How important is power efficiency in mining?

[ttul]

Here's a thought experiment for you all:

If a new mining technology comes along that provides the same MHash/s performance with 1/50th of the power consumption of the current state of the art, would everyone switch to this new mining technology? Also, assume that this new technology fits into the same approximate physical space as a similarly fast old-school mining rig, and that the cost is about the same per MHash/s.

A 5970 consuming 300W at full load is doing about 2MHash/s/W. The cost of the electricity at $0.10/kWh for this device is about $0.72/day.

If some new device comes along that consumes 6W (i.e. 1/50th), then the cost of electricity for this same quantum of mining falls to $0.014/day.

Annualized, the savings from the low power device would be $257, or roughly 1/3 the capital cost of the 5970.

Would this be enough to cause people to switch en masse to this hypothetical new mining technology?

[1714864742]

1714864742

[1714864742]

1714864742

[1714864742]

1714864742

[1714864742]

1714864742

[1714864742]

1714864742

[BitterTea]

Yes, I think. In fact, if Bitcoin catches on, I envision purpose built machines for mining and other related activities. For mining, that could come in the form of custom ASIC chips (like ArtForz had made), but mass produced. For spending, that could be in the form of a dedicated box that stores and distributes the block chain and allows for easy, safe storage and extraction of private keys.

[nster]

Here's a thought experiment for you all:

If a new mining technology comes along that provides the same MHash/s performance with 1/50th of the power consumption of the current state of the art, would everyone switch to this new mining technology? Also, assume that this new technology fits into the same approximate physical space as a similarly fast old-school mining rig, and that the cost is about the same per MHash/s.

A 5970 consuming 300W at full load is doing about 2MHash/s/W. The cost of the electricity at $0.10/kWh for this device is about $0.72/day.

If some new device comes along that consumes 6W (i.e. 1/50th), then the cost of electricity for this same quantum of mining falls to $0.014/day.

Annualized, the savings from the low power device would be $257, or more than 1/2 the capital cost of the 5970.

Would this be enough to cause people to switch en masse to this hypothetical new mining technology?

fixed

[qed]

If are asking "will FPGA replace video cards for mining?" the answer is a big no.

[nster]

If are asking "will FPGA replace video cards for mining?" the answer is a big no.

only because FPGA is too expensive and doesn't provide large hashrates.... if they were only a little more expensive and could provide 300 Mh/s+, it would be another story

[deadlizard]

Why is everyone so down on FPGA. of course it's not going to be economical on an individual basis.
The idea is to prototype the fastest possible FPGA design and then make cheap ASIC chips and build a multi-chip device to sell to miners. And not just bitcoin miners but the whole distributed computing community.

[true]

Would this be enough to cause people to switch en masse to this hypothetical new mining technology?

Considering most "miners" are people who are using equipment they already have, probably not. Dedicated miners, who knows? Probably?

I know that if an ASIC design came out that was fast and cheap enough, I might buy into it, but it'd have to be my view of fast and cheap enough. I don't have cards dedicated just to mining, they also serve the purpose of LAN gaming. But I have interest in BTC, so back to the first sentence...

[cdhowie]

Would this be enough to cause people to switch en masse to this hypothetical new mining technology?

Considering most "miners" are people who are using equipment they already have, probably not. Dedicated miners, who knows? Probably?

I know that if an ASIC design came out that was fast and cheap enough, I might buy into it, but it'd have to be my view of fast and cheap enough. I don't have cards dedicated just to mining, they also serve the purpose of LAN gaming. But I have interest in BTC, so back to the first sentence...

I agree.  I for one am just using my gaming card and don't really have the budget to buy anything new just for mining.

[rezin777]

Would this be enough to cause people to switch en masse to this hypothetical new mining technology?

Certainly! The casual miner might not make changes, but anyone who is interested in increasing their efficiency would.

[deadlizard]

mabye we should start an ASIC bounty 

[cdb000]

Would this be enough to cause people to switch en masse to this hypothetical new mining technology?

Considering most "miners" are people who are using equipment they already have, probably not. Dedicated miners, who knows? Probably?

I know that if an ASIC design came out that was fast and cheap enough, I might buy into it, but it'd have to be my view of fast and cheap enough. I don't have cards dedicated just to mining, they also serve the purpose of LAN gaming. But I have interest in BTC, so back to the first sentence...

In December and January I spent money on a bunch of 5870s. I also paid for some heavy power supplies. I already had mainboards and CPUs to run them. The resulting cluster runs 9 5870s and has mined a reasonable number of Bitcoins -- the outlay has been covered and profit has been made.

The increasing difficulty means that the rate of Bitcoin mining is slowing down - I currently expect to mine a block about every 30 hours. The 50 Bitcoins are worth roughly £22.8. The power used costs roughly £6 so each block makes a profit of about £16.8.

Difficulty tends to rise in steps of about 40%, so the time taken and power used to mine a block can be expected to rise to 42 hours+£8.4 then 58.5 hours+£11.76, 82.3 hours+£16.46, 115 hours+£23.05.

So, after 4 difficulty increments, assuming we see 40% increases each time, it will no longer be worth the cost in power for me to mine Bitcoins at their current value using the 5870s. Of course, as the expected return on investment in GPUs becomes smaller, we might expect that less people will be investing in mining hardware -- the difficulty might not actually increase so rapidly.

