Bitcoin Forum

It seems as soon as cablepair announced his bASIC product, BFL started shifting gears.

Then we get the bASIC announcement that his devices can do 54 GH/s instead of the original 27.

Now Tom stated in his initial announcement he was going to be very conservative, so his announcement of double makes sense in that light.

BFL comes out and doubles all their product estimates.

Okay, so this is great for the network, but does anyone know the technical aspects of this and how this is really working?

Are we talking more chips or just underestimating chip hash rates ?

Is this the beginning of the price war everyone discussed earlier ?

Really curious on this.

Tom mentioned plans to have 2 designs (2nd one with double the amount of chips). Thats why there are now 2 bASICs once with 27gh/s and another with 54gh/s. These are minimum values and based on conservative theoretical estimates not actual prototypes so there is still a possibility that these values could go up.

BFL may be on actual prototype specs (but they haven't confirmed this) merely said that the "plan all along was to update the firmware to allow for increased hashrate" should a price war or competition get intense.

Do you have a link to that quote? Just that it wasn't on his website last week.

https://bitcointalk.org/index.php?topic=79637.msg1229735#msg1229735

https://bitcointalk.org/index.php?topic=79637.msg1229735#msg1229735 (https://bitcointalk.org/index.php?topic=79637.msg1229735#msg1229735)

Ah ok on the 28th September.

I was hoping he had made that quote before 'Avalon' spec's were announced.

Nope. Though I am not sure how much the AValon specs influenced these decisions since there is only an initial run of 300 and the 2nd mass run will be $2k

https://bitcointalk.org/index.php?topic=114339.msg1234039#msg1234039

In a nutshell or laymen terms.  It's the GH/s war.  The 3 big ones (BFL, bASIC, NgZhang) are all competing with each other.

giggedy giggedy

they compete, we win :)

Well, No, Actually its all relative IMO, The only people that will win are the first to receive.

Correction: the only people that will win are the ones that make the hardware.

+1+1+1+1

 Correction: the only people that will win are the ones that create the machines that creates the hardware. ;D

On BFL's part its just business smarts.

They overestimated specs on a previous line and took a hammering over it.

This time around they seem to have underestimated to prevent a reoccurance.

This has had the added ability of keeping competitors in the dark.

bASIC et al put their cards on the table, BFL showed theirs and came out on top.

The GHz battle continues apace :)

BFL just increased the value of their conservative estimate.. there was no actual testing of physical hardware... Right now bASIC is ahead of $/Gh/s ... but what matters most is the power draw, which no one has actual numbers of.. just estimates based on simulations (ie bullshit)

Haha! Classic, I would love to see that phrase in a companies literature somewhere :)

Correction: the only people that will win are the ones that do not lose money ;)

back to the main thread:

chips performance is well known in the design phase, is known for the amount of logic gates and the technological process. Doubling the speed of chips in the production phase is either magic or clever marketing procedure. Firmware change can have a big effects in the FPGA but not ASIC.  if I'm wrong, correct me giving reasonable arguments.

Another similarity between Bitcoin and gold. During the gold rush era the first people to profit were the ones who sold the miners and other gold rush followers the tools and supplies they needed.

+1 This is spot on and I said something like this in a different thread. The only way they could increase out put from what I understand is to increase chips on the board. That should not be hard it would just increase the power draw, and require a moderate change. Cost to a have a PCB custom designed? 100 USD.......

I think some are missing my point.

Irregardless of whether the output of the chip was known as a hard fact at the design phase, any monkey can state a smaller number as a public estimate.

This allows for a *hey presto! extra output!* occurance without having to add more chips.

Ah yes, because it is common to understate your initial performance estimates by 50% then magically decide to increase them under competition.

I think it is simply very cheap to add chips.  If the profit margin on product A @10GH is $200, and one can instead switch over to produce B @15GH which has a margin of $175, why not?

The question is, why would a company do this?  Well, they started with product A because the margin is higher.  They switched to product B because competition came in.  Simple as that.

Remember, with ASIC's, the cost for each additional unit is very small after you've put down the costs for the manufacture of the first one.  So whether you order 1,000 or 100,000 doesn't make a ton of relative difference to your outlay for the chips.  The thing that becomes expensive in the long run at things that run up the price without adding performance like cases, USB cables, shipping, power cords.... etc...

Just imagine that AMD, increases the power of their chips by 50%. 7970, instead of 900Mhz -> 1350Ghz (without increasing power consumption), Nvidia goes bankrupt ;) It is a pity that AMD does not know the tricks such as BFL know;)
Maybe BFL should cooperate with AMD, we would have several times faster and super energy efficient graphics card ;)

I know that's how I buy all of my electronics lol The paper said it was a 30 inch t.v. but when I got there they gave me a 60 inch t.v., man I love Best Buy.

Chips are 60% cheaper when they get the die made and start churning them out. So that 300 dollar Stratix-3 they were buying has now became a 120 dollars. I even know a company that does chips for 0% NRE, they just charge you for the first full order upfront, If you were basing them off of a spartan-6 like the other guys are, the FPGA's cost 158.99 brand. 158.99/2 * 1000 "min order" = cost for first order. They send you 10 sample chips to put int he boards and test then once you guys agree they work they make and ship the rest.

well for BFL im sure they had the numbers set a little lower then it really was.

Personally I wish BFL would just be honest and stop playing pr games. They should just admit they aren't as far into fab as they have been saying and state that they redesigned their board to be more competitive with others.

Obvious lie is obvious.

And see this is the thing the vast majority fails to see, these asic's as far as everyone knows are designed after old low end model chips. When a designer coughs up the 10k plus for a virtex-7 and builds that to ASIC, all there products will be null.... 150k LE VS 2Mil LE lol

There's always a better product coming or possible. These aren't static. I mean when Apollo launched the computer it used to go the the moon and back is overpowered by modern watches now. In a couple of years these will be paperweights. In 10-15 years we'll have some sort of personal data device/phone/watch/whatever than can do circles around anything we have now.

The only real question is will you get a return on your investment? It's a game of chicken.

True

This is really interesting. Let's talk about that.

Well... during a gold rush, you want to be selling the pick-axes... when there is very little gold to find.

We know there are Bitcoins to find.  They're out there, just waiting for your nonces.  The whole thing is a gamble though because we're still in the mindset of taking a BTC and thinking of it as a small bundle of dollars that could potentially become a larger bundle of dollars.  This will change exponentially when people start transacting in BTC, and this is what the vast majority of people don't understand.

I'm sure if we tracked the part of the economy going into BTC right now, it would be next to zero.  That is fine... as long as it grows exponentially.  When we get to 1% of the world economy, we'll be just a hop, skip, and jump away from 100% of the world economy.  This might take 50 years... but so what.