Re: [LABCOIN] IPO [BTCT.CO] - Details/FAQ and Discussion (ASIC dev/sales/mining)
Shareholders are loosing trust here. Please respond to the two points below ASAP!
So, you guys really have no idea how this will turn out? Incredible gamble. There are so many wholes I dunno where to start.
You haven't yet demystified that you took the simulation of one core and just copy/pasted it 15 extra times to make the chip and didn't bother simulating the end result. Where are the power specs being properly simulated? This is not as simple as 0.8W@200MHz equals to 0.8W*16@300MHz! That is really naive! A chip working at 1.5V uses 56% more power than one at 1.2V!
First, it's a "sea-of-hashers", then it's a "sea-of-gates" with rolled cores. Rolled cores means at best the chip produces a bitcoin hash at at half the frequency rate, since each core will need to do SHA256 twice for one hash (2 clock cycles). This means that a 300MHz chip with 16 cores will hash at 2400MH/s. Or is the chip gonna work at 200MHz like you also said? That's 1600MH/s.
A BitFury chip has 756 rolled cores but needs 65 cycles/hash, an equivalent of "normal" ~11.63 unrolled cores computing one hash per clock cycle. They are barely able to get 2GH/s from this design choice from the incredible signal interference inside the chip from skipping the necessary analog simulation but look how they estimated it to have between 2.8 and 10.4GH/s.
In case ppl don't know, a "sea-of-gates" chip is also a type of structured ASIC where only the metal layer needs to be customized.
Concerning the shares, it's not difficult:
I just need a list of accounts to lock. (feel free to PM or email me.)
Cheers.
Quote
[11:43] Alright,the 130nm IC is designed to work at about 200 Mhz for core, with a total consumption of 0.8W per core, hashing performances of about 200 mhash per core, total speed will depend on the overall output grade
Quote
[11:53] <+labcoin_dev> Chip frequency : 300Mhz, Process : 130nm CMOS, Die size: 4160,0000um2, VDD : 1.2-1.5 V adjustable
Quote
[11:57] So you've arranged them to fit more on the same size?
[11:58] <+labcoin_dev> we adopt a "sea-of-hashers" approach
[11:58] from vbs: "For that to be possible, not only each Labcoin core would have to be ~42% smaller [65/130*(6.5^2)/(7.1^2)] than each BFL core but also the Labcoin chip would magically operate at a higher frequency (300MHz vs 250MHz) while keeping the same power draw..."
[11:58] <+labcoin_dev> i remind you guys that the bitfury chip has up to 750 cores
[11:59] <+labcoin_dev> they also use the same approach, we chose this design because one of our guys was close to bitfury developments 2 months ago
[11:58] <+labcoin_dev> we adopt a "sea-of-hashers" approach
[11:58] from vbs: "For that to be possible, not only each Labcoin core would have to be ~42% smaller [65/130*(6.5^2)/(7.1^2)] than each BFL core but also the Labcoin chip would magically operate at a higher frequency (300MHz vs 250MHz) while keeping the same power draw..."
[11:58] <+labcoin_dev> i remind you guys that the bitfury chip has up to 750 cores
[11:59] <+labcoin_dev> they also use the same approach, we chose this design because one of our guys was close to bitfury developments 2 months ago
Quote
[12:13] 6. How is it possible to create chips with a such much better performance than the competitors?
[12:14] <+labcoin_dev> ThickAsThieves, rolled cores instead of unrolled, sea-of-hashers approach
[12:15] <+labcoin_dev> as in sea-of-gates
[12:15] <+labcoin_dev> you can google this term and find out more about what it means technically
[12:14] <+labcoin_dev> ThickAsThieves, rolled cores instead of unrolled, sea-of-hashers approach
[12:15] <+labcoin_dev> as in sea-of-gates
[12:15] <+labcoin_dev> you can google this term and find out more about what it means technically
So, you guys really have no idea how this will turn out? Incredible gamble. There are so many wholes I dunno where to start.
You haven't yet demystified that you took the simulation of one core and just copy/pasted it 15 extra times to make the chip and didn't bother simulating the end result. Where are the power specs being properly simulated? This is not as simple as 0.8W@200MHz equals to 0.8W*16@300MHz! That is really naive! A chip working at 1.5V uses 56% more power than one at 1.2V!
First, it's a "sea-of-hashers", then it's a "sea-of-gates" with rolled cores. Rolled cores means at best the chip produces a bitcoin hash at at half the frequency rate, since each core will need to do SHA256 twice for one hash (2 clock cycles). This means that a 300MHz chip with 16 cores will hash at 2400MH/s. Or is the chip gonna work at 200MHz like you also said? That's 1600MH/s.
A BitFury chip has 756 rolled cores but needs 65 cycles/hash, an equivalent of "normal" ~11.63 unrolled cores computing one hash per clock cycle. They are barely able to get 2GH/s from this design choice from the incredible signal interference inside the chip from skipping the necessary analog simulation but look how they estimated it to have between 2.8 and 10.4GH/s.
Technical Details (Translated from various Bitfury Posts)
Each chip is capable of 2.8-10.4Gh/s using a 756 core design
Estimated Chip Power Consumption: 1.96-7.26W (0.7W per 1Gh/s)
Estimated Power Consumption at Wall: 1.4W per 1Gh/s (<200W per 120Gh/s Device)
- The design is built on the 65nm UMC Process (http://www.umc.com/English/process/a.asp)
- Bitcoin Engines: 756 Rolled cores (65 Cycles per Hash)
- Expected operational frequency: 250Mhz-900Mhz
- Packaging: QFN48
- Conservative design, could be 40% smaller
- Risk interconnect and transistor variations +/- 20%
- Core implemented using full custom design process (some global place & route)
- Number of transistors per "core": 55,000
- Power estimate obtained from hspice simulation
- Design optimised for low power and minimum size rather than high clock rate
Each chip is capable of 2.8-10.4Gh/s using a 756 core design
Estimated Chip Power Consumption: 1.96-7.26W (0.7W per 1Gh/s)
Estimated Power Consumption at Wall: 1.4W per 1Gh/s (<200W per 120Gh/s Device)
In case ppl don't know, a "sea-of-gates" chip is also a type of structured ASIC where only the metal layer needs to be customized.
Until I see where the 3 million shares are, I consider this a scam. Why aren't you showing us your portfolios?
Concerning the shares, it's not difficult:
I just need a list of accounts to lock. (feel free to PM or email me.)
Cheers.