Guesstimate thread for total ASIC pre-order hashing power.

[DeathAndTaxes]

Interesting chart.  IMHO (which doesn't really mean anything don't take it the wrong way) I think it is too aggressive.  I would consider it a worst case scenario.  I find it unlikely the growth would be more aggressive than that.  I added it to the OP.  If you can provide numbers I would like to add that as an alternate viewpoint for the OP.

Some comments in no particular order

KNC: Shows shipping 2.2PH/s in roughly 30 days.  Based on breakdown on known orders, the average unit is ~0.25 TH/s.  That would mean shipping ~8,880 units in 30 days.  While not impossible it seems improbable if for nothing other than a manpower perspective.  To go from zero to ~300 units per day would be impressive.  Also I am interested to know where the 2.2 PH/s.  Based on their probable die size that would be a massive number of wafers and represents about $38 million in sales.  For the record I think other estimates of very small die are inaccurate.  Given the package size and power consumption it would seem to be unlikely.

Avalon:
~1.2 TH/s by end of year is about 3x all known rigs and chips.  Curious why you would think Avalon will sell that many chips?  Especially now given their massive delays, lack of price drop, and poor "rep" in the community.  Since Yifu update, among the major bulk orders a grand total of 3 orders totaling about 24K chips have been shipped.  That works out to about 3.3 TH/week.  To ship another 1000 TH by the end of the year (~18 weeks) would require shipping 20x as fast.  At the current rate it will take two years just to ship the known 275 TH/s of chip orders.  I am not sure what the problem is with Avalon but I don't think they will sell that much capacity nor will they ship it that fast.

ASICMiner:
1TH/s by end of year seems ambitious given they have had difficulty keeping the 50 TH/s existing farm operating anywhere near peak capacity (falling from 47TH/s to 30 TH/s over the course of two weeks and despite recovering in the last week not exceeding 40 TH/s).  BTW I am an ASICMiner shareholder and I believe in Friedcat but I would think 1 TH/s by end of year would be a best case scenario for shareholders (worst case scenario for everyone else) and I wonder how likely that is.  Granted ASICMiner additional capacity is likely a 2nd gen chip and thus should have on the order of 4x (55nm) to 8x (28nm) power efficiency and density (hashpower per square foot of warehouse space).  Still 1TH/s even w/ 2nd gen chips would represent growing the farm's footprint (space & power) 500% to 1000%.  Given the difficulty in maintaining 50 TH/s today I am not confident that will be easy.  A lot could change in the next couple weeks though.

Bitfury & Hashfast:
The growth seems probable but the max capacity seems light (kinda hard to tell with the scale of graph).  For HF they will have to have another 880 TH/s to cover excess chips for MPP.  Either way those chips will be used in January.  Either they 1st batch will make ROI and HF will sell the excess chips or they won't and the excess chips will be given out as part of the MPP.

Cointerra:
Could be the graph but it looks like more than 2 PH/s.  I also think showing delivery starting 1 DEC is at this point impossible; "in Dec" is ASIC code for 31 DEC (and likely 1-2 weeks late but that will be close enough that we don't look that bad).  Similar to KNC I don't think 2 PH/s per month in delivery is likely.  Although based on KNC "rollout" we will have a better idea for Cointerra. 

Still I think it is great, it got me thinking about other options.  It also points a pretty gloomy picture as a worst case scenario.  If it unfolds like that I don't see anyone except the chip producers having a % ROI.  Any chance you could share the raw numbers behind the graph and your rationale?  I am interested more in the thought behind the numbers than the actual numbers.

[gkm22d]

It was my first draft, just wanted to get some numbers on there just to help people visualize.

I will edit some this afternoon.

I have KNC down for a total of 2 ph, it probably looks like more because when there is an addition underneath it throws everything else above it off visually. I picked 2ph instead of 500 th just to go for worst case scenario.

Hashfast and icedrill, used their numbers, will add more for January hashfast. Bitfury, pretty much just used this thread's numbers.

Why do I think Avalon will have so many chips? Their next gen coming out in October, no preorders, lots of nre money from previous success and mining with people's equipment or not shipping to keep difficulty down. I will adjust downward.

Asicminer, I am fairly confident in that number. Friedcat knows what he is doing and has more money than anyone. Comitted to 20% of the network. Shrewd business man. 28nm chip 8x more power efficient, so they can be up to 400th right there. Double that with their leasing plan, and throw in 200th for sales.

As for shipping dates, purely just guessing. It would take a lot of research to nail down better estimates, which most people would disagree with anyway.

I will pretty it up though and post update.

[ImI]

Bitmine?

http://www.coindesk.com/bitmine-to-drop-4phs-of-asic-power-onto-bitcoin-network-before-april/

Maybe late, but plenty.

[bcp19]

I love the chart, it is very well made.  I do have to admit that you predicted the 'explosion' to happen about a month after I did, but we have more information at this time than I did when I made mine.  Very well done.

[gkm22d]

http://s21.postimg.org/mqfwmlz93/Photo_2013_08_27_4_50_54_PM.png

I edited a little bit, as per discussion. Don't know if it is more proper to edit post or make new post, also,  might have undersized.

Feel free to suggest changes.

[DeathAndTaxes]

Bitmine?

http://www.coindesk.com/bitmine-to-drop-4phs-of-asic-power-onto-bitcoin-network-before-april/

Maybe late, but plenty.

Thanks, that is a new one.  I think I will change the OP to break it into deployed in 2013 and deployed in 2014.  The reason is that due to falling revenue per GH/s each month it is more important to get the short term accurate then the long term.  4 PH/s in March really doesn't affect someone who has a current pre-order because honestly if you haven't hit 100% ROI by March you probably aren't going to and if you are profitable you likely will make most (75%, 80%, 90%) of your lifetime revenue.

[DeathAndTaxes]

Anyone have a good idea on how to figure out ASICMiner sales other than try to look historically for every group buy and direct sale in the past couple months?

If not posting links and totals form various group buys would be a start.  If any reseller wants to help me out that would be awesome.

Something like

Date: [#####]
Reseller: [name]
Hashrate: [# GH/s]
USB Eruptors: [# units] 
Eruptor Blades: [# units]
[Link to presale thread]

[gkm22d]

Finally found a source for 1000th for asicminer, last paragraph
http://thegenesisblock.com/cointerra-expects-to-deliver-2-phs-of-asics-in-december/

[AussieHash]

http://www.coindesk.com/bitmine-to-drop-4phs-of-asic-power-onto-bitcoin-network-before-april/

The third generation ASIC, labeled CoinCraft, will be a 28nm in-house effort. It has been co-designed by a number of players, including Synapse, Innosilicon, and Verisilicon.
....

This chip will cost $6.8/GH in fully assembled boxes, which like the Avalon and Bitfury boxes will be available directly to end-users. Bitmine will take delivery of 100,000 of these chips from Global Foundry in November, giving it 4 Petahashes of power (not including the BitFury chips).

It expects to ship 1-1.5 Petahashes of power to customers who want to run the boxes themselves. However, this is really a chip for the cloud, as it enables the firm to dynamically adjust power consumption in a large-scale computing operation, and that’s where the other 2.5-3 Petahashes of power is going.

That’s where the Asian datacenter and a $12m, four Megawatt supercomputer comes in.
....
Massive Luck Investments and Bitmine are jointly building a data center in China that will host mining power using the third-generation chips. That facility, which will cost just under $1 million, will be finished around February, although it will start hosting CoinCraft processors in small quantities in December.
....
However, the other supercomputer is scheduled for installation in March at the Bitmine/Massive Luck data center, and it’s where the firm’s 2PH of remaining ASIC hash power will go.

[geofflosophy]

I edited a little bit, as per discussion. Don't know if it is more proper to edit post or make new post, also,  might have undersized.

Feel free to suggest changes.

Is that y-axis difficulty or hashrate? I think that 10 Petahash/s is a solid estimate for the beginning of January, if it's difficulty you might be overshooting some.

Edit: looking at the value for 8/24 makes it clear to me that you're measuring hashrate.

[DeathAndTaxes]

Finally found a source for 1000th for asicminer, last paragraph
http://thegenesisblock.com/cointerra-expects-to-deliver-2-phs-of-asics-in-december/

Updated thanks.

[DeathAndTaxes]

http://www.coindesk.com/bitmine-to-drop-4phs-of-asic-power-onto-bitcoin-network-before-april/

Added thanks.

[geofflosophy]

http://www.coindesk.com/bitmine-to-drop-4phs-of-asic-power-onto-bitcoin-network-before-april/

Added thanks.

Bitmine can be ignored for these calculations through January because they are already factored into Avalon and Bitfury.

[sickpig]

just want to say thanks for the effort that everybody involved in the thread have made to gather this precious information. kudos.

[erk]

ASICminer are going to have to get up to 1PH/s by Xmas to remain in the game.

Assuming they can get a 16x power reduction from 110nm to 28nm die reduction, they are going to need a 20x hash rate increase to stay where they are now. That would mean replacing their entire farm with 28nm tech and expanding it by Xmas.

[willphase]

BFL is likely to be more than 2PH.

The basis for this is taking the amount of hashrate that's been delivered to the network since around June time (when BFL started shipping a lot) and comparing that to the public order list - this gives a 'hidden' order list of around 6x - meaning BFL would be shipping between 3 and 4 Ph.

This estimate doesn't include any 28nm tech

EDIT: FWIW my mining investment decisions are based on a guess that there will be around 16Ph on the network by May 2014.

Will

[DeathAndTaxes]

BFL is likely to be more than 2PH.

The basis for this is taking the amount of hashrate that's been delivered to the network since around June time (when BFL started shipping a lot) and comparing that to the public order list - this gives a 'hidden' order list of around 6x - meaning BFL would be shipping between 3 and 4 Ph.

This estimate doesn't include any 28nm tech

EDIT: FWIW my mining investment decisions are based on a guess that there will be around 16Ph on the network by May 2014.

Will

If you can show some links or references I will raise the estimate.  BFL isn't the only one who has been delivering hashpower.  I hated in school when teacher said "show your work" (because I could do it faster in my head), but "show your work".

[DeathAndTaxes]

ASICminer are going to have to get up to 1PH/s by Xmas to remain in the game.

Assuming they can get a 16x power reduction from 110nm to 28nm die reduction, they are going to need a 20x hash rate increase to stay where they are now. That would mean replacing their entire farm with 28nm tech and expanding it by Xmas.

That is their stated goal and as a shareholder I agree it is necessary, but I am also concerned about the ambitiousness of it.  The recent rise in hashing power above 50 TH/s is a good sign.  If they can get the network stable above 50 TH/s for a couple weeks that would be a good sign they have the power and space to handle the upgrade.  Also I think efficiency gain will be lower than 16x.  If it is say 12x then it becomes even more ambitious.    Still the purpose of the thread is to document planned hashpower increases so I have AsicMiner down for 1,000 TH/s in 2013.

[willphase]

If you can show some links or references I will raise the estimate.  BFL isn't the only one who has been delivering hashpower.  I hated in school when teacher said "show your work" (because I could do it faster in my head), but "show your work".

it's basically a lot of assumptions, which is why it's more of a guesstimate than anything accurate.  Not sure showing my working will help other than to stress how much we don't know about BFL order queue.

I'm assuming that BFL have been 75% of the hashrate delivered in a three month period between June and now.  This is around 100 - 550 which is 450Th and 75% of this is around 330Th.

I'm then looking at the public order list for these months and adding up the Th based on MR orders until the current date for MR (28 July)- 22 orders and only 2/3 minirigs delivered (from Jody's blog they are delivering 500Gh at a time) so 15 * 1.5 = 22.5Th then minirigs form average of 277/(277+36+150) = just over half (60%) the BFL hashrate is minirigs so that's around 40Th (including others) delivered from the public list.

so 40/330 means that the factor of 'hidden' orders is around 6-8 times (depending on the estimate of 75% above).

so now take the total number of orders which is around 500Th and multiply by 6-8 and you get the 3-4Ph that BFL will deliver.

I know it's a TON of guesswork here - which is why I give an error margin of 3-4Ph from BFL.

Will

[Syke]

so now take the total number of orders which is around 500Th and multiply by 6-8 and you get the 3-4Ph that BFL will deliver.

I know it's a TON of guesswork here - which is why I give an error margin of 3-4Ph from BFL.

A while ago they had 75 wafers on order, with 1000 chips per wafer that's around 300 TH/s. Unfortunately they haven't announced ordering any more wafers. They're going to have to order a lot more wafers to reach 4 PH/s.