ASIC power consumption estimates

[Jutarul]

Moore's Law and higher nm processes.

True, but say you had a basic $1200 budget:

2010: ~10-15Mh/s
2011: ~1400Mh/s
2012: ~2200Mh/s (achievable with cheap GPUs)
2013: >50000Mh/s

As much as Moore's law etc will have an impact, I don't think we'll ever see hashrate/$ go up >20x in a single year again. Hope not anyway 

yep. ASICs are endgame.

[jamesg]

Moore's Law and higher nm processes.

True, but say you had a basic $1200 budget:

2010: ~10-15Mh/s
2011: ~1400Mh/s
2012: ~2200Mh/s (achievable with cheap GPUs)
2013: >50000Mh/s

As much as Moore's law etc will have an impact, I don't think we'll ever see hashrate/$ go up >20x in a single year again. Hope not anyway 

yep. ASICs are endgame.

I never meant to suggest that a 20x increase would happen again. But doubling every 18 months is still an interesting thought. Plus, we are seeing that most ASICs are on old nm processes, so they can be improved more quickly.

[Inaba]

Why does no on believe me!  The BFL Minirig QC will be out RSN!

[kano]

The singles ended up using 4 times as much power per Megahash as BFL claimed they would initially. What's to stop that happening again?

I usually learn lessons the hard way, but once they are learned, they are not repeated. I hope it is the same with BFL.

...

Actually ...

This time they have to be at least 10x better than they were last time ... do people usually get that much better per iteration?

Before was 400%

If this time they are 10x better at estimating, i.e. only 40%, then the 1.5kW becomes 2.1kW = 19.1A on 110V .........

[Bogart]

The singles ended up using 4 times as much power per Megahash as BFL claimed they would initially. What's to stop that happening again?

I usually learn lessons the hard way, but once they are learned, they are not repeated. I hope it is the same with BFL.

...

Actually ...

This time they have to be at least 10x better than they were last time ... do people usually get that much better per iteration?

Before was 400%

If this time they are 10x better at estimating, i.e. only 40%, then the 1.5kW becomes 2.1kW = 19.1A on 110V .........

[MrTeal]

Or they run two 1200W PSUs instead of trying to source a 2kW model

[Meatball]

Moore's Law and higher nm processes.

True, but say you had a basic $1200 budget:

2010: ~10-15Mh/s
2011: ~1400Mh/s
2012: ~2200Mh/s (achievable with cheap GPUs)
2013: >50000Mh/s

As much as Moore's law etc will have an impact, I don't think we'll ever see hashrate/$ go up >20x in a single year again. Hope not anyway  

Yeah, I don't think you can use Moore's law in this case.  Moore's law considers the same type of technology (CPU's) increasing in effiiciency/power over time.  We're actually looking at jumps to new types of technology.  I think you need to compare the jump in power/efficiency from CPU/GPU/FPGA/ASIC mining to something more akin to the jump from Horses to Cars to Aircraft...

[VeeMiner]

• BFL - 1 watt/Gh +- 10% source
• Avalon - 2-10 watts/Gh source
• ASICMINER - 6 watts/Gh waiting on source
• bASIC -???/Gh
• DeepBit "Reclaimer" -???/Gh

The ASICMINER source says 4.2 Joule/Ghash:
https://bitcointalk.org/index.php?topic=91173.msg1215501#msg1215501

(BTW, fix your units, it's either Watt/Ghash/s or Joule/Ghash  </science-nazi>)

I really hope that BFL can deliver

[dunand]

Avalon announced a maximum of 400w for 66Gh/s.

We can narrow the estimate for Avalon to : 2-6 watts/Gh

[jamesg]

Avalon announced a maximum of 400w for 66Gh/s.

We can narrow the estimate for Avalon to : 2-6 watts/Gh

Done.

[kaerf]

from cablepair:

bASIC's will do 54Gh/s at under 100watts

http://www.flickr.com/photos/88427916@N07/8075593408/in/photostream


EDIT: the description has been modified with power numbers removed.

[creativex]

Because BFL is using a 65nm process and Tom is using 130nm? That's a 4x difference in power use right there.

Actually, I'm not sure what Tom is using. Maybe I'm thinking of Avalon. Anyone know for sure?

Tom is using 90nm and has further stated that the bASIC 1 will run on a barrel type connector or molex, this means under 110w.

https://www.bitcoinasic.net/index.php?route=product/product&product_id=51

Given the complexities/risks associated with designing a 45 or even 65nm chip I think a 90nm ASIC was an excellent choice for a product introduction.

[jamesg]

Looks like Tom has posted power estimates after all.

https://www.btcfpga.com/forum/index.php?topic=4.msg132#msg132

We are expecting the 27Gh/s units to use between 50-60 watts and the 54Gh/s units to use between 100-120 watts give or take

this is estimated data - and will not be completely correct but it gives you a ballpark and as close as an estimate as our competitor friends have on their units

Quoted here for posterity. I have also updated the OP.

[jamesg]

More info has come out about BFL ASIC. 

https://forums.butterflylabs.com/content.php/125-BFL-ASIC-Update

[Morblias]

Josh said 1st week of December for shipping. Tom also said 1st week for the bASIC. Are people sweating now over their who will ship first bets? 

[CoinHoarder]

Josh said 1st week of December for shipping. Tom also said 1st week for the bASIC. Are people sweating now over their who will ship first bets? 

It seems that they will start shipping so close together that it won't even matter. Except for maybe the people that are very first in line.

[MrTeal]

Josh said 1st week of December for shipping. Tom also said 1st week for the bASIC. Are people sweating now over their who will ship first bets? 

I would guess only the very early BFL customers who were relying on being first. Most bASIC customers went into it expecting to ship after BFL.

[gyverlb]

It seems that they will start shipping so close together that it won't even matter. Except for maybe the people that are very first in line.

There's the reliability of the estimation to account for too:

• Tom said Nov/Dec from the start and seems still on track. He delivered the modminer quads in a timely manner too from what I can tell.
• BFL originally said October then November and now December and has a past history of optimism.

BFL seems to have far more customers being the first announcing ASICs. So even with more hands to handle the task and an optimistic early start Tom customers as a group might get their ASICs before the BFL customers do and overall see a better ROI. So the luckiest (first delivered) BFL customer might have the best ROI of all but a random customer might be better off with Tom's offer.

I hope we'll have a clear picture at the end of December.

[ninjaboon]

So guys...If I ordered a BFL Mini Rig SC today, when can I expect it to reach my house (SE Asia) ?

[PuertoLibre]

Oh, by the way, the prototype picture has been released.



Courtesy of: www.butterflylabs.com

Re-posted under the FAIR USE ACT for discussion purposes.

It is unknown (as of yet) if this is simply a prototype for the case design or if it is a fully working prototype.

If it is the latter, then obviously the burning questions is: Did it meet the power and efficiency ratio of 1Gh per watt? I'll let others ask....