Pages:
Author

Topic: The Open Source Block Erupter Project - page 2. (Read 18967 times)

newbie
Activity: 9
Merit: 0
June 19, 2014, 01:28:53 PM

The current understanding is that DataTanks cannot be purchased independently. Rather, they are in groups of six per shipping container.
This option for high density mining will be aimed more towards franchisees, who have access to both the funds and electricity to maintain it.

From the DataTank Mining Prospectus
http://www.datatank-mining.com/files/DataTank_Mining_Ops_Prospectus.pdf

You can technically purchase as little as one unit (0.01kW) of a DataTank if you like, so no its not aimed only for big buyers.

[HAVELOCK] DataTank Mining: 1.2MW 3M Novec Immersion Cooled 2PH Mining Container
https://bitcointalksearch.org/topic/datatank-mining-12mw-3m-novec-immersion-cooled-2ph-mining-container-655464
hero member
Activity: 560
Merit: 500
June 16, 2014, 06:21:35 PM
Here is the X24 blade immersed.

http://vimeo.com/98321051


so a DataTank holds 50 of x24s? 24*10Gh*50= ~12000-13200 Gh/s or 12-13.2 Th/s if my math is right?
this needs 10200-11880 Watts of power?  (100 amps? @120V)
how much is a DataTank? when is it available for purchase etc...?

The current understanding is that DataTanks cannot be purchased independently. Rather, they are in groups of six per shipping container. Based off of this document (on pg 9-10), here are the supposed specifications:

Example 2.4PH 1.2MW Mining Container (ASICMiner Gen3)

● 6 tanks per container, 200kW-240kW each
● 400 boards each tank, 500W per board, 2,400 boards total
● Equivalent to 600 4U boxes
● Shipping container footprint instead of 60 racks in high density facility
● 2-3PH/s at estimated hashrate (based on reports from China24)
● No thermal throttling, possible performance increase to be tested with actual hardware
● Fixed and predictable cost, no matter if fully or only partially populated

This option for high density mining will be aimed more towards franchisees, who have access to both the funds and electricity to maintain it.
hero member
Activity: 489
Merit: 500
Immersionist
June 16, 2014, 06:12:29 PM
so a DataTank holds 50 of x24s? 24*10Gh*50= ~12000-13200 Gh/s or 12-13.2 Th/s if my math is right?
this needs 10200-11880 Watts of power?  (100 amps? @120V)
how much is a DataTank? when is it available for purchase etc...?

What I posted is one of many immersion tests to test various configurations/speed settings on different boards (not only X24).
The X24 blades have no relation to DataTank, they are with air cooling, with ATX power connectors and fans (all of which is not required for immersion).



DataTank:
Your math is off. Here are the design guidelines for DataTank (PDF):
https://docs.google.com/uc?export=download&id=0ByWHHc0u_thNdzB3c2hvVzJkcTQ

donator
Activity: 2352
Merit: 1060
between a rock and a block!
June 16, 2014, 11:43:34 AM
Here is the X24 blade immersed.

http://vimeo.com/98321051






so a DataTank holds 50 of x24s? 24*10Gh*50= ~12000-13200 Gh/s or 12-13.2 Th/s if my math is right?
this needs 10200-11880 Watts of power?  (100 amps? @120V)
how much is a DataTank? when is it available for purchase etc...?
hero member
Activity: 489
Merit: 500
Immersionist
June 16, 2014, 09:37:41 AM
It's a reference design as far as I know. It's not originally made for immersion.
full member
Activity: 224
Merit: 100
June 16, 2014, 09:35:49 AM
Here is the X24 blade immersed.

http://vimeo.com/98321051






Which leaves me with the impression the X24 is additionaly the weapon of choice for AM mining and franchising.
hero member
Activity: 489
Merit: 500
Immersionist
June 16, 2014, 09:04:34 AM
Here is the X24 blade immersed.

http://vimeo.com/98321051





newbie
Activity: 11
Merit: 0
June 15, 2014, 02:01:23 AM

Hi friedcat,

Is there open source cgminer driver? Or some other driver code could be referenced?
legendary
Activity: 1288
Merit: 1004
June 14, 2014, 10:52:30 PM
Exactly.  I hope this unit does that.  It looks to be built in such a way that it would benefit from pulling the air through.

It is definitely directed in an efficient manner from the looks of it over the heatsinks.
I wonder is the fan pushes or pulls the air over them?
I know it some cases like the SP-10 Dawson having the fans pull helps for a more even cooling.
This miner is very elegant in it's design I am impressed I cannot wait to get one or more. 
Great work Friedcat.

Heat rises due to convection.  What if you just turned in 90 degrees with the fan on the bottom to push air up?  That way, the fan assists the natural flow of heated air upward.

I think convection should be neglegable with good airflow.

Looks to me like each board has the same heatsinks and airflow.

