Mining a block takes 10 minutes in average. But it may take 10 seconds or 10 weeks as well. Did anybody make a calculation, what's the probability that it will take a week to mine a single block under normal conditions (i. e. when difficulty is adequate to the total hashing power)?
I've seen many guesses and opinions about but a real calculation done by someone who understands the subject has not yet been done, or at least published. Bitcoin mining is a poisson distribution. It is far easier to just use a poisson distribution calculator (online, graphing calculator, mathematica). In 1 hour the expected # of events (blocks) is 6. The probability that there will be 1 or less is ~1.7% In 2 hours the expected # of events (blocks) is 12. The probability that there will be 1 or less is ~0.01% In 24 hours the expected # of events (blocks) is 144. The probability that there will be 1 or less is essentially 0 (4.198 E-61) |
|
|
|
your answer is yes. I remember a rep from hashfast saying that there would be overclocking temperature logic built into the chips themselves. They will adjust the hashrate and intensity if a lower temperature is detected. Best believe it makes a lot of sense to mine these in something akin to an icebox or freezer just to get a huge overclock.
Not a good idea. a) compressors on fridges are small. They are designed to keep cold stuff cold not remove a 1.2KW heat load. b) cooling below dew point = condensation onto components which probably don't enjoy getting wet. |
|
|
|
|
Updated to reflect increased HashFast order book (2.24 PH/s excluding IceDrill all batches combined). Reduced KNC Sept/Oct back to 0.6 PH/s (1500 reported units @ 300 TH/s average - mixed units). Estimated KNC Nov batch @ 2 PH/s.
|
|
|
|
It's hard to verify gold holdings because of security concerns and metal authenticity testing. Bitcoin holdings are easy to verify. Sign a message from the address. Wouldn't surprise me if these "Trusts" throw out some BS "security" "auditing" excuse for not revealing the actual address. I read the winkelvoss prospectus. No address in there that I could find. http://www.sec.gov/Archives/edgar/data/1579346/000119312513393903/d562329ds1a.htmWhy would a trust which doesn't yet exist and own no coins list an address? This is our empty Bitcoin addres .... Also anyone thinking it will be a single address needs to get their head checked. The ETF allows direct redemption I doubt the funds want to unlock a wallet with a single address containing potentially hundreds of millions of dollars someday for every redemption. Risk management and all that. I mean MtGox doesn't use a single address why would an ETF? |
|
|
|
Everything changed. The shareholders must suffer 3 more diff increases, each with 50 %. The price will fall at the same value. You know for a certainty that each of the next 3 difficulty changes beyond the currently pending one will be >50%? |
|
|
|
Let's be clear, I am no lawyer and I always follow the laws in all things.
This is just speculation. A company, not based in the USA, whose principals (CEO, etc.) is not living in the USA, and whose assets are not in the USA would be pretty hard to stop or prosecute if it was being traded on a networked, distributed exchange.
Don't let hubris cloud judgement. Outside the US isn't black and white, more like shades of gray. On one hand you have places like Somolia and on the other you have places like the UK whose government is very friendly to their runaway colony. Case in point. http://en.wikipedia.org/wiki/United_States_v._ScheinbergA company outside the US operating servers outside they US, with employees outside the US, regulated and registered outside the US, and owning bank accounts outside the US. The DOJ with the help of friendly courts in more than 14 countries seized over 87 bank accounts with assets totaling at least half a billion dollars. Not a single penny was in a US bank account but the DOJ got every last cent along with IP, patents, trademarks, domain names, software source code and sought more than $3B in damages (mostly settled out of court). Players might have their funds returned to them in 2014 if everything goes according to plan. |
|
|
|
what device would one use to connect the USB connector from the Sierra to an ethernet wall jack?
