BFL update-29 March 2013

[repentance]

I had planned on updating everyone with a video of a board hashing here in KC tonight, but I haven't been able to get that together yet, so I'm probably going to have to push it off until tomorrow. We are targeting a start of shipment next week, but I'm not quite ready to commit to that at the moment, given our past estimates. It's imminent, though. We may miss our power targets, that's been part of the hold up... we think there's a problem with the power consumption and we're trying to figure out where it's having an issue. However, in the interests of time, we are going to be shipping what we have and going back and fixing while we are shipping.

The power is still far less than any other unit, so it's not like it's something crazy or anything, but it's not 1w/GH and we're trying to locate the source of the power drain. the rest of the 6 wafers, we have been holding off on the last 5 layers for the rest of the chips to be sure we don't need to make a tweak in the metal layer due to the power issue. I think we've pretty much settled that the power issue is NOT in the chip. If the cooling becomes an issue, like I said, we'd scale it back and ship multiple units. Worst case is 195w, but the power systems aren't rated to handle that, so we'd back off the hashrate before we'd let it get that high. We will ship the purchased hashrate regardless of what it ends up costing us.

https://twitter.com/BFL_News

Obviously if they have to ship multiple units to meet the hashrate, fewer Batch 1 orders will be filled from the first 5,000 chips.

And some charity's going to be 1000 BTC better off.

https://forums.butterflylabs.com/content/123-bfl-offers-1000-btc-charity-if-they-miss-their-power-targets.html#comments

[Crystallas]

Ah, I was condensing the chat facts as you posted. My bad. Can a mod merge my post with this one?   TYIA

[redcomet]

Exciting, hope things go well for them.  And I hope the power issue is sorted before they ship mine later this year...

[TwinTurbo]

When they say purchased hash rate, do they mean 40 or 60? If memory serves me correctly they were originally advertised to deliver 40 GH, then upped to 60 GH when competition arose.

None of this surprises me.

[Crystallas]

When they say purchased hash rate, do they mean 40 or 60? If memory serves me correctly they were originally advertised to deliver 40 GH, then upped to 60 GH when competition arose.

None of this surprises me.

I'm guessing it's tied to the unit purchased.
ie: Jalepeno is the 4.5GH
Little SC (30)
Single SC (60)
MiniRig SC (1200)

[TwinTurbo]

When they say purchased hash rate, do they mean 40 or 60? If memory serves me correctly they were originally advertised to deliver 40 GH, then upped to 60 GH when competition arose.

None of this surprises me.

I'm guessing it's tied to the unit purchased.
ie: Jalepeno is the 4.5GH
Little SC (30)
Single SC (60)
MiniRig SC (1200)

Oops, I should have been more specific. In the very beginning, the Single SC's were advertised as 40 GH. They then increased it.

[veryveryinteresting]

When they say purchased hash rate, do they mean 40 or 60? If memory serves me correctly they were originally advertised to deliver 40 GH, then upped to 60 GH when competition arose.

None of this surprises me.

I'm guessing it's tied to the unit purchased.
ie: Jalepeno is the 4.5GH
Little SC (30)
Single SC (60)
MiniRig SC (1200)

Oops, I should have been more specific. In the very beginning, the Single SC's were advertised as 40 GH. They then increased it.

The MiniRig SC is (1500)

[Dargo]

A great leap in progress, glad to hear it.

[Fuzzy]

Worst case is 195w

Is that for a 60 GH Single SC? That can't be right, that's garbage compared to the 60w originally advertised.

[MrTeal]

Worst case is 195w

Is that for a 60 GH Single SC? That can't be right, that's garbage compared to the 60w originally advertised.

Well, Avalon is pulling 620W for their 67GH/s units, or 108MH/J. 195W for a 60GH/s Single would be 308MH/J. Just based on process node you'd expect BFL to be about 2.9x as efficient as Avalon, so those numbers aren't completely out of line. The only reason that 1J/GH was reasonable was BFL implementing a full custom layout as opposed to a standard cell. Hopefully their problems are just with the board and it's not the chips drawing the extra power.

[||bit]

Cool. Seems like it will be better to have the less efficient ones since you get more hardware in return. Sounds backwards - I know. But when/if the power drain issue is resolved, maybe there will be a fix that can be done at home. Then you can increase the clock back.

[rudrigorc2]

so the avalon party is over? 

[Red_Evil]

Maybe but Bfl doesnt say directly the ahipped now or ?

[Frizz23]

Worst case is 195w

Is that for a 60 GH Single SC? That can't be right, that's garbage compared to the 60w originally advertised.

Their estimates for their FPGAs were way off too. So 195W for a Single SC sounds plausible.

[NEO2012]

Worst case is 195w

Is that for a 60 GH Single SC? That can't be right, that's garbage compared to the 60w originally advertised.

Their estimates for their FPGAs were way off too. So 195W for a Single SC sounds plausible.

YEAH AND IF THEY WOULD SHIP ANITHING THIS YEAR WOULD BE OWSOME

REMEMBER THEY HAVE 1000 CHIPS

NOW THE WILL BE DIVIDED IN 2

SO BASICALLY 5OO CHIPS

ONE ORDER OF SC  = 16 CHIPS

THEY WILL SHIP 100 ORDERS OUT OF  25000   
HOPEFULLY BEFORE MAY
JOSH SHOULD FED TO ALIGATORS FOR THE LIEND SCUMM HE IS

[johnyj]

Looking forward, lots of uncertainty:

BFL has 75,000 chips, each doing 8Gh/s, that is 600T, plus 260T ASIC miner and 100T Avalon, 40T GPU&FPGA, network will top 1PH by autumn, 17X difficulty increase

Suppose one single hashes 4.5 BTC per day at current difficulty, by September, it will hash 0.265 BTC per day, back to the performance of a GPU 2 years ago

For winter we might see helveticoin's ?PH 28nm ASIC products

[Frankie Delaney]

Looking forward, lots of uncertainty:

BFL has 75,000 chips, each doing 8Gh/s, that is 600T, plus 260T ASIC miner and 100T Avalon, 40T GPU&FPGA, network will top 1PH by autumn, 17X difficulty increase

Suppose one single hashes 4.5 BTC per day at current difficulty, by September, it will hash 0.265 BTC per day, back to the performance of a GPU 2 years ago

For winter we might see helveticoin's ?PH 28nm ASIC products

All for the exact same 3600 coins per day. Money well spent.