The Open Source Block Erupter Project - page 4.

jimmothy

hero member

Activity: 770

Merit: 509

Quote from: Bicknellski on June 12, 2014, 02:48:04 AM

Put up a spreadsheet and show us where these will make money for the fabricator and the miners. Please I would love to see your projections and do not forget to add all the costs in for the end users please.

Here is my ROI estimations (starts in 3 difficulty jumps)

http://btcinvest.net/en/bitcoin-mining-profit-calculator.php?diff=11756551917&dcosts=750&diff_mincrease=15&blpbtc=25&dhsmhs=750000&diff_mincreasedecrease=3&btcusd=650.80&dpowcon=750&btcusd_mincrease=1&pcost=0.05&calcweeks=32&dleadtime=3&action=calc

This assumes miners are being sold for $1/gh. With production costs at less than $0.8/gh there is clearly profit for the manufacturer.

Quote

Your estimations are based on what?

Given what numbers are available most smaller miners need to be moving out of Sha256. You can hope and dream these units will be cheap but no one is going to make them if they know it is going come back and bite them when they can't sell them at the price they need to. Just a few of those those smaller board makers who would target the remaining and every dwindling small miner market.

What are your estimations based on? What numbers are available? Where are small miners moving to from sha256?

I don't see a dwindling small miner market at all. Almost every manufacturer has a small miner for sale.

Quote

The numbers we have run do not look too promising even at greatly reduced price for BE200 or 300 chips.

I think you are also assuming that there is some next gen chips that will blow these miners out of the water which will likely not be the case for a few months (5-7 months I would guess).

Bicknellski

hero member

Activity: 924

Merit: 1000

Your estimations are based on what?

Given what numbers are available most smaller miners need to be moving out of Sha256. You can hope and dream these units will be cheap but no one is going to make them if they know it is going come back and bite them when they can't sell them at the price they need to. Given that there are only just a few of those those smaller board makers left who would target the remaining and ever dwindling small miner market I don't see them taking on this risk unless it is a Chinese based fabricator and even then doesn't look too promising.

If you got the numbers let us see it. Put up a spreadsheet and show us where these will make money for the fabricator and the miners. Please I would love to see your projections and do not forget to add all the costs in for the end users please. The numbers we have run do not look too promising even at greatly reduced price for BE200 or 300 chips.

jimmothy

hero member

Activity: 770

Merit: 509

Quote from: Bicknellski link=topic=641957.msg7264981#msg726498

The average miner shouldn't be investing in mining given the numbers currently.

I disagree. There are plenty of people with cheap electricity. You don't need a giant mine to make a profit. You simply need a good $/gh.

Quote

These things are architecturally interesting but in terms of what is in the market already for air cooled it is on the wrong end of the curve price and design wise given the modularity data center air cooled versions already out.

What on the market is more modular? Cubes are about as modular as it gets. Not to mention it can be disassembled in to blades.

Quote

If it were profitable for Asicminer to build them they would.

I also disagree. These are almost certainly profitable for AM to sell. Hashratios production cost is less than $0.8/gh for a much less cost effective miner (bulky case + extra fans and heatsinks).

I would guess AM could produce these for around $0.5/gh which is on par with spondoolies sp30 BOM.

At $0.5 or even $0.8 there is plenty of room for profit considering the market rate is ~$1.3-$2/gh.

Problem is this would take tons of capital AM doesn't have. And it would create a situation where AM is directly competing with its partners/customers.

I think it is better for AM to focus on providing the chips and let the chip integrators battle it out.

Bicknellski

hero member

Activity: 924

Merit: 1000

AM is not stopping their immersion development. So anyone building the cubes will be in direct competition with the DataTank boards likely with the BE300 chips or the next gen they might have on the drawing board after that. Gotta keep this in perspective if AM is going to produce 100s of PH/s worth of chips or more a significant amount is going into data tanks via that single wide or double wide immersion blade. Welcome to the end home miners.

And by the way... there is an air cooled version not unlike the spondoolies SP10 / SP30 configuration have a look in the link above. Enjoy your cubes. Not sure if you can get the chips on a board configuration that narrow given the power going through the copper for air cooling but I bet someone thinks you can do it and have an air / novec cooled boards in the same exact configuration.

antirack

hero member

Activity: 489

Merit: 500

Immersionist

Speaking of immersion cooling, this thread wouldn't be complete without the design guidelines for the 240kW tanks:

https://docs.google.com/uc?export=download&id=0ByWHHc0u_thNdzB3c2hvVzJkcTQ

Here are some more files:
https://drive.google.com/a/allied-control.com/folderview?id=0ByWHHc0u_thNT293cTl6OXBVZms&usp=drivesdk#

Here some AM history:

Jutarul

donator

Activity: 994

Merit: 1000

Quote from: friedcat on June 11, 2014, 02:30:40 AM

The CAD of the mechanical design for X24 has been put in the repo:
...
Suggestions, modifications and brand new designs are all welcome.

