Topic: [LABCOIN] IPO [BTCT.CO] - Details/FAQ and Discussion (ASIC dev/sales/mining) - page 648. (Read 1079974 times)

Haha look at their share price doubling, even though nothing actually happened

And now it drags on over the whole september, then the whole october, and then most of november...

Interesting, eASIC actually just confirmed the ActM deal: http://www.easic.com/vmc-uses-easic-to-achieve-24-756-ths-bitcoin-miner/

I think we might see some pressure on Labcoin shares as people sell to buy ActM.  I think ActM's share price should continue go up in very short term, and hopefully I'll be able to pick up some more cheap labcoin shares.

I have a question for those who are selling, how many do you sell ?
I would like to know what the minimum price to buy to give.
I collect bitcoiny to build a house. 

Remember, this was the sample chip run. Perhaps they're working with the "worst case" scenario to test the limits of the chip without proper cooling? (I dunno, devil's advocate, etc.)

But this whole discussion is why I keep coming back here, despite all the other noise.  Thanks, guys.

Yeah, I was just pointing out the size of the packages is not the major determining factor.  A "chip scale" flip chip BGA package would probably be tiny but still be able to remove a lot of heat (from what I understand)


It is strange. It would be helpful to have more information from them on why they decided to forgo the heat pad. Maybe they think the plastic will be able to dissipate the heat.  

The thermal conductivity of plastic is about 0.2-0.5 W/mK, compared to 200 for aluminum.

But, remember the 'm' stands for the thickness of the material, not the surface area.  If the package material is very thin, it can still conduct a decent amount of heat.

If you have a surface area of 1cm (0.0001m²), thermal conductivity of 0.5 W/mK, 50C temperature difference, and 0.25mm thick layer of plastic, you should be able to move

-(0.5W/mK) * 0.0001m² * (50K / 0.0025m)
=-0.5W/K*0.0001m*(20000K/m)
=-0.5W/K*0.0001*20000K
=-0.5W*0.0001*20000
=-1W.

So, just one watt of cooling using those parameters. Maybe 2 watts if you count both sides.

It's also possible that they could be using a type of plastic with a higher thermal conductivity as well.

And, it's also possible my math could be completely off.  The units did cancel properly, but I could be doing that totally wrong

EDIT: A 12-inch wafer is 0.775mm thick, and a lqfp44 package is 1.2mm thick (not counting the pins).  So the thickness of the packaging should be 0.21mm on either side. But there could be other materials in there as well. If so, that should increase the amount of heat able to be emitted package a great deal.

Anyway, we'll find out in a few days what's actually possible.

Well, I laughed.

Normally the  heat pad is also for ground connection. I would wonder if Avalon did not use it this way too since good power and ground connections are quite important as half of  the transistors is toggeling at high speed.

...and that ^^^ makes me nervous.

While the die size determines the heat transfer at the silicon junction, the size and more importantly type of package is also important as the thermal energy must pass through the package as well and the lower the thermal conductivity the higher the core die temp is going to reach.  The bizarre thing is that they reported using "QFP packaging, 44 pin, no exposed heat pad" for the size of the die that makes absolutely no sense @ >12W and is still hard to believe even at 5W.  They do make packages with exposed metal heat pad to improve heat transfer.  For example this is the Avalon chip.



Inside the package the die is pressed against the large square metal pad in the center of the package.  The center pad is only used for heat transfer, the other pins are used for electrical power, ground, and signal.  The pad will be surface mounted to a non-electrical pad (copper plate) on the PCB to conduct heat away from this chip.  Using a multi-layer board the connection can extend from the top layer through the PCB to the bottom layer which is used as a heat dump.  As a side note this is why the heatsink is on the "back" of an Avalon board.  The heat is conducted through the heatpad, through the PCB, to the heatsink on the other side.  With a QFP and no heat pad essentially the entire chip is encapsulated in insulating plastic and that greatly limits the amount of power that can be dispersed.  

Thanks!

I agree. And even if they only achive half of what BitFury did, it's still fine with an estimated cost of $ 9-10 chip.



I'm aware of that. Labcoin 1st gen has a die size of 6,5 mm x 6,5 mm and Avalon 7 mm x 7 mm. Also: "... underperform ... in comparison to the original announcement ... would approximately match a deployment of 3 TH/s+ [as lately announced]". Please don't get me wrong, my only intention was to provide a reasonable context, nothing more.


The following might have been a hint from TheSwede75:

Thanks for your post.

