You should be looking at W/$. Capital investment in bitcoin mining hardware will likely be similar in the ASIC era as in the GPU/FPGA era, adjusting by block reward and exchange rate of course. So forget the hashrate. $1070 of GPUs use about 800W of power. $1070 of BFL FPGAs use about 150W of power. If a $1070 bASIC uses 405W, we're taking a step backward to almost GPU levels. $1070 of BFL ASICs should use only 50W of power, a significant step forward. I believe Inaba was thinking along this line when he said that.
|
|
|
They've collected millions of dollars in pre-orders. They can afford to throw around $12k on something that will encourage more sales.
|
|
|
90nm... my wild guess is ~2 W/GH.
|
|
|
Here are the changes I see for the $1070 unit:
Was: Multiple SHA256 hashing cores for a total Bitcoin mining speed of 54Gh/s Now: Two Bitcoin SHA-256 Hashing Clusters with Twelve 90nm ASICs (6 ASICs per cluster)
Was: Thermal sensors and on chip heat sinks Now: Dual High Performance heat sinks and fans.
Was: DC barrel connector for power Now: Optional 4 Pin molex or DC barrel connector for power
Was: Compact board design allows easy stacking and cooling Now: Board dimensions 4" x 9"
Was: bASIC Mining boards are projected to ship in November/December of 2012 Now: bASIC Mining boards are projected to ship in December of 2012
New: 54Gh/s is the gauranteed minimal hashing speed - we were very conservative here and it's likely that our software development team will be able to increase the mining speed. New: Energy efficiency information is coming soon.
|
|
|
Specs should be released in just 2 hours though. Three hours. EST is an hour behind EDT. Heh I bet he meant EDT.
|
|
|
Right now it looks like we are just over 800 units ordered (including both the 27gh/s and 54gh/s models) Just to be safe I am going to stop taking pre-orders at midnight tonight EST.
If people are seriously concerned about this - I will allow another 24 hours to get in your pre-orders
If people feel like there should be some time to digest the additional specs before I stop taking pre-orders than that is fine. I will stop taking pre-orders at midnight EST 10/22/12 - or when we hit about 900 units ordered, whatever comes first.
Given the context, I think he meant pre-orders stop at the end of Oct 22, not beginning, so 26 hours from now. I do wish people would avoid vague phrases like "midnight 10/22". Specs should be released in just 2 hours though.
|
|
|
It all depends on relationships. They may not have the face time in China to work out all the kinks. They may have to work through an intermediary who drives up costs. They may be unhappy with various stages, or all stages, of the work that is done overseas.
I have no idea why they've decided to take on specific parts of their assembly themselves, but we can all guess. Maybe Chinese business ethics prevents them from sanding off all identifying marks on chips stolen from BTCFPGA's fab.
I've been fortunate to work in tech-heavy locations, but I've still had to travel to China to kick off relationships, projects, etc., to make sure things work out OK. When we haven't had someone on the ground, we've gone back and forth with revised tech packs, killing time to market and margins.
ROFL! No offence meant toward ngzhang and crew. It's just that I've done business with other Chinese who act like it's totally normal to recycle old computer parts and sell them as "new" as long as they're cleaned up enough to look new, and sometimes even if obviously used. I have a feeling that's quite common over there.
|
|
|
I saw mining with GPUs as a safer bet than buying and holding bitcoins, even if it often didn't make much more profit than holding bitcoins. Because even during the crash from $32 to $2, difficulty dropped so I was able to profitably mine most of the time, and for a short time when I couldn't, I simply shut off the GPUs and didn't lose. And if the price had kept plummeting, I'd have sold the GPUs to gamers without losing nearly as much as an investment in the currency would have. In other words, while the price was going up, mining was profitable, and while the price was going down, mining was either profitable or neutral.
Of course ASICs will change that, and miners will be more exposed to the risk of currency failure.
|
|
|
Currently, it seems network hashrate stabilizes around 2 TH per $1 market price, for example currently 22 TH and $12/BTC. When the market was around $5 for a while, hashrate stabilized around 12 TH. After the new hardware is out in mass and the block reward is halved, I expect the new normal to be 40 TH per $1 market price. At $12, that's ~500 TH. If the market goes up to $20, expect hashing to increase to ~800 TH within a couple months. Hashing increases do seem to lag price increases though, so while the market is going up, mining is more profitable (if you don't compare to the profit of simply buying and holding the currency).
|
|
|
I think the X6500 is the most power efficient miner on the market today, at 16.4W for 400 MH/s. But when the block reward is 25 BTC and the network hashrate is 500 TH, which may happen by the end of this year, you'll need to be paying less than $0.09/kwh to break even. I'm assuming $12/BTC. And even if your power is free, you'll be making a piddly $1/month. Is it worth the time to keep it running? And what's the value of a mining device that makes you $1/month? $10? Maybe $20? The device may be worth more for scrap or recycling at that point. I believe most rational GPU and FPGA owners will quit by then.
|
|
|
Why are you guys still wildly guessing? Josh already said indirectly that they'd sold ~200 TH as of a few days ago. And if anything, I'd expect him to understate that figure, as a higher number discourages additional sales. I believe that by the end of the year, the network will be composed 90% of ASICs.
