NotFuzzyWarm
Legendary
Offline
Activity: 3766
Merit: 2677
Evil beware: We have waffles!
|
|
April 12, 2017, 12:27:24 AM |
|
Hey Guy! Please post how you arrived at "15% power savings". What are the components of the practical loss versus theoretical gain? How much worse is the highest achievable overclock or lowest achievable undervolt? How much worse is the useable manufacturing yield? Is rather what I was thinking about. Ok, AB allows for somewhat faster processing of a possible hash by being able to internally come up with a new hash to work on vs getting new work from a pool/server right? But - since most of the power is consumed when the ASICs actually clock to crunch the numbers, where does energy saving come in? When they are twiddling their thumbs waiting for new work power consumption drops quite a bit. so I would think it rather evens out, no?
|
|
|
|
2112
Legendary
Offline
Activity: 2128
Merit: 1073
|
|
April 12, 2017, 12:41:16 AM Last edit: April 12, 2017, 01:52:57 AM by 2112 |
|
Ok, AB allows for somewhat faster processing of a possible hash by being able to internally come up with a new hash to work on vs getting new work from a pool/server right?
But - since most of the power is consumed when the ASICs actually clock to crunch the numbers, where does energy saving come in? When they are twiddling their thumbs waiting for new work power consumption drops quite a bit. so I would think it rather evens out, no?
ASICBOOST whitepaper computes savings purely by counting one-bit logical operations (gates) in a purely combinatorial (fully unrolled) implementation. That's how they arrived at their numbers: | Number of colliding work items | 1 | 2 | 4 | 5 | 8 | 16 | | Gain in percent | 0.00 | 12.50 | 18.75 | 20.00 | 21.9 | 23.4 |
I have no non-bullshit information about any actual implementation. I presume that in a pure-ASICBOOST chip the savings would come from less combinatorial logic. In a mixed, switchable chip (that can mine both raw and boosted) the power savings would come from disabling clock on a portion of circuitry that computes terms that are common under boosting. Edit: added table formatting
|
|
|
|
NotFuzzyWarm
Legendary
Offline
Activity: 3766
Merit: 2677
Evil beware: We have waffles!
|
|
April 12, 2017, 01:48:53 AM |
|
Bitmain's variant of ASICBoost https://www.google.com/patents/CN105245327A?cl=enAnd yer right. The abstract mentions disabling a part of the logic during some operations. Aside from that is wayyyy beyond my depth.
|
|
|
|
-ck
Legendary
Offline
Activity: 4242
Merit: 1644
Ruu \o/
|
|
April 13, 2017, 10:30:35 PM |
|
Thanks for that Guy. I wasn't convinced they were using it anyway but that adds much useful data.
|
Developer/maintainer for cgminer, ckpool/ckproxy, and the -ck kernel 2% Fee Solo mining at solo.ckpool.org -ck
|
|
|
Guy Corem (OP)
Donator
Legendary
Offline
Activity: 1414
Merit: 1051
Spondoolies, Beam & DAGlabs
|
|
April 14, 2017, 03:32:36 AM Last edit: April 14, 2017, 04:13:05 AM by Spondoolies-Tech |
|
Thanks for that Guy. I wasn't convinced they were using it anyway but that adds much useful data. They are not. They didn't even know about the covert method Gmaxwell found out until he published. They did careful analysis afterwards of what needed, and got to the conclusion that they will need to tapeout custom 130nm and add it to the miner board to support this method.
|
|
|
|
-ck
Legendary
Offline
Activity: 4242
Merit: 1644
Ruu \o/
|
|
April 14, 2017, 05:22:42 AM |
|
Thanks for that Guy. I wasn't convinced they were using it anyway but that adds much useful data. They are not. They didn't even know about the covert method Gmaxwell found out until he published. They did careful analysis afterwards of what needed, and got to the conclusion that they will need to tapeout custom 130nm and add it to the miner board to support this method. Which makes their violent reaction to a BIP proposal that makes asicboost not work somewhat baffling. I guess they were more objecting to the intent in the move's aggression rather than the actual BIP, but it made most of the community convinced they were using it covertly.