USB to Cat5/6 extender Something like this. http://www.monoprice.com/Product?p_id=6042&format=2It is still USB just on cat5 wire. You can't connect it to an ethernet switch. Cat5/6 is just a cable. Ethernet is a protocol. Ethernet runs on cat5 but with the right hardware so can a lot of other things (music, hdmi, usb, etc). |
|
|
|
Anybody wanna sell 2 or 3 BTC at $180 a piece? I'll pay using PayPal. Thanks.
Why would anyone want to sell $24 below market rate  Cause you get the cash now. Some don't wanna wait 1 week for coinbase. And as far as Mt. Gox, it's a few weeks and many cant get their $$$ out. And by the Time that week goes by they may lose much more if the price tanks. So I guess anyone who thinks the price may drop and wanna lock in $180 and get paid now I guess would go for this deal. If not then I'm gonna wait, a sell off has to come, way too much way too fast on no real news. I'm not gonna chase Bitcoin when I know how volatile it is. When you sell on coinbase (or any other exchange) you lock in the price at the current price. The fact that ACH will take 3-4 days to clear doesn't change the price sold at. |
|
|
|
Honestly I don't know what to expect as far as volume. It just seems like coins with less coins out and similar output have much higher dumps from pools than ixCoin. Like DogitalCoin, it only has 8 million coins out and there's massive multi pool dumping. The number of coins in supply has nothing to do with miners dumping. If miners only mine 13K coins per day how could they sell more than 13K coins per day? Counterfeit coins? |
|
|
|
You get Modules but you would have to purchase the empty rigs.
Or build/expand your own. Other than the processing boards/modules everything else is off the self. |
|
|
|
OT: DeathAndTaxes have you invested in HF ?
I have purchased a single BabyJet for evaulation. For full disclosure I also have purchased an evaluation KNC Saturn and Bitfury "Starter". I intend to experiment with immersion cooling and using the waste heat for hot water and heating applications. If/when I make a large purchase it will be in the future when hopefully hashrate is more predictable and prices ($/GH) are lower. For my needs the Hashfast boards seem to be optimal (I don't really care about cases, power supplies, or cooling units). The high energy density and abilitity to connect an arbitrary number of hashing boards to one host (potentially hundred feet away outside the immersion cooling tank) is a good fit. Well at least on paper. Lets see what they deliver. |
|
|
|
So when MPP kicks in and HF sends out bare chips, how is everyone planning on building out the rest of the components???
HashFast clarified it won't be bare chips. It will be processing boards. I believe the OEM water cooling unit HashFast is using is also available retail from Corsair. Any CPU waterblock should be compatible if one wants to build a custom system. The boards connect by USB so any embedded system (rPi, Beagle Board) or entry level PC can be used as a host. Power is by 6 pin PCIe connectors so any good quality PSU should be fine. |
|
|
|
... I don't know the terms of Hashfast contract or what their timeline with TMSC is. It is possible the date they gave was the latest possible date. Regardless once again the delay has nothing to do with the foundry. |
|
|
|
Yes HF may not be responsible for the delay ( still waiting for the pics of something that is not the chips), but for a customer they totally are. A customer does not care if the delay is by the foundry, by the pcb supplier or by any other.They want their units on time, as simple as that. I think you misunderstood what I was saying. The delay has NOTHING to with the foundry. Nothing. I was correcting that mistaken claim that the substrate is part of the raw wafer process. It isn't and even if it was it isn't something that HashFast would have any control over. They don't manufacture their own raw silicon ingots, cut them into wafers, polish them, etc. All that is done prior chip production. Yes HashFast is responsible for their customers. Period. I was just correct a false statement about this having something to do with the foundry.
