21e6, LLC raises $5 million to manufacture Bitcoin ASIC | Hacker News Reader

21e6, LLC raises $5 million to manufacture Bitcoin ASIC

bitcointalk.org

31 pointsmrb16 comments

Now we have a Bitcoin ASIC bubble in addition to the regular Bitcoin bubble. I have a feeling some of these guys are not going to recoup their NRE (especially at 20 nm) when customers are only willing to pay $0.50/GH/s. Never mind, I'm sure suckers will buy them.

Is anyone actually shipping anything performing in the sub $6/GH/s range? Seems like everyone claiming that has failed to ship. Then again by the time these guys do ship something it may not matter much given the amount of hashing power coming online in the next 90 days (assuming knc, hashfast and others actually hit their targets).

The price scales inversely to difficulty which is roughly doubling every month. Anything shipping right now is probably worth $10-12/GH/s. December delivery pricing is around $6/GH/s, January is around $3/GH/s, etc.

Right, I agree about both this ventures potential for covering and at the same time there are tons of suckers. My comment was just more to the fact that companies currently claiming they can deliver at $3/GH/s haven't actually been able to. I'll pick on cointerra since they have a banner that calls out their price of $3/GH/s. They were supposed to have delivered in November or December and now it's late January. When they do actually deliver in February (if they deliver at all) the miners that come on line will likely be producing at a loss for their buyers. KNC is delivering at $6/GH/s in December and they are sold out.

It is estimated that most ASICs cost less than $1/GH/s to make and the current prices just reflect demand outstripping supply. Thus by January I expect all vendors (including e.g. KnC who has known working chips) to be selling for $3/GH/s.

AFAIK Cointerra's batch 1 delivery date has always been December.

The difficulty is climbing at an incredible rate because (a) lots of ASIC miners are coming on-line, and (b) a lot more people have started mining.

At some point very soon, mining Bitcoin will become unprofitable overall simply because an essentially fixed number of output Bitcoins are being divided across a larger and larger number of miners.

The only way to mine profitably will be figuring out how to mine both more efficiently and significantly more quickly than the general mining population.

Essentially if you are newly investing in Bitcoin now, you need to believe that (a) the mining solution that you're buying into will be at least an order of magnitude faster than the current generation products, or (b) Bitcoin will increase in value so much that even a small number of Bitcoins will be very valuable.

If this is just another Bitcoin ASIC, it's seems a very risky investment.

> Essentially if you are newly investing in Bitcoin now, you need to believe that (a) ... or (b) Bitcoin will increase in value so much that even a small number of Bitcoins will be very valuable.

If you can't mine profitably, yet believe that the value of a bitcoin will increase, then you're better off just paying money directly for bitcoins rather than inefficiently mining them.

Nice name! matches the number of total bitcoins mined + unmined.

I didn't even realize that. Nice catch.

I was thinking maybe Unicode code point, looked it up, and got ⇦--your explanation makes a lot more sense.

Last time I looked at ASIC design a few years ago, that might get you a couple of prototypes and not much else. The software, staff and fabs cost a crap load of cash.

"A few years ago" is a long time ago. Thanks to Moore's law, designing an ASIC, getting a full mask set made, and the first few wafers etched only costs $500k at 65nm:

http://blog.zorinaq.com/?e=73

Other Bitcoin companies (KnCMiner, etc) have apparently done 28nm with merely $1-3 million.

That matched my experience as well in trying to price out my frame buffer design.

Of course they aren't buying the 'fabs and staff' just a chip designer with a decent design suite that has the geometries and processes for their chosen fab partner pre-loaded (probably TSMC or Global Foundaries). Get the design done for 250-300K (what is the annual salary for a chip designer these days ? $150K? $200K?) and then its pretty much point and shoot at the fab. Hand over the files, run the layout, build an alpha mask set, do a couple of wafers with it, tweak to the pre-production masks, so an A0 run, fix what is left and build an A1 "production" part.

Given modern simulators you can simulate the chip pretty completely, synplicity can do the physics simulation of the process to verify geometry I believe.

I have been looking at what it would take to 'mass produce' a simple ARM frame buffer design. There is quite a bit of support out there for "free" (its really back end loaded into the process costs since once you qualify on TSMC it is harder to move to GF and vice versa)

Question, is it at all possible to mine bitcoins on a personal laptop?

Not with any kind of profit - even with a high end graphics card in the laptop, such as a HD 8970M, you would be looking at best a 0.0004 BTC daily return - this would definitely outweigh any profit after you take power consumption in to account. Any changes in difficulty (which is often), will lower your return.

Reply on news.ycombinator.com