Not exactly.  The W/mm² issue has to do with the size of the physical chip, not the size of the packaging.  There's only so much heat that can be removed from silicon for a given temperature gradient. For the packaging, you can use materials like copper or aluminum that have higher thermal conductivity then Silicon.


Yes, I think this is correct. They expect 2,000 chips and 3-4H/s.  That comes out to just 1.5-2gh/s/chip, not 4.8.  So they are already expecting about half the theoretical max output of their chips. And, at 2.7W/gh/s, that comes out to just  4-5.4W, not 12.8.

4.8Gh/s and 12.8W may be the design capacity of their chips.

Their recent announcement of 2,000 chips, 3- 4TH/s, and 2.7W/Gh/s comes out to just 1.5 to 2Gh/s and 4.05-5.4W per chip

It's likely that cooling system you'd need to use to get 4.8Gh/s is just more expensive then simply using two chips at half the power.

dexX7, I really appreciate your posts, contrary to most here, they're always really informative

I too expect the chips to underperform. Actually, everything above 70% would really surprise me.
But I don't think it matters much. The share price is still way too low if they have working chips, even if they just perform at 40%.
In every scenario where they actually start hashing, we're good.

The heat issue though, that's where I'm worried.

If they get hashing, they'll have to prove that they are hashing, not just pay dividends.

Hey all,

I looked up the specifications of other ASIC hardware to get a feeling about power consumption, heat, etc. to check what might be possible for the upcoming sample batch.

---

What Labcoin said:

Latest chip specifications:



Amount of chips:



The post from Sam is a bit older, though.

Expected hashrate:

---

The following illustrates the amount of chips needed and hashrate / chip to fullfil the goal of 3-6 TH/s with 1000-2500 chips.


For example Labcoin would need 2000 chips with a hashrate of at least 1,7 GH/s per chip to deploy 3,4 TH/s. This would translate to 4,59 W power consumption per chip or 9,180 kW altogether.

Looks fine till now, even if they underperform hard, they would accomplish their goal, but ...

---

Some said 12.8 W/chip will melt the chip etc., so I looked up already available chips. Note: this is not about efficiency, but only about the limits of power consumption.

• BitFury chips are hashing at 2,7 GH/s with 0,8 W/GH/s which results in 2,16 W/chip (reference), 55 nm, QFN 48 packaging, 7,0 mm x 7,00 mm.
   
• Avalon chips are packed in a QFN 48 package with a chip size of 7,0 mm x 7,00 mm and a transistor size of 110 nm. burnin's overclocked Avalon chips are running stable till somewhere near 430 MH/s with 1.3 V and a power consumption between 3,85-4,35 W/chip (reference #1, reference #2) with a air cooled block cooler or water cooler.
   
• The crasiest thing I've seen done till now is an overclocked Block Eruptor USB to 672 MH/s by mjgraham (reference), but it's not really a BE anymore, but a giant cooling block with additional hardware attached. Anyway, this thing would run at 10,49 W/chip and AM chips have a transistor size of 130 nm.

So green in this picture means the chip hashes stable and red equals unstable or meltdown.


I think it's not wise to say "because BitFury can't do it, Labcoin can do it neither", but it's intended to get a broader picture.

---

Pictures of the custom cooling solutions to achive this performance (BitFury, Avalon, AsicMiner #1, , AsicMiner #2):

---

Again the relevant quotes:



QFP 44 is a bit smaller than the packaging used by competitors (= more W/mm²) and no exposed heat pad basically means the heat is kinda trapped in the packaging.

TL;TR: I think the sample chips will underperform greatly in comparison to the original announcement and might only hash at 1,5-1,7 GH/s/chip. This relates to a power consumption of 4-4,5 W which would approximately match a deployment of 3 TH/s+ with 2000 chips.

Disclaimer: I'm not really familiar with this topic at all, but I tried to connect the given dots.

Work screwed me out of about 4-5 BTC again. Let me just say that trading is more profitable than work. Missed the "news" and the opportunity to sell at .0038+ and buy back in to increase my shares by about 20%.

This idea that Labcoin could be manipulating stock price in order to generate dividend payments is pretty interesting - anything is possible, I guess.

100THs @ Picostocks

I couldn't understand before why ASICMINER share prices was dropping, I had no idea they were being destroyed like this. I thought up until a few minutes ago that they still led the way.

Bitfury is a monster, this is amazing. Do they have public shares or anything like that?

Yes

Yes, that's correct. They've been building hashpower at a very rapid rate.