|
|
|
Keep in mind that even if a 54 GH bASIC uses 150W (2.5x) and a 60 GH BFL Single uses 60W, they're competitive for the first year of operation (assuming $0.12/kwh), regardless of the difficulty factor: bASIC: $1099.99 + $157.68, $23.29/GH BFL: $1333.00 + $63.07, $23.27/GH And surely both will be obsolete within a year, so don't tell me about how many pennies you'll save after the first year. And if it turns out the bASIC will use 250W, CablePair can simply lower the price by 10% or overclock to 60 GH to regain competitiveness. The important factors to be considered when investing in ASICs is price per GH and delivery speed. The big question one should ask right now, is will a new BFL order be delivered before or after a new bASIC order, as that will affect the bottom line far far more than power cost. We have some estimates when each will ship the earliest orders, but in BFL's case especially, the delivery date of new orders is a wild guess. The sooner BFL figures out and publishes estimates of when they'll catch up with all pre-orders, the more sales they'll take from bASIC. This is a good point, but I think your missing the point of the ASICs. If you invest in a high power consumption ASIC, it ceases to become profitable much faster than a lower power one. You say that they will both be obsolete in a year, but that just doesn't make sense. They become obsolete when they no longer have a positive ROI - so naturally, the one with the lower power consumption becomes obsolete later. In the case of a 2x bump in power consumption that means your device becomes completely useless much sooner than a comparable device. I dunno about you, but I'd rather mine on a device for 3 years than 1 year and have to fork out more money to stave off obsolescence. Not everyone can keep investing money year after year. That's why power is the most important aspect of an ASIC device and it literally defines if the device is viable or stillborn. Tom is basically the only hold out on the ASIC front as far as power figures go and it makes me question why that would be. Either he or his engineers have a rough estimate (or at least I would hope so!), and putting that out there would be the transparent thing to do, even if it's with the caveat that "it's only an estimate and may change by 20%" or something similar. That's a good point also. But if the next generation of hardware (after the upcoming ASICs) is 10x better in terms of $/GH (i.e. 500 GH for $1000), even the first-gen BFL SC line will soon become unprofitable at average power cost. BFL can probably turn a profit selling a 500 GH unit for $1000 with even the first-gen BFL SC chips, now that NRE is covered. But obviously it can't obsolete itself, so a new chip will need to be made on a smaller process node with much less power draw, for example moving from 90nm to 45nm when 45nm is affordable.
|
|
|
Keep in mind that even if a 54 GH bASIC uses 150W (2.5x) and a 60 GH BFL Single uses 60W, they're competitive for the first year of operation (assuming $0.12/kwh), regardless of the difficulty factor: bASIC: $1099.99 + $157.68, $23.29/GH BFL: $1333.00 + $63.07, $23.27/GH And surely both will be obsolete within a year, so don't tell me about how many pennies you'll save after the first year. And if it turns out the bASIC will use 250W, CablePair can simply lower the price by 10% or overclock to 60 GH to regain competitiveness. The important factors to be considered when investing in ASICs is price per GH and delivery speed. The big question one should ask right now, is will a new BFL order be delivered before or after a new bASIC order, as that will affect the bottom line far far more than power cost. We have some estimates when each will ship the earliest orders, but in BFL's case especially, the delivery date of new orders is a wild guess. The sooner BFL figures out and publishes estimates of when they'll catch up with all pre-orders, the more sales they'll take from bASIC.
|
|
|
Only problem is it's just an uncut wafer....
Did it look anything like this? Yeah, but uncut.
|
|
|
(Is this the right forum section?)
With PayPal, one can login and easily create any arbitrary invoice and have it emailed to a customer. One doesn't need to make a webpage to do so. Does BitPay offer the same convenience? I understand it may be a bit tricky because the 15 min countdown needs to be initiated only when the customer clicks a link in the email, not when the merchant creates the invoice.
I ask about this because it seems BFL is not going to accept bitcoins for upgrading orders from Jalapenos to Little Singles; they seem to only want to send PayPal invoices. I realize they might not want to bother adding a webpage to handle the various combinations (1 Jally to 1 LS, 3 Jally to 1 LS, 5 Jally to 2 LS, etc). So I'm hoping BitPay does, or can, offer the same convenience PayPal offers to merchants in this situation.
|
|
|
Don't forget BFL will probably ship months before Avalon does. That makes a huge difference. As for power consumption, if that is important to you, consider that even a single BFL unit can be run by a computer that uses less than 50W, so 110W total, still far better than the (conservative) estimate of the Avalon.
That said, I still think buying an Avalon is a good hedge against failure of BFL and BTCFPGA to deliver quickly and as advertised. If somehow Avalon manages to ship first, it won't matter how much power it uses. I have an Avalon on order for just that reason.
|
|
|
Yifu Guo has been very helpful with my orders, and responds to emails quickly! Just want to put in a good word. I'm looking forward to receiving some of the first shipment of Avalons in a few months.
|
|
|
Flip a coin 288 times. Will you get exactly 144 heads? Between 138 and 150 heads?
Difficulty adjusts every 2016 blocks. An increase in hashing can result in more blocks per day for a short period.
|
|
|
(600x54) + (50x27) = 32400GH/s
Add that to the hashes BFL's ASIC units will be adding to the network when they ship and you gotz a whole buncha crunchin'.
R.I.P. GPU mining.
The BFL Pre-order thread has confirmed ~48TH/s. This is another 32Th/s, and 300 Avalons is another 18TH/s. That's a minimum of 100TH/s by the time they all ship. Prolly closer to 150TH/s. That's still not even a 10x difficulty increase. I suspect it will be quite a long time before we get to 45million difficulty numbers. Part of this is recorded numbers, part of this is speculation, so should be taken with a grain of salt. Inaba posted confirming ~85th in all pre-orders so far (about a week iirc). So we're look at... at least 5x just on BFL hardware alone. if we see another 100th/s from other manufacturers... we're 10x. According to my maths, fpga will still make profit between 10x and 15x. but gpu's are stone cold dead at about 10x. Josh recently said, indirectly, that they have about 200 TH of orders so far.
|
|
|
Actually, is the size of the first batch that important? It's not like you can change whether you're in it or not, at this point. And if you didn't order in the first month, I can almost guarantee you won't get a first batch shipment.
|
|
|
|