU) blew a capacitor.  Of course the fact that a high current switching supply SHOULD operate that way should be ignored.  The obvious explanation is that Corsair doesn't know how to build power supplies.  

The car was designed to use 4 wheels.  Adding 4 wheels or removing 3 of them IN production would be equally stupid.  Glad you agree.

The time for version 2 is once orders have shipped and the new CHEAPER (which benefits KNC not customers) 4 VRM version is extensively tested.   Not on the fly right in the middle of production.  You know that.  There is absolutely no reason to switch in production.  Nobody is saying 8 VRMs today, 8 VRMS tomorrow 8 VRMS forevers but the board was designed, simulated, and tested one way.  Then at the last minute the margins upon margins where chopped off with minimal testing.   Oops it turned out to be a bad idea.

Simple.  KNC chose to use an off the shelf power module.  This means rapid prototyping but it also means higher cost.  Those modules run $25 each.  8 per board means $200 per board in cost just for the power supply.  That is $800 per Jupiter.  If they leave 4 of them off that is an extra $400 in profit per unit.  1500 units @ $400 ea = $600,000.

There is absolutely no other reason to change the extensively designed, tested and fully operational boards right in the middle of peak production.