Yes



It shouldn't be a problem it is just pointing out that the SATA/Molex/Peripheral connectors are also on rails 4 & 6 so you shouldn't use PCIe on them as well to avoid overloading.  You aren't hooking up giant banks of hard drive or multiple CD Rom drives so it is a non issue.  You can use any of the six high current connectors at the bottom.  Optimally you want them spread out on as many rails as possible.


You can but it will require modifying or making custom cables.  You have two EPS connectors.  You can't use them directly they won't fit and have different pin out configuration.  However if you cut the EPS end off and replaced it with a 6 pin PCIe connector it would work just fine.  12V is 12V.  The PSU doesn't really care if it is powering a motherboard or hard drive or another couple KNC boards.   Alternatively you could rig a custom cable using the Molex or second motherboard connector.

Note unless you know what you are doing you could do some bad stuff.  You will need to do some self research find out the current limit on each 12V rail, and then come up with a plan from their.  The proper way would be to buy blank PCIe connectors and pins.  Solder the pins and attached them to the connector.  You may be able to rig up something use splice (butt) connectors and an existing PCIe connector.  If all this seems complex and scary you likely shouldn't try.  It isn't rocket science but you should have a good understanding of basic electrical work.

cgminer doesn't control the voltage of any miner to date.  All existing miners (and almost certainly this product as well) operate on a fixed voltage.  Changing the voltage requires changing a resistor on the board.

That is a good simplification.  I would change "correct" to "good enough" as correct implies there is only one possible solution.  For any given block there is a nearly infinite possible solutions.  Miners are collectively just looking for the first one which is "good enough".   We measure good enough by the target, which is a 256 bit number directly proportional to difficulty.  The miner constructs a valid block, hashes it and checks to see if it is "good enough" that is the resulting hash is smaller than the target.

Right down the target is: 000000000000001CDC2000000000000000000000000000000000000000000000.  There are ~2x10^60 hashes smaller than this value and any will meet the block difficulty requirements.

How much have you thought about it because it not a trivial problem.

A proof of work must have the following characteristics:
a) have an adjustable "difficulty" which works under a very large range (bitcoin difficulty is ~150,000,000x higher today than at genesis and likely will go another 10x to 100x over next year).

b) be very fast to verify (<100ms).   Bitcoin has 52K blocks per year so someon bootstrapping 10 years after genesis woould need to validate 520K blocks in a timely manner.

c) the input of the proof must be linked to a prior proof to prevent precomputation (Bitcoin uses the prior block hash as input for current block hash to prevent solving "future" blocks).

d) the proof should be relatively compact (Bitcoin uses a 32 byte hash)

e) the solution should be probabilistic.  This means that given two competitors and one has 10x the computing power the smaller competitor should find ~9% of the solutions.  Many problems are not probablistic such that the faster competitor will always arrive at a solution first.  That would be bad from a security standpoint.

f) require no outside "trusted" source or central authority for issuing or assigning work.

h) have a mechanism to prevent  duplicated work across the network without any communication to a peer or outside data source.

Most simplistic "answers" fail this test.   For example using folding@home would work as a proof of work as long as you are willing to have the administrators of folding@home be the central bank with complete control and authority over the currency.

This is a good point and one that I think will become more common in the future.   In residential scenarios there is something called "the last mile".  It is relatively easy to drop a multi-gigabit data connection into a neighborhood but the installation and maintenance of the last mile into thousands of residences (which you will only collect $30 to $100 monthly) is a bottleneck.  The good news is that datacenter bandwidth is a magnitude cheaper and continues to get cheaper at a faster rate. 

Full node has (relatively) high bandwidth requirements
Users personal tx and confirmations have low bandwidth requirements.
Move the high bandwidth portion to where bandwidth is both cheap and available.

I imagine we will even see the development of ultra light clients which communicate to a specific trusted peer (probably one run by the user).  For example a user could have bitcoin wallets on mobile phone, laptop, desktop, and some hardware device which all communicate via encrypted and authenticated channel to a full node peer operated by the user. Best of both worlds.

Um if you could do all that well you could rip off full nodes as well.  Only the portion related to peer selection is relevant.  SPV are more vulnerable to an isolation attack.   SPV nodes should have very good peer selection algorithms.

Yes just yes.

