Author

Topic: Who designs a Mining Board for BFL Chips? (Read 2051 times)

legendary
Activity: 1190
Merit: 1000
July 01, 2013, 11:36:38 PM
#21
Intron said he would have a look at the BFL https://bitcointalksearch.org/topic/m.2583534

However, cant see there being much interest. Avalons are supposedly couple of weeks away, and if Bitfury chips come through at a reasonable price, it will wipe the floor with everything.

1)As far as we know besides a few testers Bitfury's chips will only be made available to Metabank and 100TH.
2)BFL has an incentive to supply chips since they collect 50% on delivery.
3)BFL chips will come directly from GlobalFoundries which will avoid many of the supply chain problems.

Trust BFL?

How did you get those two words to be in the same sentence?
Black magic!
hero member
Activity: 924
Merit: 1000
Intron said he would have a look at the BFL https://bitcointalksearch.org/topic/m.2583534

However, cant see there being much interest. Avalons are supposedly couple of weeks away, and if Bitfury chips come through at a reasonable price, it will wipe the floor with everything.

1)As far as we know besides a few testers Bitfury's chips will only be made available to Metabank and 100TH.
2)BFL has an incentive to supply chips since they collect 50% on delivery.
3)BFL chips will come directly from GlobalFoundries which will avoid many of the supply chain problems.

Trust BFL?

legendary
Activity: 966
Merit: 1000
- - -Caveat Aleo- - -
Intron said he would have a look at the BFL https://bitcointalksearch.org/topic/m.2583534

However, cant see there being much interest. Avalons are supposedly couple of weeks away, and if Bitfury chips come through at a reasonable price, it will wipe the floor with everything.

1)As far as we know besides a few testers Bitfury's chips will only be made available to Metabank and 100TH.
2)BFL has an incentive to supply chips since they collect 50% on delivery.
3)BFL chips will come directly from GlobalFoundries which will avoid many of the supply chain problems.
newbie
Activity: 5
Merit: 0
No custom milling needed if you just provide the right mounting holes on the board and position the chips correctly under the heatsink.

It is entirely possible to create 1.0v power supplies that deliver up to 120 amps with very little difficulty using polyphase buck converters. For instance, a 6 phase system need only handle 20 amps per mosfet/inductor, but once you have gone that far, it doesn't cost that much more in parts to have an independent 12-15 amp power supply for each chip.
newbie
Activity: 46
Merit: 0

It would be awesome if one of these can be attached. I have a spare H60v2 just lying around, waiting to get to work. Also, from the one on my CPU, I can guarantee that these dissipate a lot of heat Grin

EDIT:
Actually, water cooling would be a great and efficient option for any ASIC chip... It's just that producing a custom milled radiator for the assembly is so expensive. Oh, well.
legendary
Activity: 2702
Merit: 1468
I thought BFL published a working design and firmware all open source,  so it's only a matter of getting boards made, sourcing that parts, and assembling, is that not the case?


It is a matter of time, before someone copies the BFL released design and puts 4 chips on it.
People complaining that power mosfets will not support 8 chips, but if you actually look at the specs of them (100A and 68A max drain), they can power way more than 8 chips.  Sure they will be hot, but show me one power mosfet that is not hot.

I have not seen one technical reason why you cannot use the released design to run at least 4 chips.
BFL invested a lot of time in this design. (revision E).  All of this try and error on new design will cost time and money.

Let's see:
  • Working firmware, check.
  • cgminer support, check
  • proven 1 chip product, check
  • sufficient power supply, check

I think I'm going to make a copy of that board and mount my sample chips.
newbie
Activity: 5
Merit: 0
I'm working on several designs as a part of a private BFL chip buy being operated for the members of the Eligius Pool. In the works are a modular design that can take from 1 to 8 chips, and an 8 chip board that is stackable up 8 boards for a total of 64 chips. The stackable boards will have per-chip programmable voltage control for over-volting/over-clocking. Chip cooling will use 2 CPU heatsinks so overclockers have the option to install water cooling or any other available CPU cooling system.