In other words: The heatsink is the case.
Well thought out - should bring down deployment costs.

Regarding air vs. immersion cooling.
Immersion cooling is still a job for the professionals. If I understand the open source nature of the new block erupter correctly, the air-version is supposed to be a DYI kit for the little guys.

Bicknellski

hero member

Activity: 924

Merit: 1000

Quote from: jimmothy on June 12, 2014, 12:25:07 AM

Quote

Heat sinks and the volume they take up is the issue here not weight. That is valuable real estate in a DC and even in a small node at home. Why did AM build themselves a modern high density cooling solution?

They already built the future. This offering is not it.

This is what the industry wants, maybe this is not what the home user wants it, but this is how the game is going to be played:

Not sure I completely agree with that. Immersion cooling is definitely the future but it is quite a ways off before it becomes economically viable for most DCs let alone most homes.

I predict that a very efficiently deployed immersion setup would cost ~$150/kw. For those with cheap electricity, it would still make sense to go with air cooling. For example if you had an an efficient setup ($150/kw) and $0.05/kwh electricity you would be saving ~1/3rd on electricity so ~$0.016/kwh or about 1 year to break even.

You simply have to admit that for the average home miner and air cooled DCs with cheap electricity, this is quite an attractive machine.

Side note: I am 99% sure that AM has already made a high density immersion solution and I'm sure they will give it to anyone who asks (not positive about that last part)

The average miner shouldn't be investing in mining given the numbers currently. Plenty of people have posted the numbers already I don't see how small scale is going to cut the mustard. These things are architecturally interesting but in terms of what is in the market already for air cooled it is on the wrong end of the curve price and design wise given the modularity data center air cooled versions already out. Density, density and density. If you have to stack 10 of these to get comparable hash rate from something a lot smaller I can't see the benefits no matter how pretty it looks. What I am talking about here is an alternative to this configuration. The cube is great style wise but again it is going to take up a lot of space where a flat plane configuration would be more ergonomic. Again it is a numbers game. If you can get the chips for pennies and the PCB's done up en masse sure ok... but that better be ready this month or next before the difficulty and other fabricators drop their offerings.

If it were profitable for Asicminer to build them they would.

jimmothy

hero member

Activity: 770

Merit: 509

Quote

Heat sinks and the volume they take up is the issue here not weight. That is valuable real estate in a DC and even in a small node at home. Why did AM build themselves a modern high density cooling solution?

They already built the future. This offering is not it.

This is what the industry wants, maybe this is not what the home user wants it, but this is how the game is going to be played:

Not sure I completely agree with that. Immersion cooling is definitely the future but it is quite a ways off before it becomes economically viable for most DCs let alone most homes.

I predict that a very efficiently deployed immersion setup would cost ~$150/kw. For those with cheap electricity, it would still make sense to go with air cooling. For example if you had an an efficient setup ($150/kw) and $0.05/kwh electricity you would be saving ~1/3rd on electricity so ~$0.016/kwh or about 1 year to break even.

You simply have to admit that for the average home miner and air cooled DCs with cheap electricity, this is quite an attractive machine.

Side note: I am 99% sure that AM has already made a high density immersion solution and I'm sure they will give it to anyone who asks (not positive about that last part)

loshia

legendary

Activity: 1610

Merit: 1000

Led Boy,
Just to remember you the thread is:
The Open Source Block Erupter Project. So if you ever build something apart of bilking leds you can do whatever you want. As you see AM wants to sell chips. AM is giving us the option to do two things:
1. Use his design and start from there he offered even assembly service
2. To build our own how we like it - chip docs and protocol spec PCB, bom and everything is available

So get back to work Led boy. If you do not like what AM is providing you here just design and build it. Is that the goal of your WASP balloon or what?
But WASP boards waiting list is so long and you are too busy posting noncecnce everywhere so it is not gonna happen soon. Correct?
If you wanna stay on topic just take a look at github there is a loot room for improvement especially the LED PART Grin

Bicknellski

hero member

Activity: 924

Merit: 1000

Quote from: jimmothy on June 11, 2014, 05:01:42 PM

Quote from: Bicknellski on June 11, 2014, 05:44:52 AM

Would they use these chips if they can get something more powerful same price or lower?

