QuintLeo
For Ethereum, the best price-performance was on the RX 470/480/570/580 cards 'till those cards got into GTX 1070 territory, at which point the GTX 1070's more-or-less equal performance to a high-end 480/580 card WHEN THOSE CARDS HAVE MODDED BIOS made the 1070 a comparable choice. Hash/watt on both sides of that is quite close, enough so that individual differences between cards could make as much of a difference as differences between the brands.
 
 The GTX 1060 can't match the hashrate of most of those cards, but it's fairly close to the bottom-end of the RX 470/570 range which started driving sales of the GTX 1060 when 470/570 prices got high enough - again, efficiency is competative.
 
 The GTX 1080 is NOT competative, the latency on GDDR 5X hurts it's ETH hashrate quite a bit, to the point it can't quite match a STOCK RX 480/580 and struggles to beat a bios-modded 470/570 - and it's not a lot higher than the 1060. At current pricing though it might be getting SOME interest just because it's AVAILABLE.
 The GTX 1080ti is higher hashrate, but not by a TON over a modded RX 480/580 - and so far the price has made it prohibitive for ETH mining.
 
 The kicker in the pack though is that ALL of those NVidia cards are very competative on a hash/$ basis EVEN AT MSRP vs AMD on mining ZCash and many other coins that have smaller market caps then Ethereum, so there is some demand from miners working those coins.
 ZCash in particular is likely to have a majority of NVidia cards working on it, as most recent AMD cards are more profitably when mining Ethereum (Tahiti based cards like the R9 280/280x and HD 7970/7950 are notable exceptions there).
 
 BALLPARK figures (these can vary quite a bit with specific GPU type, especially on how high you can clock the MEMORY - and I might be off a bit as these figures are from memory in many cases and a lot of them are based on "reported" figures):
 First figure on ETH is for a stock card, second for a BIOS-modded card. NVidia cards to date can't be BIOS modded.
 AMD cards with 8GB memory will usually be about 2 Mh/s faster then 4GB due to the 8GB cards using 8000 Mhz memory vs 6600 or 7000 on the 4GB cards.
 
 Card model           ETH hashrate    ZCash hashrate
 GTX 1060 6GB      22 Mh/s           300 sol/s
 RX 470/570 4GB   22-26 Mh/s      280 sol/s
 RX 480/580 4GB   24-28 MH/s      320 sol/s
 GTX 1070             30 Mh/s           420 sol/s
 GTX 1080             25 Mh/s           500 sol/s
 GTX 1080ti           35 MH/s           720 sol/s
 
 I'm not sure what the hashrates of the RX 460 are, but I'm pretty sure it's noticeably less than the GTX 1060 6GB.
 The GTX 1060 3GB does a good bit less on hashrates than the 6GB, but I don't remember any actual figures offhand.
 
 Note that current estimate for Ethereum is that there are over 2 MILLION GPUs mining it, the majority of those sold in the last 2-3 months and probably 60% more or less being AMD cards in that timeframe.
 Given that Ethereum network hashrate more than doubled in the last 3 months, that's over 1 MILLION cards sold to miners in less than 3 months for just Ethereum mining.
 I think I remember seeing an estimate somewhere that AMD's TOTAL production capacity is ballpark 4 million GPUs a quarter.
 
 Current estimate for ZCash is that there might be as many as 1 Million GPUs mining it but likely a good bit less due to the NVidia dominance there and the high sol/s on high-end NVidia cards.
 My best estimate for ALL OTHER GPU-mineable cryptocoins combined is less than 1 Million GPUs in use and more likely 500k ballpark.
 
 For perspective, if all of Team Curecoin on [link=mailto:Folding@Home]Folding@Home[/link] were strictly using GTX 1070 cards, we would have slightly less then 1000 cards working on Folding - and 5000 GTX 1070 would cover ALL of the PPD generated by [link=mailto:Folding@home]Folding@home[/link] participants in the last month.
 
 
 

QuintLeo
Like any cryptocoin, breakeven depends on a lot of factors.
 
 Right now, the only rig I have mining Ethereum runs 3 x R9 290 - which are high-performance cards on ETH, 28-29 Mhash/s under current conditions, but they eat a little over 300 watts each to achieve that - high overclock mitigated somewhat by useage of a modded BIOS from TheStilt that undervolts them a little and disables some of the video-specific parts of the card/GPU.
 
 At my current power cost (which is one of the 3 lowest "small business" all-up rates in the entire US), the entire rig eats a little less than $1.25 a day, but generates right at $10 a day - but that $10 depends on difficulty (which has been rising fairly rapidly the last 3 months), and the price (which jumped a TON 'till about a week back, then settled back down by a quarter - peak was near $400 but it's running right about $300 per coin at the moment).
 
 A GTX 1070 will manage the same to a little more hashrate, while using rough ballpark half the power, as will a BIOS-modded RX 480 8GB or RX 580 8GB (the 4GB versions of those cards use slower memory that can't clock as high, which drops their hashrate some).
 
 If current conditions continued, one those 3 modern cards would generate a gross intake of almost $3.50 a day and net a bit over $3.25 at my power cost - someone paying 10 cents/KWH would net more like $3/day while someone paying 25 cents/kwh (I think some parts of California are at or above that all-up cost) would net around $2.50/day
 
 However, folks keep adding more and more rigs, the total add seems to be limited by available GPU supply of the "good mining cards" after factoring in the "competition to get the cards" vs. everyone else trying to buy the same models for gaming and other usages. Ethereum difficulty has more than doubled in the last month due mostly to this factor, which means at the same price your gross income is half for the same hashrate.
 
Ethereum has some factors specific to it as well that are reducing profitability - the developers implimented something called "Ice Age" that is increasing the time between blocks that can be mined and therefore drops the number of coins AVAILABLE to be mined in a day. They also have announced plans to move Ethereum to a Proof of Stake model in multiple stages on a "phased-in" plan, with the first stage targeted for the end of this year - but they've been talking a move to PoS for quite a while now and their previous plans have been postponed a few times, no clue if that will happen again or not and nothing announced on how long or how many "stages" they plan to go through before they end up at 100% PoS. The first stage was announced to be 1% PoS 99% Proof of Work (PoW is the current method) which will have almost no effect on miners, but later stages will drop profitability noticeably, and the final move to 100% PoS will kill ALL mining for Ethereum.