masterOfDisaster
|
|
October 15, 2013, 08:24:22 PM |
|
Just because no one posted it before - and although some might have noticed it already - here is the data sheet that has been released by bitmine today: https://bitmine.ch/wp-content/uploads/2013/10/Bitmine-CoinCraft-A1.pdf (and here is the corresponding news article: https://bitmine.ch/?p=2226) Seems like the ASIC works more or less like expected Abstract of some charachteristics: Hashing power of 25 GH/s in nominal and up to 40 GH/s in Turbo mode. Power usage of 0.35 W/GH in low power, 0.6 W/GH in nominal and 1 W/GH in Turbo mode. Supply voltage of 0.5V in low power, 0.65 V in nominal and 0.75 V in Turbo mode.
|
|
|
|
GodfatherBond
|
|
October 15, 2013, 08:35:45 PM |
|
Just because no one posted it before - and although some might have noticed it already - here is the data sheet that has been released by bitmine today: https://bitmine.ch/wp-content/uploads/2013/10/Bitmine-CoinCraft-A1.pdf (and here is the corresponding news article: https://bitmine.ch/?p=2226) Seems like the ASIC works more or less like expected Abstract of some charachteristics: Hashing power of 25 GH/s in nominal and up to 40 GH/s in Turbo mode. Power usage of 0.35 W/GH in low power, 0.6 W/GH in nominal and 1 W/GH in Turbo mode. Supply voltage of 0.5V in low power, 0.65 V in nominal and 0.75 V in Turbo mode.
"The nominal hashing power is 25 GH/s and not 20 GH/s as initially announced, this is due to some last minute optimizations that have been made with the design, a free 25% increase!" Very nice
|
|
|
|
RoadStress
Legendary
Offline
Activity: 1904
Merit: 1007
|
|
October 15, 2013, 08:36:52 PM |
|
Just because no one posted it before - and although some might have noticed it already - here is the data sheet that has been released by bitmine today: https://bitmine.ch/wp-content/uploads/2013/10/Bitmine-CoinCraft-A1.pdf (and here is the corresponding news article: https://bitmine.ch/?p=2226) Seems like the ASIC works more or less like expected Abstract of some charachteristics: Hashing power of 25 GH/s in nominal and up to 40 GH/s in Turbo mode. Power usage of 0.35 W/GH in low power, 0.6 W/GH in nominal and 1 W/GH in Turbo mode. Supply voltage of 0.5V in low power, 0.65 V in nominal and 0.75 V in Turbo mode.
This was already on their website before that pdf.
|
|
|
|
masterOfDisaster
|
|
October 15, 2013, 09:07:21 PM |
|
As far as I remember there have been some characteristics based on simulations. I thought this would be a data sheet based on metered characteristics - at least that is what I took it for... The only former news article dealing with numbers has been https://bitmine.ch/?p=2049 so far. Not having "Electrical specifications" included and looking at the revision history makes me think this might be just another estimation... But at least [quote https://bitmine.ch/?p=2049] "Running at 1 GHz clock for over 32GH/s" [/quote] is slightly different from 40 GH/s... ...as is 0.35 J/H compared to [quote https://bitmine.ch/?p=2049] under 0.3 J/H [/quote] in low power mode. Yet these are still only numbers. Will be interesting to see the characteristics when the devices are ready...
|
|
|
|
giorgiomassa (OP)
|
|
October 16, 2013, 10:15:31 AM |
|
As far as I remember there have been some characteristics based on simulations. I thought this would be a data sheet based on metered characteristics - at least that is what I took it for... The only former news article dealing with numbers has been https://bitmine.ch/?p=2049 so far. Not having "Electrical specifications" included and looking at the revision history makes me think this might be just another estimation... We will release an updated version of the datasheet soon, as for now we focused mainly on the protocol and the packaging, the electrical characteristics are being lined up currently
|
|
|
|
Kartaiv
Member
Offline
Activity: 61
Merit: 10
|
|
October 16, 2013, 04:43:46 PM |
|
As far as I remember there have been some characteristics based on simulations. I thought this would be a data sheet based on metered characteristics - at least that is what I took it for... The only former news article dealing with numbers has been https://bitmine.ch/?p=2049 so far. Not having "Electrical specifications" included and looking at the revision history makes me think this might be just another estimation... We will release an updated version of the datasheet soon, as for now we focused mainly on the protocol and the packaging, the electrical characteristics are being lined up currently Are the wattage numbers calculated at the wall or at the chip?
|
|
|
|
RoadStress
Legendary
Offline
Activity: 1904
Merit: 1007
|
|
October 16, 2013, 05:16:24 PM |
|
As far as I remember there have been some characteristics based on simulations. I thought this would be a data sheet based on metered characteristics - at least that is what I took it for... The only former news article dealing with numbers has been https://bitmine.ch/?p=2049 so far. Not having "Electrical specifications" included and looking at the revision history makes me think this might be just another estimation... We will release an updated version of the datasheet soon, as for now we focused mainly on the protocol and the packaging, the electrical characteristics are being lined up currently Are the wattage numbers calculated at the wall or at the chip? At chip of course
|
|
|
|
aerobatic
|
|
October 16, 2013, 05:20:21 PM |
|
As far as I remember there have been some characteristics based on simulations. I thought this would be a data sheet based on metered characteristics - at least that is what I took it for... The only former news article dealing with numbers has been https://bitmine.ch/?p=2049 so far. Not having "Electrical specifications" included and looking at the revision history makes me think this might be just another estimation... We will release an updated version of the datasheet soon, as for now we focused mainly on the protocol and the packaging, the electrical characteristics are being lined up currently What is the performance of the chip when running in Low Power mode? The spec states quite clearly the performance in Normal and Turbo mode, but there's no mention of low power performance. Also, please can you say how the chips will be cooled? In Turbo mode, each chip will run at 40 Watts... can that be cooled passively (just a heatsink) or will it require a 'cpu cooler' approach with heatsink and fan on each chip like an intel cooler? -- Jez
|
|
|
|
Puppet
Legendary
Offline
Activity: 980
Merit: 1040
|
|
October 16, 2013, 05:41:14 PM |
|
Also, please can you say how the chips will be cooled? In Turbo mode, each chip will run at 40 Watts... can that be cooled passively (just a heatsink) or will it require a 'cpu cooler' approach with heatsink and fan on each chip like an intel cooler? -- Jez
There is no way you can cool a (tiny) 40W chip passively without an expensive monster cooler. You will need a fan. Typically anything over 10W needs a fan of some sort.
|
|
|
|
aerobatic
|
|
October 16, 2013, 05:50:56 PM |
|
There is no way you can cool a (tiny) 40W chip passively without an expensive monster cooler. You will need a fan. Typically anything over 10W needs a fan of some sort.
lets say you want to have a 400-500 GH unit to compete with the usual offerings from BitFury, KnC, HF, CT etc.. so youd need 12 or more of the chips running in turbo mode... generating 400-500 Watts of heat... So... can you just plop a heatsink on each of the 12 coincraft chips, and then run a fan or two at the end of the enclosure to move the air past all the chips...? OR... do you need active cooling, ie: a Fan on a heatsink on each of the 12 chips? (thats starting to get BIG). 40 Watts is an awkward amount of heat. its not so small, like a bitfury chip.. where you dont need anything... and its not so large, like an Intel chip (or more likely, a KnC chip) that you need a heatsink with fan strapped to it... so its somewhere in the middle.. and the cooling solution for the coincraft chips will be interesting because it doesnt seem to fit with either the 'lots of cool chips' nor the 'a few big hot chips' bracket. it seems somewhere in the middle... and thats why im curious about the cooling solution
|
|
|
|
Puppet
Legendary
Offline
Activity: 980
Merit: 1040
|
|
October 16, 2013, 05:58:48 PM |
|
lets say you want to have a 400-500 GH unit to compete with the usual offerings from BitFury, KnC, HF, CT etc.. so youd need 12 or more of the chips running in turbo mode... generating 400-500 Watts of heat...
So... can you just plop a heatsink on each of the 12 coincraft chips, and then run a fan or two at the end of the enclosure to move the air past all the chips...? OR... do you need active cooling, ie: a Fan on a heatsink on each of the 12 chips? (thats starting to get BIG). Doesnt really matter, as long as you have sufficient surface area (heatsink) per chip and enough airflow over the heatsink. Cooling a single 40W chip isnt exactly difficult, but if you cram 12 of them on a small PCB, you will face the same problems as everyone else. 40 Watts is an awkward amount of heat. its not so small, like a bitfury chip.. where you dont need anything... A single bitfury chip consumes only ~2W solution for the coincraft chips will be interesting because it doesnt seem to fit with either the 'lots of cool chips' nor the 'a few big hot chips' bracket. it seems somewhere in the middle... and thats why im curious about the cooling solution The avalon approach should work just fine, large shared heatsinks with case fans providing the airflow. ALternatively the BFL approach could work just as well, grouping the chip closely on the PCB and putting a traditional HSF on that.
|
|
|
|
RoadStress
Legendary
Offline
Activity: 1904
Merit: 1007
|
|
October 16, 2013, 07:50:24 PM |
|
Just because no one posted it before - and although some might have noticed it already - here is the data sheet that has been released by bitmine today: https://bitmine.ch/wp-content/uploads/2013/10/Bitmine-CoinCraft-A1.pdf (and here is the corresponding news article: https://bitmine.ch/?p=2226) Seems like the ASIC works more or less like expected Abstract of some charachteristics: Hashing power of 25 GH/s in nominal and up to 40 GH/s in Turbo mode. Power usage of 0.35 W/GH in low power, 0.6 W/GH in nominal and 1 W/GH in Turbo mode. Supply voltage of 0.5V in low power, 0.65 V in nominal and 0.75 V in Turbo mode.