Of course not if it is cheaper and equally or more efficient. But if it is the same price they would probably still go with AM simply due to the fact that FC delivers.

Quote

Just asking? Would they want to make something that required all that aluminum? Even Marto isn't selling boards now with all his new designs... guess chips cost too much to make a profit compared to Bitmain offerings. Guess that small fabricator is going out of business as well.

Where exactly do you see "all that aluminum"? I just downloaded the cad files and it looks like it uses only 0.64 lbs of aluminum (excluding heatsinks).

Also it uses a single 120mm fan for cooling. I don't know if it gets better than that. Maybe a single 180mm fan would be better although there is a much better selection on 120-140mm fans.

Heat sinks and the volume they take up is the issue here not weight. That is valuable real estate in a DC and even in a small node at home. Why did AM build themselves a modern high density cooling solution?

They already built the future. This offering is not it.

This is what the industry wants, maybe this is not what the home user wants it, but this is how the game is going to be played:

jimmothy

hero member

Activity: 770

Merit: 509

Quote from: Bicknellski on June 11, 2014, 05:44:52 AM

Would they use these chips if they can get something more powerful same price or lower?

Of course not if it is cheaper and equally or more efficient. But if it is the same price they would probably still go with AM simply due to the fact that FC delivers.

Quote

Just asking? Would they want to make something that required all that aluminum? Even Marto isn't selling boards now with all his new designs... guess chips cost too much to make a profit compared to Bitmain offerings. Guess that small fabricator is going out of business as well.

Where exactly do you see "all that aluminum"? I just downloaded the cad files and it looks like it uses only ~~0.64~~ 0.32 lbs of aluminum (excluding heatsinks).

Also it uses a single 120mm fan for cooling. I don't know if it gets better than that. Maybe a single 180mm fan would be better although there is a much better selection on 120-140mm fans.

CanaryInTheMine

donator

Activity: 2352

Merit: 1060

between a rock and a block!

Quote from: necro_nemesis on June 11, 2014, 04:29:33 PM

Quote from: CanaryInTheMine on June 11, 2014, 06:21:01 AM

Quote from: necro_nemesis on June 11, 2014, 05:42:26 AM

Quote from: friedcat on June 11, 2014, 02:30:40 AM

Suggestions, modifications and brand new designs are all welcome.

I would interconnect the four board's power requirements into a single pair of PCIe connections.

I would injection mould or extrude a plastic clip to interlock the handles so they can be assembled into a matrix.

I dunno... Current design leaves power flexible... You can repurpose all kinds of older/cheaper PSUs
AND when next gen chip is out, you could upgrade one board at a time too instead of replacing whole device if power is interconnected

Seen. Edit I would put a provision in to interconnect the four board's power requirements into a single pair of PCIe connections. The idea is to eliminate the requirement to purchase splitters or adapters for larger PSU's to power the system which could be viewed as a negative.

There are many of us that have acquired pretty substantial power supplies due to previous system inefficiencies. I would also look at something that can take advantage of connecting to the abundance of used server supplies available out there. If you can demonstrate a cost savings for the entire system through intelligent design and save the customers in areas that won't effect your own bottom line it's a more attractive product at the cost of thinking the full implementation through.

The ability to nest miners with relatively inexpensive clips also becomes a savings in not requiring racks to form a stable rig set up. It also predisposes people to continue to buy the same product since it can be added to in a modular manner.

AM Lego. Grin

I assume this system is designed to have a host controller like a Raspberry Pi. I would anticipate busses to be able to interconnect as well again reducing the number of lines to the system.

I think because of the plethora of different PSUs out there, it's most economical to just use longer pci-e 6/8 pin cables to connect to source PSUs and modify/vut solder as needed. breakout boards for dell/hp/ibm servers come to mind...

yes, a host controller can be a computer, a Pi or an x24-specific microcontroller (perhaps will be released here too), 32 boards can be linked together (8 devices) the controller connects to one board, rest are cascaded (serial connection).

necro_nemesis

full member

Activity: 224

Merit: 100

Quote from: CanaryInTheMine on June 11, 2014, 06:21:01 AM

Quote from: necro_nemesis on June 11, 2014, 05:42:26 AM

Quote from: friedcat on June 11, 2014, 02:30:40 AM

Suggestions, modifications and brand new designs are all welcome.