I'll be more than happy to provide fab services to any group buy that needs them. I expect to have pricing information within the next 2 weeks. Turn time from chip delivery to shipping is 3 weeks.
sr. member
Activity: 308
Merit: 250
I'm hard at work on a BFL design along with another Engineer doing FW/MCU, and it's going quite well (IMO). At the current time we are just pushing to get a small 4 chip board done to characterize the BFL chips and the DC/DC convertor, after which we will be building full mining boards.

I expect to release more information early next week as the test board goes out to fab, but a few of the planned details of the prototype board are below. Some features might not make it to production depending on how testing goes, but it should be quite similar other than the number of chips.
  • 4 BFL ASICs
  • LGA1155 HS mounting holes
  • Circuitry for reading the on-die thermal diodes (if they work)
  • External clock input
  • 3 phase Intel VR11.1 compliant programmable power supply - comfortable at 60A
  • ARM M3 MCU core

Samples are on order from BFL, so the hope is that development boards should be ready mid-July and test production boards by the end of the month, perhaps mid-August depending on results and how many revs are needed.

Sounds promising! If you can sell these populated with the ASICs you'll make a killing.
legendary
Activity: 952
Merit: 1000
legendary
Activity: 1274
Merit: 1004
I'm hard at work on a BFL design along with another Engineer doing FW/MCU, and it's going quite well (IMO). At the current time we are just pushing to get a small 4 chip board done to characterize the BFL chips and the DC/DC convertor, after which we will be building full mining boards.

I expect to release more information early next week as the test board goes out to fab, but a few of the planned details of the prototype board are below. Some features might not make it to production depending on how testing goes, but it should be quite similar other than the number of chips.
  • 4 BFL ASICs
  • LGA1155 HS mounting holes
  • Circuitry for reading the on-die thermal diodes (if they work)
  • External clock input
  • 3 phase Intel VR11.1 compliant programmable power supply - comfortable at 60A
  • ARM M3 MCU core

Samples are on order from BFL, so the hope is that development boards should be ready mid-July and test production boards by the end of the month, perhaps mid-August depending on results and how many revs are needed.
sr. member
Activity: 479
Merit: 250
What about treating a Jalapeno as a base to expand on? They are low cost, functional if not touched, expandable (still to be confirmed) and shipping.

There is the option of ordering a Jalapeno and using additional 3 chips, JTAG flashing with the released firmware and getting a 20 GH Miner for $274 + $150 + rework cost.

Actually with a 2 chip upgrade for $100 plus rework cost you may even be able to run 16GH with no upgrade to power or heatsink, as the enclosure was designed for 60W.

A Jalapeno and Chips ordered today may arrive around the same time (100+ days )

There are 6 spare pads on the Jalapeno which would need to be solder stenciled, reworked with the new chips, and flashed via JTAG.

I understand the solder stencil and rework will be risky with an already densely populated board though. Anyone else think this might fly?





I will be doing the above service for anyone going this route as I repair xbox360 ps3 and laptops with a scotle ir 6000 v4
hero member
Activity: 630
Merit: 500
There is a guy organising bfl chips group-buy and also stating he is working with a company developing a PCB board? https://bitcointalksearch.org/topic/bfl-chip-groupbuy-275chip-50-btcpaypalsepa-420-sold-0100-batch-4-238052
full member
Activity: 237
Merit: 100
I think no one is bothering because there is no reason to chose BFL over Avalon.
The BFL chips cost about half as much per GH/s, assuming you're using a discount, of which there are thousands for sale dirt cheap. Also, they use less power per GH/s. So ya, I'd say there's a few reasons to look into it.

Look?