I don't know anything about chip design, but if they put extra work or anything extra in order to have the Low power and the Turbo mode i think it was useless. Let's say i buy 10 chips. Why would i want to run them in low power for 250Gh/s at 87.5W instead of running at 400Gh/s at 400W? The difference power cost is 1.12$ per day while of the difference of the bitcoins mined is a lot bigger and i don't really find a reason to run your chips at low or at nominal power. If you can't pay for the power cost at Turbo mode i don't think it's worth to keep them mining at low power mode. Am i wrong?
|
|
|
|
|
Puppet
Legendary
Offline
Activity: 980
Merit: 1040
|
|
October 16, 2013, 09:18:23 PM |
|
Am i wrong?
Not yet. But you will be wrong ~6-12 months from here when electrical efficiency will become pretty much all that matters. Bitmine certainly didnt waste that effort, once the network reaches dozens of PH, they will still have a chip thats worth buying, assuming they can hit those numbers.
|
|
|
|
RoadStress
Legendary
Offline
Activity: 1904
Merit: 1007
|
|
October 16, 2013, 09:45:18 PM |
|
Am i wrong?
Not yet. But you will be wrong ~6-12 months from here when electrical efficiency will become pretty much all that matters. Bitmine certainly didnt waste that effort, once the network reaches dozens of PH, they will still have a chip thats worth buying, assuming they can hit those numbers. Maybe i wasn't clear enough. TL;DR version is: If you can't afford to keep the chips running at turbo mode then you are loosing more money running them in low power mode. At least that's what i think.
|
|
|
|
Puppet
Legendary
Offline
Activity: 980
Merit: 1040
|
|
October 16, 2013, 09:54:14 PM |
|
Maybe i wasn't clear enough. TL;DR version is: If you can't afford to keep the chips running at turbo mode then you are loosing more money running them in low power mode. At least that's what i think.
In that case: you are wrong . What you probably meant to say is something like "If you can't afford to keep the chips running at turbo mode its going to take a VERY long time to break even on your investment". That would be true, at least with todays hardware prices.
|
|
|
|
RoadStress
Legendary
Offline
Activity: 1904
Merit: 1007
|
|
October 16, 2013, 10:09:15 PM |
|
Maybe i wasn't clear enough. TL;DR version is: If you can't afford to keep the chips running at turbo mode then you are loosing more money running them in low power mode. At least that's what i think.
In that case: you are wrong . What you probably meant to say is something like "If you can't afford to keep the chips running at turbo mode its going to take a VERY long time to break even on your investment". That would be true, at least with todays hardware prices. I would like more than one opinion.
|
|
|
|
GandalfG
Sr. Member
Offline
Activity: 259
Merit: 250
Dig your freedom
|
|
October 16, 2013, 10:44:28 PM |
|
40W from chip 12x12mm Most heat go as usual in QFN package by vias to bottom PCB layer. If you seat 10 chip on board in Eurocard form-factor (100x160mm) you need heatsink and enough air flow to dissipate 400W. Its serious amount of heat if you consider 24h/day continuous work.
For comparison Overclocked 7970 about 250-280W Only HD6990 and GTX590 monster consume near 400W If believe BFL spec Monarch take only 350W for 600Gh/s
|
Want to say thanks? 16ragydppe9QFRVhrdwEUjgfMS7KCfEFGY
|
|
|
papamoi
|
|
October 17, 2013, 04:48:53 AM |
|
so guys
what is your roi and when are you planning to deliver?
did you reduce your price of 40% because the difficulty have increased of the same or??
|
|
|
|
Puppet
Legendary
Offline
Activity: 980
Merit: 1040
|
|
October 17, 2013, 06:59:51 AM |
|
I would like more than one opinion.
Who needs opinions when you can have math? Power figures at the wall will be higher than whats stated for the chip but lets assume the stated power figures apply at the wall with an electricity price of $0.25/KWH. Assume BTC exchange rate of $150 Once difficulty reaches 10B, per month each chip: in turbo mode, 40W 40GH, it will mine $9.17 and cost $7.2 in electricity -> $1.97 profit in normal mode, 15W 25GH it will mine $5.73 and cost $2.7 in electricity -> $3.03 profit Once difficulty reaches 20B in turbo mode the chip will operationally run at a loss, while in normal mode it will still be profitable. Depending on BTC exchange rate and your electricity cost (and PSU/PCB efficiency), the point where one mode makes more sense than the other changes, but there will always be a difficulty level were turbo mode is less profitable than normal mode unless you have free electricity. Also note that if you need to actively cool your house/datacenter, effective electricity cost could be 2-3x higher and the point where normal mode makes more financial sense could be achieved at a difficulty of only a few billion, a difficulty level we are likely to see early 2014.
|
|
|
|
|