|
Developer/maintainer for cgminer, ckpool/ckproxy, and the -ck kernel 2% Fee Solo mining at solo.ckpool.org -ck
|
|
|
Guy Corem (OP)
Donator
Legendary
Offline
Activity: 1414
Merit: 1051
Spondoolies, Beam & DAGlabs
|
|
April 14, 2017, 05:42:59 AM |
|
Thanks for that Guy. I wasn't convinced they were using it anyway but that adds much useful data. They are not. They didn't even know about the covert method Gmaxwell found out until he published. They did careful analysis afterwards of what needed, and got to the conclusion that they will need to tapeout custom 130nm and add it to the miner board to support this method. Which makes their violent reaction to a BIP proposal that makes asicboost not work somewhat baffling. I guess they were more objecting to the intent in the move's aggression rather than the actual BIP, but it made most of the community convinced they were using it covertly. Try to read the BIP proposal from their perspective. They distrust and dislike the author. He makes bogus inflated claims. The entire community believes they hold unfair advantage. I hardly blame them. Yes, their PR is awful.
|
|
|
|
klondike_bar
Legendary
Offline
Activity: 2128
Merit: 1005
ASIC Wannabe
|
|
April 15, 2017, 04:16:31 PM |
|
Thanks for that Guy. I wasn't convinced they were using it anyway but that adds much useful data. They are not. They didn't even know about the covert method Gmaxwell found out until he published. They did careful analysis afterwards of what needed, and got to the conclusion that they will need to tapeout custom 130nm and add it to the miner board to support this method. Which makes their violent reaction to a BIP proposal that makes asicboost not work somewhat baffling. I guess they were more objecting to the intent in the move's aggression rather than the actual BIP, but it made most of the community convinced they were using it covertly. so let me understand: asicboost essentially is an added logicgate at chiplevel that determines whether a specific nonce is worth hashing, and if not, the chip doesnt do any other work for that cycle (hence the savings). So its not any faster at mining, it just saves some power?? if thats the case, it actually serves to decentralize mining by making minimal savings for the big (cheap-power) miners but significant savings for smaller miners who have more expensive electricity. ie: someone paying $0.03/kwh saves maybe $5/month, while someone paying $0.10/kwh saves $15/month
|
|
|
|
Guy Corem (OP)
Donator
Legendary
Offline
Activity: 1414
Merit: 1051
Spondoolies, Beam & DAGlabs
|
|
April 15, 2017, 07:46:13 PM |
|
Thanks for that Guy. I wasn't convinced they were using it anyway but that adds much useful data. They are not. They didn't even know about the covert method Gmaxwell found out until he published. They did careful analysis afterwards of what needed, and got to the conclusion that they will need to tapeout custom 130nm and add it to the miner board to support this method. Which makes their violent reaction to a BIP proposal that makes asicboost not work somewhat baffling. I guess they were more objecting to the intent in the move's aggression rather than the actual BIP, but it made most of the community convinced they were using it covertly. so let me understand: asicboost essentially is an added logicgate at chiplevel that determines whether a specific nonce is worth hashing, and if not, the chip doesnt do any other work for that cycle (hence the savings). So its not any faster at mining, it just saves some power?? if thats the case, it actually serves to decentralize mining by making minimal savings for the big (cheap-power) miners but significant savings for smaller miners who have more expensive electricity. ie: someone paying $0.03/kwh saves maybe $5/month, while someone paying $0.10/kwh saves $15/month No, your description of AB is completely wrong. Please research online, there are enough resources these days.
|
|
|
|
Guy Corem (OP)
Donator
Legendary
Offline
Activity: 1414
Merit: 1051
Spondoolies, Beam & DAGlabs
|
|
April 15, 2017, 07:56:53 PM |
|
I have been told that an attack has been found on Cuckoo hash recently (trading memory for more work). I didn't look into it.
