That's easy.  Just take cards out of a bunch of boxes, and fully populate the other ones.  Then throw away a HUGE PILE of extra cases, base boards, and power supplies, and you still paid less.  Or hell, sell the PSU's and cases for scrap and you got it all for even cheaper!

This sounds about right to me.  Those types of PDU's usually have a couple of "always-on" outlets for other things, and the limit of 2500W for the device comes down to the supply cabling to the PDU itself.  I'm not sure what standard wall connectors are rated at in germany, but 2400W is the practical limit for a standard outlet here.  So it's not the device itself that is limited to 2500W, it is the wall plug.


You calculated right.  Most datacenters max out at around 5KW per rack... many are lower than that.  If you're lucky, you MIGHT be able to find some that will go up to 7KW.  This maximum is derived from typical uses cases for hosted customers, and the entire datacenter is built around this assumption.  All of the power cabling, cooling systems, etc, is designed around that maximum. So it is tough for them to go over that without significant work.


You're kind of out of luck.  I would look at smaller power supplies, and more of the switched PDU's.

Steamboat-

Thanks for all the updates... the hardware looks awesome!

I'm contemplating an upgrade from DIY kit form to fully-assembled and tested boards.  What sort of difference in shipping time do you expect between these options? i.e., how many days to assemble/test each batch of boards, and will you ship out the DIY kit form ones first?

* USB is a much simpler (and cheaper) interface to work with than PCI-E for development of this type
* A single host computer can have a lot more USB devices connected to it than PCI
* The host doesn't even have to be a PC- it could be a raspberry pi or TP Link router
* The boards in development now use the PCI-E power connectors anyway that GPU's use

Ah, ok- I hadn't seen that update yet. 450mhz drops the price per GH down to $23, so it certainly is better.

If 320mhz is the real limit of the K16, the "break even" point is around 425mhz for a 12-chip board.

The bigger problem: 94W for 20 chips is 4.7W per chip... I don't know if that would work even with 12 chips.

From BKK's testing, the last board revision with modifications was good for 320mhz stable... the new board revision should be that good and maybe even a little better.  From other's testing, the chips are only really stable to 409mhz on either water OR air cooling. Since you'll need to buy extra boards, the math just doesn't work out for under-populated boards. It works out to roughly $25 per GH for 12-chip boards and $24 per GH for 16-chip boards.. and you'll have to deal with the instability and shorter life of the OC'd ones too.

Yes.  I can't find any exact threads, but I have read reports of chips not performing correctly in avalon-built units.  Of course it depends on what the actual failure mode is, but I wouldn't count it out.

Yes, that will work.  6-pin connectors are rated to deliver 75W.  8-pin are rated to deliver 150W.  Also note that these are nominal ratings- the physical connector is OK for more than double that.  It would *probably* be safe to run three or four K16's off of a single 6-pin, but that will require some experimentation.

Lots more information here: http://www.tomshardware.com/reviews/power-supply-specifications-atx-reference,3061-12.html

Also, one of the most important things to look at when picking a PSU- make sure it is single rail.  With cheap multi-rail power supplies, the current handling capability will be divided up across those rails, and maybe in ways that don't divide evenly by 32W.

Or you could have just put the numbers:

10 GH - 18.88 BTC
5 GH - 9.88 BTC

Not worth it.  But people will probably pay it anyway, since it's ASICMiner....

Whoever said september was just pulling numbers out of thin air.  The first chip batches were sold at the end of April.  Avalon specified a lead time of 9-10 weeks, which works out to the end of last month.  

Also- if you look at Avalon's BTC address for chip orders, there's approx. 75,000 BTC there.  This works out to around 97 orders of 10,000 chips, or just short of a million chips.  So these pictures may very well be almost ALL of the orders of avalon chips so far.

Avalon's address: http://blockchain.info/fb/1fgaftz

Since I'm a nice guy:



Whole album, with translated captions:
http://imgur.com/a/GR0e3

Is this satire? It sure sounds like satire.  I really hope for for your sake that this is not a serious post.

The stuff I read on here sometimes.  You do realize that the higher the network hash rate is, the more stable bitcoin is?  One of the largest risks to any blockchain-based cryptocurrency is the danger of a 51% attack.  The higher the overall network hash rate is, the less likely it is for a 51% attack to succeed.  So no, invalidating asics won't make people "take bitcoin seriously."

If anyone here was expecting fast ROI... you're in the wrong game.  This group buy ALONE is for 14,000 GH worth of hash rate.  At the quoted 282mh/s, that's close to 10% of the entire bitcoin network right now.  If you are running your calculations now, you should be putting in a number like 100 million for the difficulty at the time you start.  I'm expecting 6 months before I hit ROI (calculated in BTC).  Could be sooner if BTC's value rises during that time.

Order number 100000014, in and paid

The TH board is functionally equivalent to four K16's linked together with i2c.

Steamboat: I have a question about the payment order queue and future changes to our order.  For right now, I'm planning on ordering JUST board assembly at $95 ea, since that is what I can afford.  If I decide later on to change my order to fully assembled boards, will I lose my place in the assembly queue?

I don't think that would really be a problem.  The PCI-e 6-pin connector has a design specification of 75W, and the pinout is:

Pin 1 - +12V
Pin 2 - N/C
Pin 3 - +12V
Pin 4 - GND
Pin 5 - Sense
Pin 6 - GND

So, looking at the wires actually delivering power you have 37.5W on each +12V / GND pair at the rated PCI-e spec.  On top of that, the connector is actually rated for higher power delivery- more than double the 75W specification.  Assuming you use proper diameter wires and good termination, one pair of power wires will be more than enough.

tl;dr: yes, it would work ok.

(edit) Just noticed you had said pin 5- you would want to use pins 1 and 4, not pin 5.

I'm planning on mounting 4 units in a slim micro ATX case, using something like this:



You do need SOME sort of standoff to keep the PCB suspended in the case, and you will need the others to hold the fan up.

Take a look at this little guy: http://www.newegg.com/Product/Product.aspx?Item=N82E16817129006

Fanless, and enough for a K16 plus a rPi, for only $5 more.