ASICMINER: Entering the Future of ASIC Mining by Inventing It

[Catmoonglow]

This is exactly why I am going to be keeping my eye out for future IPO's. How many times have we said to ourselves, "Shit! If only I bought Microsoft/Apple/Google stock back when it was cheap...?" Who knows, the next Amazon may be just around the corner. This is an exciting time to be in bitcoin!

[necro_nemesis]

The relationship of power vs performance when overclocking/underclocking an ASIC is a significant factor when determining it's value. Power consumption can be significantly lowered underclocked greatly extending positive returns especially at high energy costs and low BTC value. I'd still like to see more on Gen3 test data to better understand it's life cycle.

[glendall]

Wow been quite the wait for these chips. I was thinking they were coming soon like 6 months back, and haven't even arrived yet.  ... But I'm pretty sure they'll be worth the wait.

[ning]

@Jutarul, Thank you.

[xhomerx10]

The relationship of power vs performance when overclocking/underclocking an ASIC is a significant factor when determining it's value. Power consumption can be significantly lowered underclocked greatly extending positive returns especially at high energy costs and low BTC value. I'd still like to see more on Gen3 test data to better understand it's life cycle.

 Please elaborate.  Where exactly is the value in under-clocking an ASIC?  IMHO, if you under-clock, you lose. 
Perhaps at the near end-of-useful-life of your ASIC it might be beneficial to under-clock if you have a time-of-day utility pricing scheme but you're talking about pennies by that point.  I believe the value of an ASIC is simply in it's power vs performance especially when the network difficulty rate is increasing exponentially.  Unless there is some sort of collusion among manufacturers (which would be illegal) to limit the distribution, then the sheer quantity of ASICs will negate any minor performance enhancements(?).  They need to run fast, low-power and NOW.

 Here's the GEN3 life cycle: over before it started. No test data required.

[necro_nemesis]

When energy consumption is higher than return, underclocking and undervolting can often be disproportionate to the reduction in performance thus altering the ratio of consumption to return and thereby extending the useful life of the installation.

[willBTC]

It seems that Spondoolies-Tech's chip is better than AM  3-generation ship on power consumption, but it's more expensive than AM.

[bitcoin.newsfeed]

It seems that Spondoolies-Tech's chip is better than AM  3-generation ship on power consumption, but it's more expensive than AM.

It seems like Spondoolies is actually selling products while ASICminer is only telling stories.

[raskul]

It seems that Spondoolies-Tech's chip is better than AM  3-generation ship on power consumption, but it's more expensive than AM.

I'm more than happy to pay for quality.

[jimmothy]

It seems that Spondoolies-Tech's chip is better than AM  3-generation ship on power consumption, but it's more expensive than AM.

Not exactly. SP hammer is 0.58w/gh at 7gh compared to AM 0.55w/gh at 12gh.

Also we don't know how much complete miners will cost but my guess is under $2/gh.

[necro_nemesis]

It seems that Spondoolies-Tech's chip is better than AM  3-generation ship on power consumption, but it's more expensive than AM.

Miners are concerned about the the chips minutest performance characteristics at the inflated prices paid at retail. This is factoring in what at retail is considered optimal in terms of cost, energy usage and heat management. If there's a cost benefit at the manufacturing level you can afford to add in additional chips and change the whole complexion of the equation. Just look at what was done to produce the S2 underclocking the BM1380 and adding in proportionately more ASICs to achieve the desired performance. They are willing to throw ASICs at the solution to improve efficiency. Cost at the manufacturing level can be crucial to altering the situation considerably. There additionally are cooling=space benefits to adding more ASICs and running at lower consumption/heat dissipation. This can be particularly exploited in mining and franchising right from the onset where markup is of no importance.

I would like to see the overall test data and as I mentioned when the ASIC specs were first released an explanation for the selectable higher clock range given what the ranges were stated at.

[xhomerx10]

It seems that Spondoolies-Tech's chip is better than AM  3-generation ship on power consumption, but it's more expensive than AM.

Miners are concerned about the the chips minutest performance characteristics at the inflated prices paid at retail. This is factoring in what at retail is considered optimal in terms of cost, energy usage and heat management. If there's a cost benefit at the manufacturing level you can afford to add in additional chips and change the whole complexion of the equation. Just look at what was done to produce the S2 underclocking the BM1380 and adding in proportionately more ASICs to achieve the desired performance. They are willing to throw ASICs at the solution to improve efficiency. Cost at the manufacturing level can be crucial to altering the situation considerably. There additionally are cooling=space benefits to adding more ASICs and running at lower consumption/heat dissipation. This can be particularly exploited in mining and franchising right from the onset where markup is of no importance.

I would like to see the overall test data and as I mentioned when the ASIC specs were first released an explanation for the selectable higher clock range given what the ranges were stated at.