|
|
|
|
Guy Corem (OP)
Donator
Legendary
Offline
Activity: 1414
Merit: 1051
Spondoolies, Beam & DAGlabs
|
|
April 16, 2017, 03:37:10 AM |
|
PSA: We released the following miners only when we were operational: SP10, SP20, SP30, SP31 and SP35. If you see anything else offered online, it's a scam. We designed but never completed the SP50.
|
|
|
|
|
Guy Corem (OP)
Donator
Legendary
Offline
Activity: 1414
Merit: 1051
Spondoolies, Beam & DAGlabs
|
|
April 17, 2017, 03:38:32 AM |
|
It's a scam and scam site
|
|
|
|
jborkl
|
|
April 17, 2017, 09:38:07 PM |
|
I enjoyed the mining rigs while they were around! lost track of what has been going around here
|
|
|
|
Zeta0S
Legendary
Offline
Activity: 1890
Merit: 1031
|
|
April 18, 2017, 09:29:22 AM |
|
I still mis Spondoolies-Tech, I wish there miners where still whit us. Spondoolies-Tech Where well build quality miners I ask my self what went wrong
BitmainTech is now (almost) allown and that is not good at all.
|
|
|
|
edgar
Legendary
Offline
Activity: 1848
Merit: 1001
|
|
April 18, 2017, 10:50:27 AM Last edit: April 18, 2017, 01:17:52 PM by edgar |
|
I still mis Spondoolies-Tech, I wish there miners where still whit us. Spondoolies-Tech Where well build quality miners I ask my self what went wrong
BitmainTech is now (almost) allown and that is not good at all.
no, its fine. we've been assured that the rampant dishonesty, delayed shipping, dusty/rusty rigs & corner cutting was all pure coincidence and in no way intentional on the part of bitmain or any of his current and/or future business associates. haven't we, guy...
|
|
|
|
Unacceptable
Legendary
Offline
Activity: 2212
Merit: 1001
|
|
April 18, 2017, 10:58:23 AM |
|
I still mis Spondoolies-Tech, I wish there miners where still whit us. Spondoolies-Tech Where well build quality miners I ask my self what went wrong
BitmainTech is now (almost) allown and that is not good at all.
no, its fine. we've been assured that the rampant dishonesty & corner cutting was a pure coincidence and in no way intentional on the part of bitmain or any of his current and/or future business associates. It makes little difference...ASIC mining at home is mostly over so Bitmain being a monopoly has little meaning.... There are other whales mining away & between them all,they will take Bitcoin where THEY want it to go...you little folks have no voice in the matter anymore
|
"If you run into an asshole in the morning, you ran into an asshole. If you run into assholes all day long, you are the asshole." -Raylan Givens Got GOXXED ?? https://www.youtube.com/watch?v=9KiqRpPiJAU&feature=youtu.be"An ASIC being late is perfectly normal, predictable, and legal..."Hashfast & BFL slogan
|
|
|
BitcoinPSU
Newbie
Offline
Activity: 67
Merit: 0
|
|
April 18, 2017, 11:11:17 AM |
|
It's a scam and scam site Also check the image, they mirrored the front, but forgot to mirror the names of the ports back
|
|
|
|
Guy Corem (OP)
Donator
Legendary
Offline
Activity: 1414
Merit: 1051
Spondoolies, Beam & DAGlabs
|
|
April 21, 2017, 02:56:01 PM |
|
|
|
|
|
-ck
Legendary
Offline
Activity: 4242
Merit: 1644
Ruu \o/
|
|
April 21, 2017, 09:30:06 PM |
|
Well this is 'offtopic' for this thread but since it's your thread you can take it whichever direction you want. Bitmain's on that list? Did Jihan pull his head out of his arse and realise he was one person against the rest of the world at last? Now if only he'd do the same thing for bitcoin - but then he'll argue the only reason he agreed on the ltc roundtable was because they agreed to a block size increase when they need it which is laughable since ltc has almost completely empty blocks and may never need it. Oh well, I'll reserve further discussion for bitcoin/
|
Developer/maintainer for cgminer, ckpool/ckproxy, and the -ck kernel 2% Fee Solo mining at solo.ckpool.org -ck
|
|
|
|