Anyway, you are telling me that you would sign a contract with a foundry like TSMC for hundred thousands of dollars and that you wont have a fixed deadline and further compensation if they dont reach that date ? Kinda going off topic but if you want to know the contract usually provides TSMC a giant window like delivery in 6-8 weeks expected with penalty clause after 14-20 weeks. The foundry holds all the cards in the negotiation (especially for small players). If you don't like TSMC terms you can go to Global Foundry and they will give you also the same terms. If you don't like any foundy terms you can always build your own $8B 28nm FAB. Of course once again so there is no confusion an issue with the substrate has nothing to do with TSMC, fabrication, or the delivery time of the wafers. Per HF statement they expect the raw wafers from TSMC on time HOWEVER you can't cut, and package the wafers into final chips without the substrate. |
|
|
|
I like the idea of shipping Baby Jets in Sierra chassis (with one PSU) because it's now likely they will need at least two additional MPP mini-boards to achieve zero/positive ROI.
If this is cost prohibitive, HashFast could offer the Sierra chassis option as a paid upgrade.
Well, I was just not suggesting to ship the BJs into sierras chasis, I suggest shipping as sierras complete units that are expected to be shipped in late november. Back in time when they made the "3 days sale" you could get a sierra with x3 hasing power than a BJ at the same price ! But when I complained about this and some customers try to upgrade or change their BJ into a sierra unit, HF said a clear NO, saying that there was a month between those batches, more than enough time for a BJ to catch the sierra in profit terms. Now that the gap between the two batches seems to be really small, it makes no sense to ship the first batch customers a babyjet. - The chip and boards are the same - The power supply is the same - The cooling solution I suppose is also the same - The controller is the same The only difference is the case, which HF has not yet bought or produced for the BJ. Indeed there is still no render about it, just for the sierra. So why no purchase a bunch of server racks cases and make the things right HF ? This way if we roi you wont have to execute the MPP. Give the first batch customers a compensation for the delay, reward them for going first into this and compensate them for the error you made by offering the sierra units one month later and no saying a word in advance or giving them the chance to upgrade their orders. It is an interesting idea and hopefully one that HF is at least considering. Based on the news, blogs, and mockups the BabyJet does seem to be taking a back seat to the Sierra anyways. Going forward (Jan onward) with rising and falling $/GH prices a 400 GH/s unit is less interesting than a 1.2 TH one so I don't see the BabyJet being a large seller in 2014. It is almost a certainty that HashFast will need to pay out the MPP so each BabyJet orders means 1 to 4 more processing boards at some point in the future. Even a halfway step of upgrading all Babyjet orders to Sierra "chassis" and including 2 processing boards, the first one paid for and the second one being a "down payment" on the MPP would put customers in a better situation. Of course Hashfast doesn't HAVE to do any of this they can just stick to the terms of service but I am pretty sure they are smart enough to know that 2014 sales depend heavily on how they "make it right" for batch 1 and 2 customers. KNC for example despite some missteps probably will have no issue lining up batch 3, 4, 5, 20 sales in 2014. Then again a lot comes down to how many chips Hashfast will have. They have 550 batch 1 orders + 500 more for Icedrill project. Not sure if the Nov sales are part of the first "batch" of chips from the fab or not. What they can do may come down to how many chips they will be receiving from TSMC in the next week. If it is 20,000 chips they they have a lot of options. If it is 4,000 chips and it needs to cover batch 2 orders as well, then there is a lot less they can pull off. |
|
|
|
At 96 coins per block we should be seeing a lot more selling than we are now and at 1.5 cents it's worth something. 96 coins per block is only 13K per day. If 100% of miners sell instantly and not wait/hope for better prices it would only be 13K in volume. Volume is ~5K to 30K per day. What kind of volume would you expect is merge miners are just selling off as they mine? Miners dumping 200K per day everyday despite only mining collectively 13K? |
|
|
|
I check this thread every once in a while, mainly to wonder why an up and coming company would hire cypherdouche, official bitwhore for Half-Fast.
I skipped the last couple of pages, in an effort to pose one quick question, posed in the midst of the following essay.
In the mid 1990's, I worked for AT&T, hired just before the Lucent spinoff.
I worked in Reading, PA, in their Micro-Electronic [ME] manufacturing sector. The department I worked in dealt directly with Substrates.
