Quote from: Fatman3001 on September 23, 2015, 08:58:19 PM

Quote from: AmityGames on September 23, 2015, 07:39:20 PM

Quote from: Fatman3001 on September 23, 2015, 07:16:47 PM

Quote from: AmityGames on September 23, 2015, 06:45:45 PM

Q: Now, what kind of company announces an amazing product but does not suggest a price for it?

A: A company who is uncertain of demand and price points

IMHO, Spondoolies has a small number of working prototypes of the SP50 and is trying to figure out how to gear up for production. They may have just barely enough demand from their hosting/merger partners to get production off the ground, and are naturally looking for ways to milk the wider market beyond at the highest price it can bear (with optimists leading the way of course).

This is marginally better than KnC, who's been stuffing every ASIC at hand in their Swedish facility guarded by buxom blondes with machine guns

Technically, I trust Spondoolies to deliver. I have looked at the guts and firmware of all their machines extensively. As an EECS myself, I think their implementation is indeed best in class (better than Bitmain and KNC). One suggestion - do not affix the heatsinks with an epoxy bead (a la SP20).

Economically, I would not pay much more than 1.0-1.2 BTC per TH for ANY new miner. At present BTC rates, this puts the economically feasible price for the SP50 around $25,000 per unit.

Pay more than that, and you are unlikely to see any ROI - which is kinda the point of the whole thing if you are a professional operator with access to low cost electricity.

Hi there,

I didn't know you cared for mining. Nice to see.

I think producers are forcing customers to realize that ROI in less than a year is not realistic any more. But with cheap electricity these new machines should be profitable for far longer than previous generations of asics. The nm race is grinding to a halt, and even if there is a breakthrough in that field it will take 2-4 years for it to make it to production.

Thanks.

Would you invest your money in mining if the baseline ROI was 1-2 years or more?

The VCs who are funding the large mining ops are definitely looking at much faster paybacks. The rational ones at least....

I'm worried about a new difficulty race. Spondoolies needs to sell only ~3000 SP50 to double the network difficulty and halve the yield (that's on top of the reward halving in June).

On the semi side, it's not that hard nowadays to step down to 20 or even 16 nm. The fab capacity and technical expertise is there, readily available.

If I had a good setup I wouldn't hesitate. That is, if I had a guaranteed access to a low cost facility and guaranteed cheap electricity. Now is the best time to do it, as the network will grow and the halving will keep overinvestment in the field at bay (I hope) you have a good chance to ROI (in terms of BTC) most of the costs (depending on the price of the blimmin miners, of course) before the halving. I do believe the BTC price has hit bottom. And I do believe it will pick up within the next 6 months. But if it stays sub $1000 for the next couple of years I would be much more comfortable with mining than speculating.

With regards to 20nm I don't know if it's relevant anymore. As I understood there were some minor efficiency advantages over 28nm, but its main advantage was the ability to run chips at higher frequencies with less leakage. I guess that's attractive if your chip designs focuses on a few massive chips, like with the KnC Neptune, rather than lots of smaller ones. But with the 14/16nm chips it will be interesting to see, if we ever get some confirmed power specs on Bitfurys, KnCs or LKETCs 14/16nm machines, how much there is to squeeze out of those processes.

I think Bitfurys claimed 0.06J/GH figure is not strictly down to fab process. They've been tweaking their chips for efficiency since day 1 and clearly have an advantage in this regard. But if it is mostly down to fab process it's worth to note that the difference in efficiency between 0.07J/GH and the specs for the SP50 is as large as between the antminer S5 and the first KnC machines, percentagewise.

In any event I would be very interested to hear your thoughts on the matter.

Thanks.

Where BTC price will go is anybody's guess. It's like trying to predict the stock market or turbulence. You can't.

What you can do however is to make evaluations of a given investment subject to certain level of risk. This is standard operating procedure in engineering or finance. You don't make your plans based on an optimistic outcome (no matter how much you believe that BTC will go up).

For my planning, I usually do a status quo scenario (BTC price and difficulty remain roughly the same, after all they've been like this for almost an year), a conservative change scenario (BTC price at $200 a nice round number, difficulty grows to reflect new capacity going online) and a bad scenario (BTC collapses sub-50, difficulty at 1 bil). The last one tells you how much you have at risk. The first two give you a reasonable estimate of your return.

This is how the VCs who fund the private ops of KnC, BitFury, SPO...etc look at the economics and pricing. If you look at their decisions, nobody seems to be anticipating a rise in BTC prices. This being said, the last 9 months have been sweet for mining, which is exactly what they've been doing with their own hardware. The situation may change though.... so beware jumping on the SP50 or the S7.

I think the 16 nm press releases from KnC and Bitfury with their reported 0.06J/GH are mostly BS, or lab results that will be difficult to scale fast. This kind of PR is intended to scare/delay capacity expansion, since these guys have the most to gain from the sweet mining status quo above.

With regard to tech, your efficiency grows as the inverse square of gate length (all other things being equal). You do have a hard limit on clock speed, but afaik none of the ASICS operates close to it - as far as I can see, all the designs are constrained today by thermal dissipation limits.

Which is why the only tangible advances on the open market so far are based on undercloking the true-and-tried 28 nm process. If your electrical design is already ironed out, it's not very expensive to order incremental runs from TSMC as long as the latter have the fab still available (and 28 nm tech is very mainstream now). So, you cram many ASICS, run them underclocked, and voila - you have the S7 or the SP50.

When the likes of TSMC step down their fabs to 20 nm or 16 nm (and this will be driven very much by the likes of Intel, not Spondoolies) you will get these improvements almost immediately in the bitcoin space. This is likely to happen sooner rather than later - look up Intel's or TSMC official plans and you'll see where the tech is heading. IMHO, this will happen around the time of reward halving...