There may be an opportunity for someone to offer me and other miners a technology that can mine Bitcoins at a lower cost in power. Of course, whatever technology that turns out to be will have to have a low enough initial outlay that I would expect to be able to recover it in a reasonable amount of time, such expectations would be influenced by the knowledge that lots of people are likely to be buying and using the same new technology.

ASICs are expensive for small quantities, cheap for large. But how many SHA256 ASICs would one expect to be able to sell? And how much performance per Watt can actually be achieved?

[nster]

Would this be enough to cause people to switch en masse to this hypothetical new mining technology?

Considering most "miners" are people who are using equipment they already have, probably not. Dedicated miners, who knows? Probably?

I know that if an ASIC design came out that was fast and cheap enough, I might buy into it, but it'd have to be my view of fast and cheap enough. I don't have cards dedicated just to mining, they also serve the purpose of LAN gaming. But I have interest in BTC, so back to the first sentence...

In December and January I spent money on a bunch of 5870s. I also paid for some heavy power supplies. I already had mainboards and CPUs to run them. The resulting cluster runs 9 5870s and has mined a reasonable number of Bitcoins -- the outlay has been covered and profit has been made.

The increasing difficulty means that the rate of Bitcoin mining is slowing down - I currently expect to mine a block about every 30 hours. The 50 Bitcoins are worth roughly £22.8. The power used costs roughly £6 so each block makes a profit of about £16.8.

Difficulty tends to rise in steps of about 40%, so the time taken and power used to mine a block can be expected to rise to 42 hours+£8.4 then 58.5 hours+£11.76, 82.3 hours+£16.46, 115 hours+£23.05.

So, after 4 difficulty increments, assuming we see 40% increases each time, it will no longer be worth the cost in power for me to mine Bitcoins at their current value using the 5870s. Of course, as the expected return on investment in GPUs becomes smaller, we might expect that less people will be investing in mining hardware -- the difficulty might not actually increase so rapidly.

There may be an opportunity for someone to offer me and other miners a technology that can mine Bitcoins at a lower cost in power. Of course, whatever technology that turns out to be will have to have a low enough initial outlay that I would expect to be able to recover it in a reasonable amount of time, such expectations would be influenced by the knowledge that lots of people are likely to be buying and using the same new technology.

ASICs are expensive for small quantities, cheap for large. But how many SHA256 ASICs would one expect to be able to sell? And how much performance per Watt can actually be achieved?

Did you not see  that the 40% increases have STOPPED? it is stabilising. You will now see 10% increments

[ttul]

ASICs are expensive for small quantities, cheap for large. But how many SHA256 ASICs would one expect to be able to sell? And how much performance per Watt can actually be achieved?

The power consumed by a 2GHash/sec ASIC would not exceed 5W; this is based on inquiries with friends who make ASICs in their day jobs, and assuming a relatively straightforward, unoptimized design. So in other words, power requirements would be 50 - 100x lower than with GPUs.

[[Tycho]]

ASICs are expensive for small quantities, cheap for large. But how many SHA256 ASICs would one expect to be able to sell? And how much performance per Watt can actually be achieved?

The power consumed by a 2GHash/sec ASIC would not exceed 5W; this is based on inquiries with friends who make ASICs in their day jobs, and assuming a relatively straightforward, unoptimized design. So in other words, power requirements would be 50 - 100x lower than with GPUs.

5 times lower than GPUs, actually.
200 MH/s per single chip in plastic package.

[ttul]

ASICs are expensive for small quantities, cheap for large. But how many SHA256 ASICs would one expect to be able to sell? And how much performance per Watt can actually be achieved?

The power consumed by a 2GHash/sec ASIC would not exceed 5W; this is based on inquiries with friends who make ASICs in their day jobs, and assuming a relatively straightforward, unoptimized design. So in other words, power requirements would be 50 - 100x lower than with GPUs.

5 times lower than GPUs, actually.
200 MH/s per single chip in plastic package.

Where are you getting your figures? Mine are based on a 10M gate ASIC using a speedy cheap process like Hard Copy or eASIC.

[[Tycho]]

ASICs are expensive for small quantities, cheap for large. But how many SHA256 ASICs would one expect to be able to sell? And how much performance per Watt can actually be achieved?

The power consumed by a 2GHash/sec ASIC would not exceed 5W; this is based on inquiries with friends who make ASICs in their day jobs, and assuming a relatively straightforward, unoptimized design. So in other words, power requirements would be 50 - 100x lower than with GPUs.

5 times lower than GPUs, actually.
200 MH/s per single chip in plastic package.

Where are you getting your figures? Mine are based on a 10M gate ASIC using a speedy cheap process like Hard Copy or eASIC.

ArtForz is already mining on his own ASICs, ask him for details.
But he never shown photos of this setup, so it's not guaranted to be true.

[ttul]

Would you put down a deposit to get first dibs on an ASIC? If so, how much would you put down for a 2Ghash/s device that consumes 5W?

[MoonShadow]

Would you put down a deposit to get first dibs on an ASIC? If so, how much would you put down for a 2Ghash/s device that consumes 5W?

If we are talking about a working card with an asic (or any other method, for that matter) that gets those kind of specs, I'd be willing to depost $50 up front if I was fairly convinced that the offer was for real and the cards were ready for production.

[Gryph]

i'd put down $100 deposit, and purchase $400 per card if they were ready to run out of the box, software included.