They’ve started assembling Baby Jets (overnight tonight). They aim by night’s end to have the first round of assembly done for 400 Baby Jets.
They took the Baby Jets’ chassis’ out of their boxes and opened them up. Removed their drive bays and other unneeded components. Unboxed the Seasonic power supplies, attached and labeled their cables, and installed the supplies in the Baby Jets. Inserted the cooling unit and radiator. Added an additional chassis fan on the back of the box. Then screwed the chassis back together, put it back in its styrofoam packing material, and stacked each on pallets. So, the bulk of this update is something that they could have done back in October. Exactly what I was thinking. I suspect the "eagerness" of Caria is a euphemism for the fact Hashfast booked production lines at Ciara. And since HF arent ready yet, they just had Ciara do some silly things like unboxing and reboxing cases without PCBs. Not exactly the most efficient way to go about it.
|
|
|
How do colored coins in any way solve the problem (if you consider it one) of the deflationary nature of bitcoin? How is it different to lend me a colored coin vs a non colored coin?
|
|
|
Its bad enough as it is, that you dont have to make false claims. Its entirely possible no monarchs will ship before "march or april" but the supplied quote gives no evidence for that. On the contrary, it states first wafers should arrive on January 9th. For a competent company, that should mean shipping units in late January, even if its only a tiny batch.
Dont get me wrong, I fully expect more delays to be announced, I fully expect bulk shipment to be closer to summer than winter, and I dont consider BFL competent, so in the end March or later for a first batch and may or later for bulk is definitely plausible, but if you are going make a post like that, please dont give the false impression that is what BFL said. They havent said it.
Yet.
|
|
|
Considering China, USA, Japan, India, and Australia all use a period '.' for decimal places, it's clear who the winner is.
1,000,000 = 1 million 1.000.000 = The number 1. No not many people use a double decimal place nor is it widely understood.
"winner"? Lol. By that logic, forget English as language, its clear that Mandarin is the winner, followed by Spanish. http://en.wikipedia.org/wiki/List_of_languages_by_number_of_native_speakersNow pls rewrite your post in Mandarin. You may also want to start saying the year is 4000 something instead of 2013.
|
|
|
edit: brainfart. Those chips are 28nm chips, just like coincraft A1. Nothing old about them, they arent even on the market yet. As for power consumption a little under 1W/GH at the wall is in line with all the other 28nm chips, and that probably includes coincraft A1 which is specced at 0.6W/GH at the chip level. Its not because AMT seems to think it will somehow be lower at the wall than what the chip is specced for, that you should bet a lot of money on that.
|
|
|
These photos look intimidating. I hope the manual is really good Its laughably detailed, fortunately and for the most part, not hard at all. Still, assembly proved to be not without some challenges. I had trouble with the thermistor (temperature probe). The supplied heat resistant sleeves proved inadequate and I ended up shorting two thermistors (velleman kindly provided new ones free of charge). Fixed it by using some kaplon tape. I also didnt tighten the extruder well enough, and as a result it leaks a little bit. If I leave it printing unattended, now and then a blob of molten plastic will ruin the model. For now I just swipe the hot end clean with some ear cleaners every 20 minutes or so. In the long run, I will need to disassemble it, clean it with a blowtorch, and try again. I cant tighten it now, because there is already plastic between the threads.
|
|
|
I was doing a test print of thin overhangs. This makes the cooling fan kick in and it results in rough, unsmooth edges and web like threads. This is fully expected, and normally you would print an object like I was printing in ABS and then you smooth it with acetone. However, what was not expected was the effect youtube image stabilization would have. Its frigging hilarious, so I thought Id let you have a laugh: http://www.youtube.com/watch?v=8PM_woiVquM#t=25
|
|
|
KnC's network protection program is quite inferior to BFLs. BFL have been protecting the network for like 18 months now
|
|
|
Sooner rather than later electricity costs are going to become very important. Absolutely. But when that happens, no one is going to want to pay $13K for a few TH. The big difference with GPU's, is that highend GPU's would always be relatively expensive compared to the electricity they use. High end GPU's are in demand for gaming etc, regardless of what bitcoin difficulty is. Bitcoin ASICs have no use besides mining, so once mining profitability drops so low that electricity cost becomes a major factor, prices per TH will have to plummet.
|
|
|
Worst case scenario? We will own the IP of the chip and it will be the first open source RTL
Im pretty sure the contract with TSMC will make that impossible.
