Let's recap on what we've seen in the past few months

[grod]

I looked at the Xilinx excel sheet yesterday. Did you?

No, actually.  I used digikey's retail prices with the assumption they're a viable ongoing concern with sustainable profit margins and extrapolated.   That has worked for me before, and I have a feeling the numbers you were looking at aren't that far off from my estimates.

edit: I haven't verified that $75000 is true for all FPGA revisions though, but the point stands. $5.26M/year is absolutely a market into which someone can spend hundreds of thousands if they believe they can capture a large portion of it.

Yup, his "millions" estimate is way off.  Half a million is more than enough to produce a run of devices cost competitive with current GPUs which require a PCIe slot to operate, but which would yield 90% gross margins at the point GPUs are no longer profitable.   In fact, a quarter million (aka 32,000 bitcoins) may be.

Anyone holding a giant stockpile of bitcoins could simultaneously drive adoption of such a product and increase their total amount by doing just this.

Also, it's 5.25M a year right now, but there is potential of wider adoption and a larger revenue stream in the future.

[1714115626]

1714115626

[1714115626]

1714115626

[1714115626]

1714115626

[Synaptic]

I looked at the Xilinx excel sheet yesterday. Did you?

No, actually.  I used digikey's retail prices with the assumption they're a viable ongoing concern with sustainable profit margins and extrapolated.   That has worked for me before, and I have a feeling the numbers you were looking at aren't that far off from my estimates.

edit: I haven't verified that $75000 is true for all FPGA revisions though, but the point stands. $5.26M/year is absolutely a market into which someone can spend hundreds of thousands if they believe they can capture a large portion of it.

Yup, his "millions" estimate is way off.  Half a million is more than enough to produce a run of devices cost competitive with current GPUs which require a PCIe slot to operate, but which would yield 90% gross margins at the point GPUs are no longer profitable.   In fact, a quarter million (aka 32,000 bitcoins) may be.

Anyone holding a giant stockpile of bitcoins could simultaneously drive adoption of such a product and increase their total amount by doing just this.

Also, it's 5.25M a year right now, but there is potential of wider adoption and a larger revenue stream in the future.

I said ASIC.

None of you seem to understand the cost to produce an ASIC.

You also don't seem to realize the risk averse nature of capital holders, and the Bitcoin market is shit compared to other investment opportunities.

Bitcoin is a scam. No-one in their right mind is going to spend any significant amount of money to see the scam through to it's completion.

[da2ce7]

I said ASIC.

None of you seem to understand the cost to produce an ASIC.

You also don't seem to realize the risk averse nature of capital holders, and the Bitcoin market is shit compared to other investment opportunities.

Bitcoin is a scam. No-one in their right mind is going to spend any significant amount of money to see the scam through to it's completion.

The market cap of bitcoin is over 60M USD.  Since more than half the coins ever to be made are still yet to be made, it is completely feasible that a group of people will want to mine/spend the 1.2M USD capital needed in producing an ASIC miner.

100% return of investment could easily be reached in one year... even if the bitcoin price drops to $1.

[grod]

Ok, show YOUR math.  For the record, you were claiming neither asic nor fpga would ever be viable not long ago.

Here is the pricing from Xilinx's EasyPath website.  $13 for 30,000 logic cells.  That's about $50 for 100 mhash (it would be far more cost effective to house 60 watts worth of chips per board than the current 7.5-15 watts) so call it *a minimum* 800 mhash device for a $400 production cost, board included.  Actual hashing output would likely be a bit higher.

How much were people paying for a 800 mhash rig recently?  You could retail these with a 100% margin and still sell tons.

The startup cost is not millions.  It's $75,000.

[ElectricMucus]

This will not happen, just enter structured asics into google and read how they are already obsolete because of higher integration of fpgas.

[da2ce7]

Ok, show YOUR math.  For the record, you were claiming neither asic nor fpga would ever be viable not long ago.

Hmmm... I was?  Please point me to any thread that I say that ASIC are anything but insanely profitable...

The maths goes:

200K for development + testing.

Do a small run of 200,000 1Gh/s chips @ around 4M transistors each.

They cost $1 ea.  So that is 200K + 600K setup.

Total 800K for a dedicated ASIC run.

Building into board, and packaging maybe 1$ each.  So total 1.2M

You will put 20 boards in a box and sell the box.  20Gh/s @ 40W.   Sale price of $1000 per box.
That will mean that you produce 10M income on say 1.2M investment.  To break even you only need to sell 20% of your production...  It is insanely profitable.