As a miner, I can tell you that I am not concerned with this at all.  I want my miner delivered on time - period.  Minutia is meaningless to the individual miner.  One difficulty retarget late in delivering and you have negated all of these efficiencies which you are dreaming about.  Spoondoolies is shipping - where is AM? It's too little and too late.  Diff will double while AM dithers.

[Newar]

Why is this thread getting shorter... what posts are getting removed... the fact that anything is getting removed is bothering me.

You can check "the archive"

https://bitcointa.lk/threads/asicminer-entering-the-future-of-asic-mining-by-inventing-it.61268/

[necro_nemesis]

It seems that Spondoolies-Tech's chip is better than AM  3-generation ship on power consumption, but it's more expensive than AM.

Miners are concerned about the the chips minutest performance characteristics at the inflated prices paid at retail. This is factoring in what at retail is considered optimal in terms of cost, energy usage and heat management. If there's a cost benefit at the manufacturing level you can afford to add in additional chips and change the whole complexion of the equation. Just look at what was done to produce the S2 underclocking the BM1380 and adding in proportionately more ASICs to achieve the desired performance. They are willing to throw ASICs at the solution to improve efficiency. Cost at the manufacturing level can be crucial to altering the situation considerably. There additionally are cooling=space benefits to adding more ASICs and running at lower consumption/heat dissipation. This can be particularly exploited in mining and franchising right from the onset where markup is of no importance.

I would like to see the overall test data and as I mentioned when the ASIC specs were first released an explanation for the selectable higher clock range given what the ranges were stated at.

As a miner, I can tell you that I am not concerned with this at all.  I want my miner delivered on time - period.  Minutia is meaningless to the individual miner.  One difficulty retarget late in delivering and you have negated all of these efficiencies which you are dreaming about.  Spoondoolies is shipping - where is AM? It's too little and too late.  Diff will double while AM dithers.

Agreed, profit margin on every individual ASIC goes down with every tick of the clock. It's looking at AM as an enterprise over the course of months given the hardware they can build.

FWIW some second hand evidence of progress does filter it's way onto the net. RM posted this today. I'll leave it to your own interpretations as to what it indicates.

[klondike_bar]

FWIW some second hand evidence of progress does filter it's way onto the net. RM posted this today. I'll leave it to your own interpretations as to what it indicates.

heres what I can estimate from the picture:

- each group of three chips probably draws 30-60W  assuming 30A-rated components like seen in most designs (both bitfury and bitmain used the 30A TPS53355 regulator and the same inductor i can see in this image).  I would assume that each chip is able to draw up to 15W but this depends on operaing voltage (anyone have this spec?)

- assuming 12GH/chip is still correct, that's about 1w/GH in the shown configuration, possible with the ability to overclock further at a loss to efficiency

- 24 chips per single PCI-e power jack. assuming reasonable loads on power supply cables it would be unwise for this board to consume more than 250W without a risk that cheap wires will melt.  that means 10W/chip, or under 1w/GH

[necro_nemesis]

RM's results as of the 14th of April.

Results:

Board:one chip testing board
Frequency:360Mhz
Volt:0.72V
Hashrate per chip:11.52Ghash
Power consumption:6.375W per chip
Power consumption per Ghash:6.375/11.52=0.5539W/Ghash
After power supply changeover:0.5539/81% = 0.684W/Ghash（at blade)
Power consumption on wall:0.684/0.8 = 0.855W/G
Adding other components loss about 1KW/Thash

Tips:this result is not very accurate just for reference.

[_mr_e]

RM's results as of the 14th of April.

Results:

Board:one chip testing board
Frequency:360Mhz
Volt:0.72V
Hashrate per chip:11.52Ghash
Power consumption:6.375W per chip
Power consumption per Ghash:6.375/11.52=0.5539W/Ghash
After power supply changeover:0.5539/81% = 0.684W/Ghash（at blade)
Power consumption on wall:0.684/0.8 = 0.855W/G
Adding other components loss about 1KW/Thash

Tips:this result is not very accurate just for reference.

Is this good?

[CoinBomb]

as we've talked (in length!) when the results came out, its not stellar, but not terrible..more testing/optimisation required.

[necro_nemesis]

RM's results as of the 14th of April.

Results:

Board:one chip testing board
Frequency:360Mhz
Volt:0.72V
Hashrate per chip:11.52Ghash
Power consumption:6.375W per chip
Power consumption per Ghash:6.375/11.52=0.5539W/Ghash
After power supply changeover:0.5539/81% = 0.684W/Ghash（at blade)
Power consumption on wall:0.684/0.8 = 0.855W/G
Adding other components loss about 1KW/Thash

Tips:this result is not very accurate just for reference.

Is this good?

Good enough to plan to produce a significant number of them according to a recent report.

[bitcoin.newsfeed]

http://www.cybtc.com/article-900-1.html

?