IT IS THE VERY FIRST STEP IN THE PRODUCTION OF AN INTEGRATED CIRCUIT.
We used sophisticated grinders to ensure they were all of a uniform thickness.
The substrates were then polished to the typical mirror dark gray mirror finish that comes to mind.
The substrates were shipped to other departments at the facility. I worked mostly with 3.5" and 5 inch wafers. Weeks later, the wafers would emerge from their reactors, having the different conductive, insulative and semi-conductive layers placed upon 'my' substrate. In my mind, 4 to 6 weeks seems to resonate.
(in the late 90's I changed course into the fiber optic manufacturing industry and have since lost touch with any manufacturing process enhancements for ICs)
We would then take delivery of the wafers and grind the remaining substrate off of the 'bottom' of the wafer down to scant microns above the etchings from weeks prior.
Let's say that the time to production has been halved in the past 20 years.
It doesn't change the fact that HF is only in early chip production.
So, I gotta know, is this an effect of poor planning, poor engineering or some karmic manifestation for the douchebaggery that was undertaken in the name of Half-Fast?
The first of the KnC miners were a few days late.
I would think that Hal-Fast will be much slower than KnC.
That is not correct or at least not using current terminology. Substrate is not part of the wafer fabrication process it comes afterwards. The raw silicon in this case is produced by TSMC. The will deliver complete wafers. That process according to Hashfast began roughly weeks ago and is (per their words) on track for end of month. A packaging house will take the complete wafers, cut them into individual dies. The dies as bonded to a the substrate and optionally a metal lid (heat spreader is attached).  From HashFast mockup image. The shiny metal part in the middle is the die made from silicon it is the part which comes from the fab (TSMC). The green part is the substate it has nothing to do with the raw silicon production and is the last step in microprocessor construcion. The part with the logo is the lid. It protects the die from physical damage and helps to distribute the heat.  Older style microprocessor. Die (metal part) & substrate (green part). There is no package lid so the die is vulnerable to physical damage. More than one Athlon XP owner crushed the die attaching the heatsink.  Newer style microprocessor. The chip on the left is how it is delivered. The package lid hides most of the die and subtrate but if you remove it, it looks very similar underneath. HF will be late. They already stated being two weeks late. However silicon fabrication is 4-6 weeks so stating they are at the beginning of production is not correct. Hashfast wouldn't be responsible for supplying raw wafers. They supplied specs to the foundry. The foundry (TSMC) will deliver final fabricated wafers. They are responsible for everything between tapeout and final wafers. No customer (including HashFast) has any control or input on how the foundry does their work. |
|
|
|
|
Ninjarobot your image is broken.
|
|
|
|
With knc shipping in volumes, i'm not sure that this is correct. KNC is just about done shipping Sept/Oct batch. About a week ago they indicated they were 500 of 1500 units complete. Figure 100 units per day they are probably >1000 units shipped and 400 to go. They could be done by the end of the week. Regardless of if it takes them a day or week to wrap up at most KNC is "only" adding another 150 to 200 TH/s from batch 1 (probably less). KNC has stated Nov batch will not ship before 15 Nov and they also promised to the Batch 1 customers that Batch 2 would be 30 days after last batch 1. I doubt it will be the full 30 days but I can't see them spitting on their batch 1 customers by shipping batch 2 (which could be up to 2 PH/s) earlier than 15 Nov. |
|
|
|
Yep. Though to be fair, thats exactly how it went with KnC, and I assume they did plenty of testing with an FPGA. Of course its entirely possible (likely even) they will run in to more hickups, like the firmware issues KnC had. If they dont get packaged chips in the next 10 days, I seriously doubt they will ship anything before the tail end of November. Why 20+ days beyond chip arrival for shipping? Since you reference KNC they went from chip delivery to shipping in 3? 4? days. Granted that was kinda impressive but even a week isn't 3 weeks. |
|
|
|
|
Show Posts