|
|
|
Of course it could be 5.5TH and BTC could be at $2K by then, who knows, who knew that KnC's 350GH/s Jupiter would end up at 650GH/s and BTC up from $100 to $1000? BTC price doesnt enter the equation. Yes, I used dollar pricing, but only because electricity is still billed in dollars and to make it more clear how relatively insignificant 0.1W/GH better efficiency still is and will be for a while. ROI calculations should be done in BTC, and any fluctuation in the price would apply equally to all asic vendors. If anything if BTC price keeps going up, the energy efficiency would become even less important. As for Neptune being 5.5TH; thats possible, but for BA Im going by nominal specs, they also promise overclockability to 2.1 GHz which would put the X3 at 2.7TH or 5.4 TH for 2 units at the same price as Neptune. NOt too much difference there either, even if Neptune ends up at 5.5TH. But I've made back what I spent on my Oct batch When you calculate your ROI in BTC?
|
|
|
Good luck banking on BlackArrow. I feel pretty confident that they will fold before releasing anything. Too late to the party!
Why would they be too late, and Neptune would not? AFAIK, BA sold out. They should also be at least as close to tape out as Neptune, and without the risks involved in a brand new and quite radically different process node thats not even deployed. Not saying Im banking on BA, Im not "banking" on any asic vendor, but if I were to make a bet, at least BA makes more sense to me. YMMV.
|
|
|
Neptunes are down to 32/1200 left up for sell.
Seems they will have sold out 1200 x 12,995.00 = WOW
I dont quite understand that either. Even if people assume Neptune will do 4TH, BA prospero X3 would cost the same per TH and in all likelihood will be delivered several months earlier. Sure, Neptune might be more power efficient, but only if KnC delivers significantly better performance per watt than they claim and/or BA delivers worse performance per watt than they claim, because BA states 0.5 W per GH at the chip level (lets assume that works out to 0.7W at the wall) while KnC is saying 30% better than Jupiter. If Im not mistaken, Jupiter does around 1.1W/GH currently? A 30% improvement over that is ~0.8W. ON paper they are pretty much on par, with arguably BA having a slight advantage. Even if you assume KnC will be one or two tenths of a watt more efficient per GH, if that is more important than mining several months earlier, then you are in trouble. @4000GH and $0.15/KWH, a 0.1W/GH better efficiency works out to 0.4KW * 0.15 = $0.06 per hour, or $43 per month. One day that will matter, but if that is significant next spring, and more significant than mining a few months earlier, good luck earning back those $13000.
|
|
|
Wow, so at 15b, a Neptune will barley be making 3 BTC per/mo... Hopefully that's worst case... I would actually consider Neptune shipping in May a best case scenario. Given all the uncertainties regarding 20nm production, I wouldnt be surprised if it shipped closer to July.
|
|
|
But the only way to force my miners to get IP addresses in a row is by setting them all static. Unfortunately the router decides to change that later on for no reason If you use static addresses, make sure you select addresses that are outside of the DHCP range.
|
|
|
Thanks for the link. I guess that settles it. Although it doesnt quite explain what appears to be exaggerated energy efficiency claims compared to bitmine's specs. I also dont understand why AMT seems to hide this info. Anyway, AMT doesnt appear to be an outright scam, but I still cant say they give me a warm feeling.
|
|
|
I am not going clear back through the thread, but I was under the impression that their 28nm devices will be based on the Bitmine Coincraft chips. These guys appear to be assemblers, rather than a straight up engineering company.
If they are using Coincraft for their 28nm miners, they should just say so. It would eliminate much of the doubt (at least when bitmine confirms this). But let me point out that AMT specs dont seem compatible with bitmine's specs. AMT is listing "300-600W" for their 1.2 TH rig. Thats 0.25-0.5W/GH at the wall. Bitmine is only claiming an efficiency of 0.6W/GH at the chip level for nominal performance targets and 1W/GH in turbo mode. They do say it can probably go down to 0.35W/GH (IIRC) but at unknown speeds, and that is still at the chip level. For 1.2TH, 300W at the wall seems impossible with bitmine chips and less than ~800W at nominal speeds, very unlikely. And why would you not use the turbo mode initially?
|
|
|
Apparently 500GH is underclocked.
At this point, under- or overclocked doesnt really have any real meaning yet. When you under or overclock a chip, its changing the clockspeed in relation to the manufacturers default speeds. These chips are still being characterized.
|
|
|
|