No one currently doing any serious design right?

this guy is serious:
https://forums.butterflylabs.com/hardware-discussion/3305-%5Bgroup-buy%5D-chip-group-buy-board-design.html
sr. member
Activity: 327
Merit: 250
Check out this GB thread by CanaryInTheMine for buying BFL chips. Theres others in there too.

https://bitcointalksearch.org/topic/closed-bfl-4-ghs-chips-06-btc-per-chip-batch-29-236103

He lists a couple of developers, but nothing tangible just yet. Perhaps we can get BkkCoins or burnin, etc. to develop a PCB and cooling solution to these chips? Considering they are ~4Gh/s each, that makes them over 10x faster and at least 5x more efficient per watt than Avalons. I have some chip vouchers I would like to see completed into full on miners, or just use my vouchers to buy chips to add to my 4 Jalapenos with 6 empty chip slots each. I lack the necessary equipment, but as more Jalapenos get shipped out, I think public interest will definitely spike if you can magically turn a 5-7Gh/s box into a ~28Gh/s machine if the firmware and cooling solutions can be met. (Even though it seems this was BFLs original plan, who knows, there might be better solutions as everything becomes open source.) I definitely see these BFL chips, as much as I hated the wait, as having a lot of potential until KNCMiner and Bitfury release their chips/crazy machines, but keep in mind we are dealing with actual hardware to work with.

Anyone else on board?
hero member
Activity: 924
Merit: 1000
I think no one is bothering because there is no reason to chose BFL over Avalon.
The BFL chips cost about half as much per GH/s, assuming you're using a discount, of which there are thousands for sale dirt cheap. Also, they use less power per GH/s. So ya, I'd say there's a few reasons to look into it.

Look?

No one currently doing any serious design right?
erk
hero member
Activity: 826
Merit: 500
I thought BFL published a working design and firmware all open source,  so it's only a matter of getting boards made, sourcing that parts, and assembling, is that not the case?
sr. member
Activity: 420
Merit: 250
What about treating a Jalapeno as a base to expand on? They are low cost, functional if not touched, expandable (still to be confirmed) and shipping.

There is the option of ordering a Jalapeno and using additional 3 chips, JTAG flashing with the released firmware and getting a 20 GH Miner for $274 + $150 + rework cost.

Actually with a 2 chip upgrade for $100 plus rework cost you may even be able to run 16GH with no upgrade to power or heatsink, as the enclosure was designed for 60W.

A Jalapeno and Chips ordered today may arrive around the same time (100+ days )

There are 6 spare pads on the Jalapeno which would need to be solder stenciled, reworked with the new chips, and flashed via JTAG.

I understand the solder stencil and rework will be risky with an already densely populated board though. Anyone else think this might fly?




yes, it will work.
full member
Activity: 237
Merit: 100
What about treating a Jalapeno as a base to expand on? They are low cost, functional if not touched, expandable (still to be confirmed) and shipping.

There is the option of ordering a Jalapeno and using additional 3 chips, JTAG flashing with the released firmware and getting a 20 GH Miner for $274 + $150 + rework cost.

Actually with a 2 chip upgrade for $100 plus rework cost you may even be able to run 16GH with no upgrade to power or heatsink, as the enclosure was designed for 60W.

A Jalapeno and Chips ordered today may arrive around the same time (100+ days )

There are 6 spare pads on the Jalapeno which would need to be solder stenciled, reworked with the new chips, and flashed via JTAG.

I understand the solder stencil and rework will be risky with an already densely populated board though. Anyone else think this might fly?


legendary
Activity: 952
Merit: 1000
I think no one is bothering because there is no reason to chose BFL over Avalon.
The BFL chips cost about half as much per GH/s, assuming you're using a discount, of which there are thousands for sale dirt cheap. Also, they use less power per GH/s. So ya, I'd say there's a few reasons to look into it.
hero member
Activity: 546
Merit: 500
Intron said he would have a look at the BFL https://bitcointalksearch.org/topic/m.2583534

However, cant see there being much interest. Avalons are supposedly couple of weeks away, and if Bitfury chips come through at a reasonable price, it will wipe the floor with everything.
hero member
Activity: 546
Merit: 500
Anyone in here can do that please? ^^

I think no one is bothering because there is no reason to chose BFL over Avalon. BFL is far riskier and sketchier operation and people are going to stick with the more reliable company and technology given that their chips are pretty much the same price.
hero member
Activity: 756
Merit: 500
Anyone in here can do that please? ^^
Jump to: