Pages:
Author

Topic: [ActiveMining] Drawingthesun's Thread. Only Known/Useful Information - page 2. (Read 8342 times)

hero member
Activity: 532
Merit: 500
https://karatcoin.co
I'm certain that miners will be able to be built to spec. Worst case scenario is that we need more space to mine if no better cooling solution can be found found due to having so much kW in a confined area.
legendary
Activity: 1176
Merit: 1015
The following conversation makes some good points about the power density of our 24TH/s miners:
(Unless really good cooling is being used or something else (liquid submerged cooling?) we might not be able to build these 24TH miners to spec)
Please remember we are in a knowledge vacuum when it comes to our chip details, as we have not been updated on changing spec.

kleeck, did you calculations account for the expansion cases? They actually have 16 PCI-e slots (assuming a miner is only one slot), though they are 5U not 4U. But I think the requirements would still be lower. Smiley

No, I didn't take that into account. That would mean 34U could handle 24Th/s, excluding any heating concerns - so much less space would be needed.

A dose of reality is badly needed here.

At best easic might match KNC's 28nm W / Gh/s performance, which is on the order of 1.5 W/Gh/s.  It's more likely this chip will perform worse than that, but let's give Ken the benefit of the doubt.

So 24 Th/s would consume 36 KW of power.  High density racks only support 10-12 KW per unit, and that is with sophisticated water cooling radiators on the doors.  36 KW requires 165A of 220V.  If you have any familiarity with the type of industrial rental spaces that Ken is pictured in, you would know that they typically have 100A of 220 available.

I sincerely feel bad for the folks that didn't bail on this disaster when I was sounding the alarm in June and July.

First off, the 24Th/s Platinum is rated at ~20,800 Watts or 20.8 KW. That's before the overclocking potential that Ken has stated the eASIC design will safely allow. The space you see Ken pictured in is the manufacturing center, NOT the data/mining center. You're right though, the space needed will probably be greater than 6 racks for DTS' scenario, which is why I plugged in the "excluding heating concerns".

Thanks for the dose of reality.  Wink

Sure, go ahead and stick with imaginary numbers based on nothing rather than real world data.  An easic is not going to get close to the power consumption numbers you are dreaming about.  

The imaginary world must be a lot nicer.  AMC has 4-5 Ph/s running since September in that world right?  Wasn't that the projection?  Is that epic spreadsheet where Ken plans to mine more than all the available bitcoin in a year still floating around?

Yes. The imaginary numbers given by eASIC. No need to get all worked up. This is all theorycrafting, anyways.

Cheers.
hero member
Activity: 532
Merit: 500
https://karatcoin.co
kleeck, did you calculations account for the expansion cases? They actually have 16 PCI-e slots (assuming a miner is only one slot), though they are 5U not 4U. But I think the requirements would still be lower. Smiley

No, I didn't take that into account. That would mean 34U could handle 24Th/s, excluding any heating concerns - so much less space would be needed.
full member
Activity: 238
Merit: 100
Please write down what the time/date based on when you estimate.

Thanks.
hero member
Activity: 532
Merit: 500
https://karatcoin.co
So beginning of October, the network was hashing with 800TH/s, now with 5PH/s. End of January this will be 14PH/s or something. 1% would be 140TH/s

The Gen1-Chips will be hashing at 20GH/s max, so we'd need to have 7,000 chips hashing.

Anyone know how many chips per machine? If we say 16, then that's 430 machines. If Ken builds 5 machines per day, that's still 3 month and the hash rate wil be 50PH/s. The official thread claims 24TH/s per machine, but that would be 1200 chips/machine. Sounds a bit high.

So either Ken is hiring in January like crazy or this will be a dud.

Each card in a machine runs 256 GH/s before overclocking. Each single case can receive a total of 6 cards before needing an expansion case, which is ~1.536 Th/s per case.  So ~92 single cases would be needed @ 16 GH/s chips x 16 chips per board X 6 boards per case.

For space we can calculate off of each single case being 4U (as per VMC data sheet specs: http://goo.gl/5Tn0l0)

4U * 92 cases = 368U. The largest racks I've ever worked with are 42U each. There may be bigger, but lets work off of that for our purposes here. 368/42 = 8.761 or, effectively 9 42U racks to achieve 140Th/s.  

Here are some external dimensions of a 42U rack: (H x W x L): 80" x 24" X 42" - so In a building with 7' ceilings (shouldn't be a problem), and a ~20'x35' area one could easily house these devices.

Running them all at the same time is a different issue.  Cheesy
legendary
Activity: 1176
Merit: 1015
The 24TH/s miners are individual machines in the same sense as the Colossus is an individual computer.

The general sentiment here is correct, the 24TH/s "machine" is a rack of mining machines with the 256GH/s cards inside. (About 94 cards or 8 - 11 machines in a rack, similar to how servers are stacked up)

I doubt it will be as large as Colossus though. Smiley
full member
Activity: 210
Merit: 100
So beginning of October, the network was hashing with 800TH/s, now with 5PH/s. End of January this will be 14PH/s or something. 1% would be 140TH/s

The Gen1-Chips will be hashing at 20GH/s max, so we'd need to have 7,000 chips hashing.

Anyone know how many chips per machine? If we say 16, then that's 430 machines. If Ken builds 5 machines per day, that's still 3 month and the hash rate wil be 50PH/s. The official thread claims 24TH/s per machine, but that would be 1200 chips/machine. Sounds a bit high.

So either Ken is hiring in January like crazy or this will be a dud.

The 24TH/s miners are individual machines in the same sense as the Colossus is an individual computer.

legendary
Activity: 1176
Merit: 1015
The official thread claims 24TH/s per machine, but that would be 1200 chips/machine. Sounds a bit high.

That is the top end machine. If you look on the website it looks like the 24TH beast is more a rack of machines rather than a single desktop sized device.

So beginning of October, the network was hashing with 800TH/s, now with 5PH/s. End of January this will be 14PH/s or something. 1% would be 140TH/s

The Gen1-Chips will be hashing at 20GH/s max, so we'd need to have 7,000 chips hashing.

Oh yes, your estimates are pretty spot on. If we start hashing at the beginning of February we will need at minimum 200TH to claim 1% of the network.

Anyone know how many chips per machine? If we say 16, then that's 430 machines.

The chips are small and many can fit on a board, unfortunately we have no idea how Ken will have the boards designed.

A single machine will have multiple cards, and those 256GH cards will have about 12 - 15 chips inside them.

I am sure a machine can take many cards.

So either Ken is hiring in January like crazy or this will be a dud.

This really depends on how much work the engineering firm will do. They are building the boards, I'm not sure who puts the chips on the boards, probably Ken.

If all Ken has to do is take eASIC chips, put them on boards and place the components into a card case he might not need many hands. Again we have no idea how much work is being outsourced.
member
Activity: 112
Merit: 10
So beginning of October, the network was hashing with 800TH/s, now with 5PH/s. End of January this will be 14PH/s or something. 1% would be 140TH/s

The Gen1-Chips will be hashing at 20GH/s max, so we'd need to have 7,000 chips hashing.

Anyone know how many chips per machine? If we say 16, then that's 430 machines. If Ken builds 5 machines per day, that's still 3 month and the hash rate wil be 50PH/s. The official thread claims 24TH/s per machine, but that would be 1200 chips/machine. Sounds a bit high.

So either Ken is hiring in January like crazy or this will be a dud.
legendary
Activity: 1176
Merit: 1015
So, it's more likely he will get the chips in february then?
If he will need whole dec+jan?

I think Ken meant he would like the entirety of Dec + Jan before we all start raging on the forum again. I felt that he was implying their would be chips before February but he needs time to build the rigs and start shipping.
sr. member
Activity: 330
Merit: 250
Expected timeline for chips:
Based on Ken's responses in the thread, the current estimate is around January.
Quote from: kslaughter
I want the whole month of December and January.

So, it's more likely he will get the chips in february then?
If he will need whole dec+jan?
legendary
Activity: 1176
Merit: 1015
Feel free to post questions and your own estimates for our company below:

This thread is meant to be either facts, well researched and well formed opinions, no FUD and no screaming.

I will remove crap because this is not a general ActiveMining thread.
legendary
Activity: 1176
Merit: 1015
Information regarding our first generation chip and it's comparisons to the competition:


Everything about the chip and the cost is mostly unknown. We know it should be rated for 16GH/s and overclockable to 20GH/s.

Entropy makes a good point about the power requirements for the racks being perhaps requiring too much power;
At best easic might match KNC's 28nm W / Gh/s performance, which is on the order of 1.5 W/Gh/s.  It's more likely this chip will perform worse than that, but let's give Ken the benefit of the doubt.

So 24 Th/s would consume 36 KW of power.  High density racks only support 10-12 KW per unit, and that is with sophisticated water cooling radiators on the doors.  36 KW requires 165A of 220V.  If you have any familiarity with the type of industrial rental spaces that Ken is pictured in, you would know that they typically have 100A of 220 available.

A solution could be to invest in a liquid cooled system such as our competitor ASICMINER. Please PM me if you think there would be reason this is not possible (Or feel free to talk about it below)
legendary
Activity: 1176
Merit: 1015
Notes

I believe in Ken and believe that this operation can be profitable, I will post my math as evidence later on today or tomorrow.-> I am leaving to go out and am busy but I will post my workings out.
legendary
Activity: 1176
Merit: 1015

Please see https://bitcointalksearch.org/topic/m.7133625 for latest information.


ActiveMining is currently under investigation from Missouri Securities Division.

Until the investigation concludes; 99% of trading and all dividends are suspended.




Please do not be offended if I delete your post!!! Most of the deleted posts are simply me consolidating information to make the thread have a higher signal:noise ratio!

This thread is not a replacement for any other ActiveMining thread, but rather a gathering of information. Feel free to post information or even ask questions.

Questions that have no answer will be considered information because a negative can still tell a story.

Because I want the thread to be very high density in information, I may delete valid posts and consolidate the information on the front page. I will leave long discussions alone as long as they are making decent progress in developing an idea and it would be difficult for me to take that discussion and summarize the facts.

See my "Notes" post below to see opinions on the information.

Please note that some information is speculation and other information is confirmed by Ken. If it's confirmed by Ken I will quote the appropriate post.



What is the expected timeline for chips?
Based on Ken's responses in the thread, the current estimate is around January.
Just going to make a little prediction here - the products which shipped were probably Fast-Hash-80's which are the avalon clones.

Has Ken got a solution to the funds being held/lost by Ukyo?
We believe Ken will seize Ukyo's shares in ActiveMining as an exchange for the lost funds.
The fact that he can seize Ukyo's shares is actually good news.
IntelliHash

I can't see how intellihash could possibly work... the only thing it possibly could be is an over-clocking utility. It cannot change a chip (Ken said it would work on avalons too), it cannot fake proof-of-work (and if it can, bitcoin is broken), so the only option is an over-clocking utility. Which would be coherent with the 20% increase quoted, as that would be about the max you could reliably overclock.

TLDR; IntelliHash is, in my judgement, an overclocking utility.
Perhaps this is on the right track?


Timeline

I've brought this timeline over from the other thread as I think it may be useful to keep a copy here.

The dates in bold are confirmed by announcements, while the others are projected from estimated timelines in the prospectus.

  • 26th July [source] - Ken estimates two weeks for NRE to be paid (during visit to eASIC)
  • 3rd August [source] - Ken returns from his trip to eASIC and the Engineering firm
  • 4th August [source] - Ken posts to say he is busy working on eASIC deal
  • (sometime before August 28th) [used for guesstimates below] - NRE Funds paid
  • 28th August [source] - Ken confirms NRE funds were converted 'some time ago'
  • 4th September [source] - eASIC issue press release
  • 12th September [source] - Avalon refund confirmed as having been received
  • 1st November [original guesstimate] - Chip samples delivered in 9 weeks;
  • 30th October [source] - Ken announces gradual hashrate increase including 'other resources' [source]
  • 25th November [original guesstimate] - Low-volume chip production starting in 12 weeks, using an e-beam process;
  • 30th November [source] - Ken announces delays due to further R&D and upcoming Intellihash(tm) technology
  • 12th December [source] - Weekly announcement states "we are continuing to bring in all the parts for mass production of our miners"
  • 18th December [source] - Weekly announcement states "working with eASIC and our engineers to get our chip and boards in full production"
  • 25th December [source] - Weekly announcement states "working on assembling [for] when our boards and chips arrive"
  • 24th December - 9th January [original guesstimate] - Normal volume chip production starting in 16-18 weeks.
  • 1st January [source] - Weekly announcement states "The new year looks bright ... as we continue to assemble our miners for quick delivery"
  • 9th January [source] - Weekly announcement states "engineers are still designing our board ... hired ... competent RTL design team to make sure ... Low-Power and ... most speed from the chip"
  • 15th January [source] - Weekly announcement states "in San Jose working with our partners ... The RTL problem has been solved ..."
  • 27th February - 10th March - assuming an 8 week delay (from Jan 9) or 12 week delay (from Nov 30), chips may be forthcoming at this point in time

* Note: As of 30th November, probable delays have been introduced into the timeline.

(I will be trying to keep this updated as more facts are revealed. In an effort to keep this thread tidy, please PM me with any suggestions.)
Pages:
Jump to: