Won't be the first time that GPU mining "died" for the most part - consider 2014.

It varies quite a bit depending on the type of item and often on the specific item, but most tech-related stuff tends to have a markup between 60% and 100% at every distribution level.

I didn't say "data hall" or "hosting" - I was talking actual electric all-cost-included rates in Chelan and Douglas counties of Washington State for residential and small business - though Chelan in particular seems to be getting unhappy about cryptocoin miners flocking there for the cheap power and it's effect on their infrastructure and BOTH have a "moratorium" in place on new larger operations (Gigawatt seems to have slipped in just before that moratorium hit with their Douglas county/East Wenatchee facility).

Grant county next door is at about 4.7 after the last rate increase, and is a lot less restrictive "so far", but appears to be starting to make some effort to force cryptocoin operations to pay a higher rate "sometime soon" for similar reasons.


For folks paying AVERAGE US rates or worse, I'd figure a lot have already shut down at least some mining gear as it has gone underwater already.
I suspect this is a lot of the reason that DASH hashrate in specific (among others) has seen a noticeable drop over the last few months since the prices collapsed.

Whether you like ASICs or not is kind of a null point considering they are being produced regardless.  I understand the argument that people will just buy more and the same amount of electricity will be used regardless.  This argument ignores all of the drawbacks to asics though and only considers the benefits of high hashrate with lower power.  Keep in mind:

1.  ASICs cost significantly more than a GPU
2.  They can only mine one algorithm

[/quote]

1. Depends on the ASIC and the GPU - there are at least 2 current Bitmain ASIC that cost less than a GTX 1070 much less the 1070ti 1080 1080ti or Vega.
  Even the current S9 models are in the SAME price range as a 1080 ti or higher-end Vega.

2. Not always the case - the Bailkal units in particular can ALL mine multiple algorithms, and the Gridseed / SFards chips were dual-algorithm.

In most cases though, an ASIC will completely blow away any GPU on what it CAN work on - it is more limited, but it is more capable by far WITHIN those limits (the Bitmain Ethereum ASIC is the ONLY exception, but even that thing is quite a bit cheaper for the same hashrate and very similar efficiency vs a 6-card GPU rig that achieves about the same hashrate) - which is why folks will continue to invest in ASIC miners despite their limitations.

And even more so in the 2 counties of the US where you can get to 3 cent power.

Nvidia didn't build those - 3'd party manufacturers built those.
Still would have an indirect effect.

The point is for folks that are NOT used to LINUX and comfortable with it to be able to use it.

It's gear as well - Innosilicon A2/KNC Titan/Alcheminer generation gear MIGHT be break-even at 3 cent power, they're losing money everywhere else.

We're pretty much back to where things were at 2 summers ago, except on the current gear generation instead of what was current then.

https://www.semiwiki.com/forum/content/6895-standard-node-trend.html

Explains a lot about the current and near-future "announced" nodes from the main names.

They're somewhere in the same county I am at - same rate I pay, and there is a TON of mining activity here and in the next 2 counties over where the rate is close to (or UNDER) 3 cents/kwh due to the very cheap electric cost.

Our rate here DID just go up a bit though - it's closer to 4.7 now than 4.6 with ALL costs included.
Local PUD (and the 2 in those "next counties overs) EACH owns one or more major Columbia River power dams - and subsidizes local rates somewhat through the sale of excess electric production to other areas that will pay quite a bit more and STILL saves money over their own local generation.
I call Central Washington State "the land of VERY VERY low cost electric" for a reason.

Yes, the US average is closer to 12, but you can find places to host gear in the area I'm in that will charge you 5-7 range including hosting fees and internet access etc.

It depends on your electric cost - for ME, an L3+ would still be profitable, but for most folks you're correct that it is not profitable any longer.

"12nm" is not a new process, it's a process improvement of "16nm" that is run on the same fab lines.
The rename is more about marketing than any significant change to the feature size.
It's also NOT going to free up any fab space or capacity on the current "16nm" lines.

Going to get even worse in the next generation - what Intel and Samsung (and IBM) are calling "10nm" and shipping TODAY is almost identical to the announced specs for what TSMC and GF are calling "7nm".

Then the generation after that - IF THERE IS ONE - may not be a silicon generation, given Intel comments about "10nm is the end of the road for pure silicon" and IBM using a mixed silicon/germanium wafer in their work on 7nm.

1 KW from a 4-battery series connect of car-type or standard lead-acid deep cycle batteries is a fairly low load - car batteries in particular are rated for "reserve capacity" in minutes at a 25 amp draw.
It would also be not bad for a 2-battery series setup at 24 amps, well within the rated capacity of the batteries though you might see a small reduction in the "rated" amp-hour capacity.

Not even close to the typical 500 VA UPS that runs from a single 12V 7AH gel cell (and DOES show a massive reduction in battery capacity due to the very heavy draw).

Mining Equihash with a R9 390 is a mistake in the first place - they are a LOT more profitable on Ethash and I don't think they've EVER matched their Ethash profitability with equihash mining.

Scrypt had the equivalent of many thousands of L3 units in total network hashrate before the L3 was released, which let it soak a ton of hashrate while staying profitable for a long time even before the Big Price Jump in Spring 2017.

The original L3 also was close to the same hashrate as the Alcheminer and only a bit over double the biggest A2 units, it mostly won based on much better efficiency than even the Titan.


ZCash isn't quite as well off, but it'll still probably take months for the Z9 and competition to drop profitability by half even though the efficiency of the new ASICs is a ton better than existing GPUs.

I'm all GPU at this point, but switching some of them to different algorithms/coins anticipatory to ZCash profitability collapsing once the Z9 starts shipping in quantity.
I do have some bitcoin on "hold" at this point for a probable future purchase of some ASIC gear - but the gear I'm anticipating hasn't shown up yet.

Almost pulled the trigger on some of those "cheap" A3 or D3 units, as the probably ROI isn't all that bad at their current cost - but the power usage is huge for what they earn.

I'm sure a lot of folks have been turning off their older generation miners over the last couple months - even at 3 cent electric the A2/Titan/Alcheminer generation is no longer profitable.

L3/A4/L21 generation is still profitable, if you don't have "high for mining" electric rates, but are starting to get a bit marginal.

Bitmain HAS in fact sold used miners in the past - usually in "close-out" shortly before they introduce a new model, and in the case of the S5 they were sold specifically AS "used" (though the ones I bought were in "like new" condition when I got them).
I think they did the same thing with the S7 about a month before the S9 was announced.

The miners used on a coin do NOT dictate the price of a coin.

5% diff increase average seems a bit high - but even a 2% average increase will see the L3 become unprofitable before it pays for itself unless you have extremely cheap or free electric, and THAT is less than the average so far this year.
The huge increase in early-to-mid December warps long-term averages, and seems to have been a one-shot multiple companies getting batches online at the same time event - plus sales have been slowing since coin prices dropped a lot after the big peak very early this year.