BTCFPGA bASIC updated from 54GH/s to 72HG/s

[Devious]

With two clusters on the larger unit, it will conservatively produce 72GH/s.  Yes, this is very good news - a 33% increase in hashing power!  Yes, this means the 27G unit now becomes a 36G unit.

From https://www.btcfpga.com/forum/

Those ordering from Tom are happy!-)

[beekeeper]

With two clusters on the larger unit, it will conservatively produce 72GH/s.  Yes, this is very good news - a 33% increase in hashing power!  Yes, this means the 27G unit now becomes a 36G unit.

From https://www.btcfpga.com/forum/

Those ordering from Tom are happy!-)

But most likely this will add 3 extra weeks to deliver. One week redesigning board layout to accommodate more chips (this includes power circuit redesign), one week (they said 8 days) to produce pcb, one week to assemble everything.

PS: I'm a customer.

[Devious]

But most likely this will add 3 extra weeks to deliver. One week redesigning board layout to accommodate more chips (this includes power circuit redesign), one week (they said 8 days) to produce pcb, one week to assemble everything.

PS: I'm a customer.

I'm a customer as well, I'm not sure what to think on the timeline,

Are you guys tracking on delivery targets?
Admittedly, its going to be tight - we lost a couple precious days to the Thanksgiving Holiday.  Tom expects to learn a lot more on Tuesday regarding schedule.

Hopefully there will be good news Tuesday.


Has BFL mentioned anything about it's shipping schedule?

[beekeeper]

Hopefully there will be good news Tuesday.


Has BFL mentioned anything about it's shipping schedule?

BFL also wait for news Tuesday, their last chip version coming from fab.

[bcpokey]

Hmm, so they are selling 72GH/sec for $1069.99 now?

[beekeeper]

Hmm, so they are selling 72GH/sec for $1069.99 now?

Yes, but with a delay.

[hahahafr]

HOLY SHIT. I support you BFL, but this is bad news for us BFL supporters.

[Meatball]

Heh, how great is it that there's a 'performance war' between companies before anything has even been released?  I imagine each ASIC vendor sees a rush of pre-orders every time one of these announcements come out.

[squeept]

Competition is good. Everybody wins.

However, this announcement essentially includes the statement "we actually have nothing done whatsoever." So until both companies actually deliver, there will be no benefits.

[Starlightbreaker]

2 words:
FUCK YES.

3 words.
SUCK IT, BFL

[ice_chill]

BFL's ASIC can match that if they want. At 40Ghash their ASIC was running at 500Mhz, and they said it can do 1Ghz but not all chips will reach it. So 72GHash is possible.

[Starlightbreaker]

BFL's ASIC can match that if they want. At 40Ghash their ASIC was running at 500Mhz, and they said it can do 1Ghz but not all chips will reach it. So 72GHash is possible.

and that's still a conservative estimate.

These initial numbers are conservative estimates, but we expect that firmware and software updates will be able to iteratively increase hashing power as well.

so it might be even faster than 72 gh/s

in normal circumstances, i strongly dislike luke-jr, but at times, he can work some magic.

[bcpokey]

BFL's ASIC can match that if they want. At 40Ghash their ASIC was running at 500Mhz, and they said it can do 1Ghz but not all chips will reach it. So 72GHash is possible.

Not sure how many in this thread read the post referenced by OP, as most people are seeing "ASIC war!" but it looks like this is simply a result of ASIC stability.

Where is the design at now?
The design of the smaller bASIC unit grew from 2 ASICs per board to 6 in a cluster, each ASIC producing 4.5GH/s (safely) for a total of 27 GH/s.  As you probably know, digital computers/microprocessors operate using the binary (base-2) number  system.  This somewhat naturally leads to architectures that most efficiently have their components (such as I/O signals, communications paths, memories, etc.) in quantities of powers of two - 2, 4, 8, 16, and so on.