[da2ce7]

This will not happen, just enter structured asics into google and read how they are already obsolete because of higher integration of fpgas.

When you post something like this you sound like a complete retard... I'm sorry, they both serve purposes.

[defxor]

Ok, show YOUR math.  For the record, you were claiming neither asic nor fpga would ever be viable not long ago.

Hmmm... I was?  Please point me to any thread that I say that ASIC are anything but insanely profitable...

He was talking to Synaptic. The guy that accuses everyone of not being able to do simple math, yet shows none himself and all the math that's been done concludes the exact opposite of what he claims

[Synaptic]

Ok, show YOUR math.  For the record, you were claiming neither asic nor fpga would ever be viable not long ago.

Hmmm... I was?  Please point me to any thread that I say that ASIC are anything but insanely profitable...

He was talking to Synaptic. The guy that accuses everyone of not being able to do simple math, yet shows none himself and all the math that's been done concludes the exact opposite of what he claims

Not at all. I don't need to post any math because I seriously don't give a fuck to prove anything. My opinions will always be borne out to be accurate.

FPGAs will never displace GPUs, and common Bitcoin ASICs will never exist though there may be some hobbyist chips that pop-up (like ArtFortz's mythological chips) though all dubious, and of little consequence.

[ElectricMucus]

This will not happen, just enter structured asics into google and read how they are already obsolete because of higher integration of fpgas.

When you post something like this you sound like a complete retard... I'm sorry, they both serve purposes.

Even when they are almost the same thing?
Come on they spare the extra silicon for the universal routing fabric, fused rom instead of sram or flash. Yeah cirtaily cheaper at high volume, a good thing if you want to build consumer cameras but for something like bitcoin nfw.
And fpgas already penetrate market areas where you would expect stuctured asics, the loss of reconfigurability and optimizations at a later date is just too much of a disadvantage. What happens if cryptanalysis discovers something new, do you throw away your chips or hope that the improvements aren't that earth shaking?

Seriously, I could imagine some effort for proprietary cryptocurrency by some entity some day which might choose this path but bitcoin is foss and thus subjected to potential change and evolution. period.

better now?

[da2ce7]

better now?

No.  Most certainly not.  You sound even more noob now than before.

We need to be talking "performance per watt" numbers here.

The best fpga may be able to produce 200Mh/s @ 4W or equiv. 1Gh/s @ 20 W (Still far more efficient than a GPU).
While a fully custom asic would produce 1Gh/s @ 2W.   This is 10x more energy efficient.

Secondly, let’s look at the cost 'per-chip' once you exclude the one-time tooling costs.

5 fpga's @ 200Hh/s fpga's would account for $20 each... so you are looking at $100 total cost, for the best case.
1 asic @ 1Gh/s would cost around $1 each to manufacture (aka cost of a 4M transistor chip on modem wafer technology).

You see... for the same hashing speed.  Fpga's work out to be 100x more expensive to manufacture, and 10x more expensive to run than fully custom asic's.   Also.  To make matters worse for your augment, I'm taking the best-case numbers for the fpga... and the average-case numbers for the asic.

I do not understand how the can be 'almost the same thing' when there manufacturing costs are 100x and performance is 10x worse.  (On first analysis, that would sound like quite a large difference).

Well anyway; it was fun showing you to be a retard again.

[ElectricMucus]

The fact that you'll have to resort to insults makes your case so much weaker though.

But you are comparing apples and oranges here. Your $1.2mil just doesn't work out with anything but structured asics which will go no where near your numbers.
Feel free to provide sources if really think you're right.

[newMeat1]

FPGA costs have come down like crazy in the past few months. Our dual boards do 200 Mhash/s for $610. If we are able to hit an order size of 1000, cost will come down by $50 or so. If we get the automatic overclocking software up and running, then it will be something like 280 Mhash for $550. At 15 watts or less. Sounds competitive to me.

My opinion is, we aren't far from seeing FPGA order sizes hitting 1000.

[kazoo]

Not so sure that I wish to wade into the toxic soup, but...

Someone please correct me if I misunderstood how far ArtForz was able to take this, BUT, to the best of my understanding:

ArtForz, a German person, successfully ordered and implemented his(?) own ASIC design. It set him back ~$60,000, and for his pains, he was mining about a 20-25% of all the bitcoin production, somewhere around May or June of this year.

The 60,000 might have been euros. The 20-25% might (and probably was) some other specific quantity.

Was this bunk, or did it happen? I read this somewhere out on the interwebs, but so what?