The question was possible.  Yes it is very much possible.  It is however highly improbable.  The probability of such an event is so low that one should spend time worrying about more likely events such as getting struck by lightning in a desert, or getting eaten by a shark in a sushi restaurant, or winning the lottery the day an extinction level event asteroid strikes the earth.

It likely will dry up future sales until they start shipping but both of the companies you listed have enough funds (preorders) to make it to delivery.  However if anyone pulls a BFL and falls on their face there is no ability to bring in more funding by extended preorders and upgrades.  My guess is both HF and Cointerra deliver.   I don't think every ASIC company today will still be here in 2015.   It is relatively easy to make a fortune when difficulty is low, demand is high and margins are 99.9% or higher.  As the difficulty skyrockets, demand falls and margins collapse not every company will make it.

KNC has never reported the die size.  If someone wants to destroy their new miner and break open the package we could find out real quick.

I didn't even mean to include your quote (will remove it).  I was just speaking in general terms.  The "you" was directed at "you".

Well 10% was a SWAG.   Without knowing the specific details of the chip it is hard to say how bad it is.  Still assumming there is a chip and it otherwise works it could have been avoided by using the right package.

I hate to say "I told you so"  I never owned a share, never will I was just interested from a technology standpoint.   There is a reason every other 110/130nm ASIC chip uses a heat heatpad package.  If you put 12W in (and their 5W estimate was likely nonsense) you have to get 12W out.  Plastic is a great thermal insulator.   You aren't getting 12W out of a chip that small without something that has a higher thermal conductivity.  They likely tried and with a low thermal conductivity the internal temp of the chip skyrocketed.   Now they have a ton of chips which probably can't handle more than 2W which means running them at 10% speed (or slower).

I was surprised the wagers were still roughly right before betting closed. I guess some people really wanted to believe.

I read your post it is nonsense.  The fan on a power supply even at 100% in inaudible over the over all the other system cooling fans (like on a Jupiter means 6 140mm fans).  Having run a 48 GPU farm for 18 months I can say from experience it is a non issue.  The efficiency curve is essentially flat for modern PSU from 20% to 100% load.

If an 850W unit is 95% load (your numbers) that means the load is 807W.  70% load would require a >1150W PSU.  Where is there an 1150W PSU from equal or better brand for $30 more?  It doesn't exist.   Under what conditions would that save $120 annually?  It also doesn't exist.  You underestimate the cost of a larger PSU and overstate the benefits.  Even if you add in the cooling costs and assume low efficiency AC 100% of the year you still don't get $120 in saved energy from going with a larger PSU.  So your $30 increased cost is nonsense and your $120 annual savings are also nonsense.

So one last time your post is nonsense.  Sure one "could" spend $100 more for a 1200W Platinum PSU and it would marginally reduce their operating costs.   However the increased up front cost and the small gain would make the break even point something like 3 to 5 years.  Nothing wrong with doing that but it doesn't mean that going with a smaller gold PSU is a bad idea either.

No.  Well I assume they will make an "upgrade" available but the MPP is for raw chips.

Ok so 3% = 25 W.  Not sure how you got a bogus $120 per year or more out of that.  If your power costs that much then you have no hope of making a profit no matter what PSU you use.

They are not socketed.   Similar to KNC and Bitfury you have a single host connected to one or more boards.   With Bitfury each board consists of 16 chips with KNC and HF (and likely Cointerra) each board consists of one chip.

The difference between 50% load and 100% load on most modern 80 Plus Gold PSU is about 1%.

So 1% of 850W = 8.5W.   That is about 74 kWh per year.   At $0.10 per kWh it is a grand savings of $7.40.

The work is useful.   If you build a depository safe, install a security system, hire round the clock guards, and as a result of the security nobody robs the depository was the resources "wasted"?  Would it have been better to try and making it do something else at the same time?  Would it be better if that compromise resulted in the depository being robbed?

Creating a decentralized consensus is a difficult problem, an incredibly difficulty problem, there has been NO solution to this prior to Bitcoin.  The work isn't wasted the work protects the network.  Bitcoins are worth more than a billion dollars BECAUSE of this solution.

They connect directly to the ground.  It allows a PCIe compliant device to perform a ground sense to ensure the proper connector has been attached.  All the power on both a 6 pin and 8 pin connector come from the exact same 3 12V pins.