***Important - please read closely***
It is for this reason that we are going to grow our hashing cluster to 8 ASICs.  With two clusters on the larger unit, it will conservatively produce 72GH/s.  Yes, this is very good news - a 33% increase in hashing power!  Yes, this means the 27G unit now becomes a 36G unit.  These initial numbers are conservative estimates, but we expect that firmware and software updates will be able to iteratively increase hashing power as well.

Might be BS, or might be true, that it's simply a matter of the fact that 8 is more stable than 6, and that's why they chose to go this route, which will end up with some happy customers. Rather than being anything to do with outdoing BFL (though that doesn't hurt).

EDIT: Reminder though that chips were apparently literally dying at overclock speeds, so before you have wet dreams of overclocking, keep in mind the horrors of murdering your ASIC chip.

[Starlightbreaker]

it'll most likely delay the production though.

well, there goes beginning of december..here's hoping close to the end of december.

[crazyates]

Wait, so their chips were running ~14GH/s, with the cheaper unit running 2 chips (2 x 13.5 = 27), and the more expensive unit running 4 of these (4 x 13.5 = 54). They found out these were clocked too high, so they had to decrease performance. To compensate, they added an extra chip to each module. How many chips is it now? 3/6? 3 chips would sound about right (3 x 12 = 26) for the smaller one, and 6 chips for the larger one (6 x 12 = 72).

A) So they added 50% more chips, but only decreased the clock rate by ~10$, and somehow this is expected to use about the same amount of power? That doesn't make sense.

B) In order to do this, someone else mentioned that they have had to redesign their board, and redevelop their firmware (which may or may not be even completed yet), which just adds time. They're not even close to assembling hardware, are they?

C) A 50% increase in the chips used per unit means that whatever initial batch of chips they ordered, they can only ship 2/3 the same number of products. If they ordered 3,000 chips, they just went from being able to ship 750 units to 500 units.

D) I am loving the increase in hash speed at the same price point. Competition at it's best!

[Starlightbreaker]

8 chips x 9gh/s = 72.

[kano]

BFL's ASIC can match that if they want. At 40Ghash their ASIC was running at 500Mhz, and they said it can do 1Ghz but not all chips will reach it. So 72GHash is possible.

and that's still a conservative estimate.

These initial numbers are conservative estimates, but we expect that firmware and software updates will be able to iteratively increase hashing power as well.

so it might be even faster than 72 gh/s

in normal circumstances, i strongly dislike luke-jr, but at times, he can work some magic.

Eh what?

[Starlightbreaker]

BFL's ASIC can match that if they want. At 40Ghash their ASIC was running at 500Mhz, and they said it can do 1Ghz but not all chips will reach it. So 72GHash is possible.

and that's still a conservative estimate.

These initial numbers are conservative estimates, but we expect that firmware and software updates will be able to iteratively increase hashing power as well.

so it might be even faster than 72 gh/s

in normal circumstances, i strongly dislike luke-jr, but at times, he can work some magic.

Eh what?

what?

luke-jr?

aside from shitting on other people that disagrees with him.

[bce]

Wow, a 33 percent increase over the previously stated speed on this yet-to-be-seen product?   That's fast!

[Easy2Mine]

The smaller unit is upgraded from 6 asic chips in a cluster to 8 asic chips for a total of 8*4.5 GH/s=36 GH/s
The bigger unit has 2 clusters of 8 asic chips. 2*8*4.5 GH/s=72 GH/s

Each BTCFPGA asic chip can do a maximum of 14 GH/s, but it won't last long at that speed.
6 chips per cluster are more difficult to implement than 8 chips per cluster.
I personnally think that they took the advice of the board designing company to change it to a 8 cluster or 2*8 cluster board.
Cablepair has already say that he is hiring one of the best board designing company to do the job.
In my humble opinion, I think they will have a greater chance to deliver on time if they go for the 8 chips per cluster board instead the 6 chips per cluster board.
But who am I?
Only time will tell.