[mizike29]

To say bitcoin will go back up is just wishful thinking.  Nothing can predict whats going to happe with bitcoin and it has been a slow long downslide that I just hope stops, stabilizes finally.  I dont see 20s again, but I liked it being semi stable in the 10 to 12 range that it say for awhile.

[Synaptic]

Not so sure that I wish to wade into the toxic soup, but...

Someone please correct me if I misunderstood how far ArtForz was able to take this, BUT, to the best of my understanding:

ArtForz, a German person, successfully ordered and implemented his(?) own ASIC design. It set him back ~$60,000, and for his pains, he was mining about a 20-25% of all the bitcoin production, somewhere around May or June of this year.

The 60,000 might have been euros. The 20-25% might (and probably was) some other specific quantity.

Was this bunk, or did it happen? I read this somewhere out on the interwebs, but so what?

I call bullshit.

[nmat]

ArtForz, a German person, successfully ordered and implemented his(?) own ASIC design. It set him back ~$60,000, and for his pains, he was mining about a 20-25% of all the bitcoin production, somewhere around May or June of this year.

I believe he only tweaked the GPU client. I never heard of ASICs mentioned. Here's the wiki article:

ArtForz is a regular on #bitcoin-dev. He was the first person to mine with GPUs, using private mining code that he wrote himself. He once held a high percentage of network computation power (~25%) but reports[1] that he now controls less than 1%. He claims that his mining software is still more efficient than all other GPU miners. He uses a heavily-modified version of Bitcoin; notably, he requires less transaction fees and he doesn't perform IsStandard checks.

[osmosis]

When the hashing power has decreased and rigs were sold. But yeah people are in denial, that's what gives scam artists alternate cryptocurrencies a free hand.

Deepbit disagrees with you.  Over the past month the hash rate there (measured, not estimated) has gone from 5 terahash to 5.5 terahash.   It was going up 50 gigahash a day as the price went below $10, and has since flatlined -- but it has not decreased in any measurable way.

*MY* remaining mining hardware is on craigslist, I don't deny it.   Already sold two 5830s for $120 each, hoping for a difficulty drop or price pop to sell the last 2.  But that won't be the general case until FPGA and ASIC miners make video cards as obsolete as CPUs.

Why do you people still believe there's going to be bitcoin ASICs? How many goddamn times does it have to be drilled into your head that the economics will NEVER work out enough to develop one until BTC are trading for well over $50/coin steadily...

You are obviously making stuff up, because the price of BTC has nothing to do with profitability if you dont even mention how many bitcoins are being generated per day. Mining profitability has always been based on the difficulty level that has been expanding with competition. In otherwords, because of the free market which maintains the equilibrium...mining always has the same profitability at entrance point. Thats why early miners made a heck of a lot more IF THEY HELD THEM...because they were speculating on the future of bitcoins.

[Synaptic]

When the hashing power has decreased and rigs were sold. But yeah people are in denial, that's what gives scam artists alternate cryptocurrencies a free hand.

Deepbit disagrees with you.  Over the past month the hash rate there (measured, not estimated) has gone from 5 terahash to 5.5 terahash.   It was going up 50 gigahash a day as the price went below $10, and has since flatlined -- but it has not decreased in any measurable way.

*MY* remaining mining hardware is on craigslist, I don't deny it.   Already sold two 5830s for $120 each, hoping for a difficulty drop or price pop to sell the last 2.  But that won't be the general case until FPGA and ASIC miners make video cards as obsolete as CPUs.

Why do you people still believe there's going to be bitcoin ASICs? How many goddamn times does it have to be drilled into your head that the economics will NEVER work out enough to develop one until BTC are trading for well over $50/coin steadily...

You are obviously making stuff up, because the price of BTC has nothing to do with profitability if you dont even mention how many bitcoins are being generated per day. Mining profitability has always been based on the difficulty level that has been expanding with competition. In otherwords, because of the free market which maintains the equilibrium...mining always has the same profitability at entrance point. Thats why early miners made a heck of a lot more IF THEY HELD THEM...because they were speculating on the future of bitcoins.

Shit son,

You're out of your fucking league...

[TraderTimm]

Shit son,

You're out of your fucking league...

I guess paid trolls are too much for people who believe in freeing themselves from financial tyranny. So, how much are you getting? Because if you aren't, you'd better offer your psychopathic services to a three letter agency real quick, son.

But again, you might just be one of those misguided people who like taking a big steamy e-dump on anything you don't agree with.

Gosh, which one could it be?