I would interconnect the four board's power requirements into a single pair of PCIe connections.

I would injection mould or extrude a plastic clip to interlock the handles so they can be assembled into a matrix.

I dunno... Current design leaves power flexible... You can repurpose all kinds of older/cheaper PSUs
AND when next gen chip is out, you could upgrade one board at a time too instead of replacing whole device if power is interconnected

Seen. Edit I would put a provision in to interconnect the four board's power requirements into a single pair of PCIe connections. The idea is to eliminate the requirement to purchase splitters or adapters for larger PSU's to power the system which could be viewed as a negative.

There are many of us that have acquired pretty substantial power supplies due to previous system inefficiencies. I would also look at something that can take advantage of connecting to the abundance of used server supplies available out there. If you can demonstrate a cost savings for the entire system through intelligent design and save the customers in areas that won't effect your own bottom line it's a more attractive product at the cost of thinking the full implementation through.

The ability to nest miners with relatively inexpensive clips also becomes a savings in not requiring racks to form a stable rig set up. It also predisposes people to continue to buy the same product since it can be added to in a modular manner.

AM Lego. Grin

I assume this system is designed to have a host controller like a Raspberry Pi. I would anticipate busses to be able to interconnect as well again reducing the number of lines to the system.

CanaryInTheMine

donator

Activity: 2352

Merit: 1060

between a rock and a block!

Quote from: necro_nemesis on June 11, 2014, 05:42:26 AM

Quote from: friedcat on June 11, 2014, 02:30:40 AM

Suggestions, modifications and brand new designs are all welcome.

I would interconnect the four board's power requirements into a single pair of PCIe connections.

I would injection mould or extrude a plastic clip to interlock the handles so they can be assembled into a matrix.

I dunno... Current design leaves power flexible... You can repurpose all kinds of older/cheaper PSUs
AND when next gen chip is out, you could upgrade one board at a time too instead of replacing whole device if power is interconnected

loshia

legendary

Activity: 1610

Merit: 1000

Yes led Boy Grin

Why?
1. Because not everyone get same prices
2. Because not every one are licking manufacturer asses
3. Because it is a market and if someone get X price for Y chip it may turn out that other one can get 5X price for same Y chip. And magic Y chip which supposed to be the cheapest one suddenly becomes the expensive one of them all

Get ready for Christmas boy. Led tree market is raising Grin

Greets Bick Wink

Bicknellski

hero member

Activity: 924

Merit: 1000

Quote from: minerpumpkin on June 10, 2014, 02:03:54 PM

Quote from: ?? on ??

Quote from: jimmothy on June 10, 2014, 01:13:37 PM

Quote from: ?? on ??

Very nice and exciting for the future to come! I'd love to work with a manufacturer and bring these to the US masses.

I'd love it if you'd do that ASAP..

First company to sell these for a decent price (~$1/gh) gets my money.

As would I!

I will definitely be exercising all options that I have to make this happen. It's a bit harder seeing as most, if not all, manufacturers are outside of the states. We all have seen what happens when a US company tries to work with an overseas company and has a bad business and communication relationship (Don't need to point fingers). With that said, I would go at this a lot differently then the big companies only going after $. A smaller scale and most important realistic prices, service and expectations. I wouldn't offer what I cannot deliver (ha! How many times have people said that before?) but my track record speaks for itself now.

Taking on more work that I can handle is something I will not do nor begin to do like a lot of these short-term money hungry companies have been.

Enough about me.

If anyone knows contributors who would like to speak, feel free to shoot them my way.

We need a manufacturer who has enough money to afford a large enough order directly with AM. We could always default to the good ol' chips group buy done by a reputable member who'll then forward the chips to a (again, reputable) manufacturer. That's going to cut the price and lower all margins.

Would they use these chips if they can get something more powerful same price or lower? Just asking? Would they want to make something that required all that aluminum? Even Marto isn't selling boards now with all his new designs... guess chips cost too much to make a profit compared to Bitmain offerings. Guess that small fabricator is going out of business as well.

necro_nemesis

full member

Activity: 224

Merit: 100

Quote from: friedcat on June 11, 2014, 02:30:40 AM

Suggestions, modifications and brand new designs are all welcome.

I would interconnect the four board's power requirements into a single pair of PCIe connections.

I would injection mould or extrude a plastic clip to interlock the handles so they can be assembled into a matrix.

bitsalame

donator

Activity: 714

Merit: 510