Don't forget,as you create a "suction",pulling air through instead of pushing it thru,you get a pressure decrease which also reduces temp.Very little,but the principle is there  Wink
legendary
Activity: 2212
Merit: 1001
June 14, 2014, 08:26:47 PM
It is definitely directed in an efficient manner from the looks of it over the heatsinks.
I wonder is the fan pushes or pulls the air over them?
I know it some cases like the SP-10 Dawson having the fans pull helps for a more even cooling.
This miner is very elegant in it's design I am impressed I cannot wait to get one or more. 
Great work Friedcat.

Heat rises due to convection.  What if you just turned in 90 degrees with the fan on the bottom to push air up?  That way, the fan assists the natural flow of heated air upward.

I think convection should be neglegable with good airflow.

Looks to me like each board has the same heatsinks and airflow.

Don't forget,as you create a "suction",pulling air through instead of pushing it thru,you get a pressure decrease which also reduces temp.Very little,but the principle is there  Wink
legendary
Activity: 1288
Merit: 1004
June 14, 2014, 07:42:35 PM
It is definitely directed in an efficient manner from the looks of it over the heatsinks.
I wonder is the fan pushes or pulls the air over them?
I know it some cases like the SP-10 Dawson having the fans pull helps for a more even cooling.
This miner is very elegant in it's design I am impressed I cannot wait to get one or more. 
Great work Friedcat.

Heat rises due to convection.  What if you just turned in 90 degrees with the fan on the bottom to push air up?  That way, the fan assists the natural flow of heated air upward.

I think convection should be neglegable with good airflow.

Looks to me like each board has the same heatsinks and airflow.
hero member
Activity: 770
Merit: 509
June 14, 2014, 04:57:51 PM
Heat rises due to convection.  What if you just turned in 90 degrees with the fan on the bottom to push air up?  That way, the fan assists the natural flow of heated air upward.

I think convection should be neglegable with good airflow.

Looks to me like each board has the same heatsinks and airflow.
legendary
Activity: 896
Merit: 1001
June 14, 2014, 04:48:09 PM
Heat rises due to convection.  What if you just turned in 90 degrees with the fan on the bottom to push air up?  That way, the fan assists the natural flow of heated air upward.
legendary
Activity: 1288
Merit: 1004
June 14, 2014, 04:44:37 PM
The top as more heat would collect up there and possibly the bottom as it does not get the airflow the other three sides do?



Quote
How does the bottom one and top one stay cool?  Is the amount of air being moved enough to prevent hot spots?
Great work!!

Why would there be hotspots on the top and bottom?
hero member
Activity: 770
Merit: 509
June 14, 2014, 04:40:42 PM
Quote
How does the bottom one and top one stay cool?  Is the amount of air being moved enough to prevent hot spots?
Great work!!

Why would there be hotspots on the top and bottom?
legendary
Activity: 1288
Merit: 1004
June 14, 2014, 04:24:14 PM
I want one of these so badly. 
Great design.  It must regulate the temp well.
How does the bottom one and top one stay cool?  Is the amount of air being moved enough to prevent hot spots?
Great work!!

We had AM_BE200_X24 boards sampled and tested. We also sampled the compatible heatsinks and supporting structures for them. It hashes at 768GH (each board 192GH) with 0.85-0.92W/G (PSU loss excluded). Power draws on higher/lower hashrate with higher/lower voltage is under testing. The design files of heatsinks/structures are going to be released a little later.



It should be emphasized that QUAD and X32 are also sampled and well tested before X24.
sr. member
Activity: 361
Merit: 250
June 14, 2014, 12:40:08 AM
#99
Thank you for the update Friedcat!
donator
Activity: 848
Merit: 1005
June 14, 2014, 12:25:07 AM
#98
Friedcat, any news regarding overclocking/undervolting?
300MHz at 0.78V works fine (need to change the 3kOhm resistors on BOM list to 3.6kOhm resistors).

Undervolting on this design is not effective as expected because as voltage gets down the efficiency of TPS53355 also decreases. Changing to a more dedicated DC/DC solution for lower voltage is recommended.
hero member
Activity: 770
Merit: 509
June 14, 2014, 12:20:47 AM
#97
Friedcat, any news regarding overclocking/undervolting?
donator
Activity: 848
Merit: 1005
June 14, 2014, 12:18:27 AM
#96
Friedcat,

Regarding the design for the new cube, will each boards of the cube require a power supply and Ethernet connection? Will the cubes need a stratum proxy to operate like your previous builds or are you updating the mining software?
For power supply: we are going to open source the PSU adapter for 4*2*6pin power sockets.

For ethernet: we are going to open source the ethernect controller based on microchip PIC-32 MCU with a built-in stratum proxy. Each controller can drive as many as 32 boards (8 devices). Remember that since the boards are chained in a string instead of connected to the controller separately, raspberry-pi or PC with limited IO ports would also do the job.
Pages:
Jump to: