They are pretty much useless hassle and represent about 100 years step back in then banking industry.

I think this is a rather bold statement so let me elaborate. 

I'm Average Joe. I earn a decent salary in Piresian Pesos (1 USD =  20 PIP). I have a bank account, a bank card and PayPal. I'm not into terrorism, money laundering or general shady business. I want to order an expensive, 300 USD coat from the US because I dig it and I can afford it.
My options:

PayPal: I pay 6100 PIP. The bank converts it to 300 USD and takes 5 USD on the conversion. Then PayPal takes about 10 USD on th transaction and about 10 USD more if the shop owner wants it in paper form. I paid 305 USD and the shop owner received ~280 USD. 25 USD was taken by the bank and the Paypal to provide the place of the transaction in a secure manner. The transaction is instantaneous. PayPal insures the transaction as well. If the coat doesn't live up the the sales pitch, I can place a claim and get my money back.   

Bank Card: I pay 6260 PIP. The bank converts it to 300 USD and takes 13 USD on the conversion. The shop owner pays about 15 USD to the bank card scheme + yearly fees. The money is on the shop owner's bank account, so no extra fees on that. I paid 313 USD and the shop owner received ~285 USD. 28 USD was taken by the bank and the payment scheme provider to secure the transaction. The transaction is instantaneous. The schemes usually provide some fraud protection and get your money back if the coat failed to show up.

Bank wire: I pay 6600 PIP. The bank takes a conversion fee and a wire transfer fee or 600 PIP  or 30 USD. The shop owner gets the 300 USD straight, no deductions because I set the wire transfer to take all expenses on my end. 30 USD was taken by the banks. The transaction is greatly delayed, up to a week to clear. The bank does provide some protection if the destination bank account address was entered incorrectly but does not care about the outcome of the order.

Cash in an envelope: The old but gold method. Your preferred choice if you want no bank involvement. I put 300 USD into an envelope and send it via a reputable courier. I endure a 12 USD exchange loss and pay 25 for the envelope, a total of 37 USD. The owner receives 300 USD. The transaction is not insured or secured in any way. If the envelope gets lost or the coat never shows up then it is gone.


Then I realize that the shop owner accepts ByteToken, the most popular crypto at the time with all the usual promises and capabilities. So
I want instantaneous transaction so I go to CoinBase and buy ByteToken using my bank card. Currently one ByteToken (BT) costs 1000 USD so I will need 0.3 BTs. Unfortunately, CoinBase likes to sell BTs with a 3% exchange markup. I have to pay 309 USD to get 0.3 BTs but then I realize that there is a transfer fee as well, which is 2% or 6 USD. That means I need to spend 315 USD to be able to get 0.3 BTs to the shop. But oh no, coinbase also pushes the 5% bank card fees to the buyer which is 15 USD. After all, I pay a total if 330 USD to get 300 USD through. So far so good, except my bank also does a PIP to USD conversion for 330 USD at the cost of 8 USD. I pay a total of 6760 PIP for that 0.3 BT to go to the shop owner. That is 38 USD in fees and conversions. Unfortunately, the fees do not stop there. If the shop owner would like to get that 0.3 BT into paper USD, he would need to pay 15 USD for the BT to USD conversion and the withdrawal. In total, of a 300 USD transaction, both paries endured a total of 53 USD fees. And for that, you get the same safety as sending cash in an envelope. If the coat does not show up, the bank will do nothing because their insurace stops at CoinBase and CoinBase gives no reversal. Also, if you somehow manage to mess up the address and the money gets lost, then it is gone.

The bottom line is: Using crypto managed to make the most expensive method, the Cash in an envelope, 40% more expensive with no added customer protection whatsoever. Practically, you pay the markup for an instant cash in an envelope transaction.

Now, why would I still want to go back hundreds of years in time and do this cash sending method?
- Because banks are evil and I give them no money! Fair point. Too bad, you need a bank account to send money to CoinBase and the bank will take its sweet conversion and transaction profit anyway.
- I don't want the Piresian Tax Authority to see my expenses! They will see that you spent 340 USDs. They will not know about your coat, but does that really matter? Coats are not outlawed or anything.
- It is more simple! Well you still enter the bank card details but you also need to register to an intermediary website and go some lengths to identify yourself. It takes a whole lot more time in reality.
- It is safe and secure! This is just non sequiter.
- It creates a unified currency system! No it does not. On the contrary, it adds an extra conversion. With normal means, it is PIP -> USD. With Crypto, it is PIP -> BT -> USD. It is about as useful as the Esperanto language. It was supposed to simplify conversations between foreigners. All it did was to add an extra language noone speaks.


And it gets worse. If you are the shop owner, you want your books nice and simple. You got the coat for 150 USD and want to hold on to 100 USD after the sale. The rest goes to tax and stuff. If the BT moves wildly around, +5% today, -15% tomorrow, +2% after etc then you must re-price the coat in every 2 hours to follow the currency movements. A currency must be steady and more or less plannable on the short and the long run. If you sold the coat at 0.3 BT today but then the price went down 5% by the time you cound convert to USD, you just lost 15% of your revenue. Just because you didn't cash in on time. So in order to avoid such losses, you add a 20% markup to buffer this out plus the conversion fees. You will sell at 335 USD instead of 300 and the buyer will eventually pay 375 USD for the 300 USD coat.   

This leaves cryptos useless for the usual daily activities of Average Joe. A bank card is a nice and simple thing Joe can use quickly, easily and still have the protection of the bank. Meanwhile with a crypto, it is fees upn fees upon conversions upon insecurity and risks on the buyers side.

But hey, it is still good to save money into. Hm why is it better than any fiat money?
- It can be stolen easier and cleaner than normal cash.
- It has no backing, just like normal cash. The "consesus" is worthless. If we all agree that ByteToken is worthless, that is 100% consensus, but still makes BT worthless. The "hashing power" is worthless. Today there is 999999 TH in the system. Tomorrow, everyone turns it off and then there is 0 "hasing power". A BT worths as much as a +10 gilded broad sword in WOW. For a WOW player, that may worth 1000 USD. For someone who never played a sec, it has 0 value.
- It doesn't hold its value any better than normal currencies. Normal currencies rarely do the roller coaster ride cryptoes do.
- It is "limited". Kinda no. ByteToken declared that it will issue 100.000.000 tokens ever. Sounds fair until, 10 years later, Consensus comes up with a fork that doubles this number because it is nice. Or just forks a newer, better currency that phases the old ByteToken out completely. Also, "limited" doesn't really add value to something. There was a limited series BMW that was nice and all until the engine blew up at 100k. Since it was limited, the engine was limited as well. If the engines gives, there is no replacement part and your car is as good as dead.     

As it stands now, the whole crypto scene is more like Wild West scenario with no law and order. This is the perfect place to make or burn money quickly with speculation, bullying, scamming etc. However, that is not development or advancement, that is 150 years regression into the lawless past. 

Now please discuss. Prove me wrong.

A 1060 6gb with Samsung RAM does around 0.125 Eth/month. That is about 32 USD/month after deducting electricity. Non Samsung does around 0.1 Eth/month and probably bring in about 26 USD.
I calculate something like ~7 months for 3Gb ~9 months for 6Gb for Samsung, 9-11 months for non Samsung.

Yes and no.
The 1050 TI's mhs/W ratio is slightly worse than the 1060 so you will have higher electricity bill for the same performance.
The hash/USD ratio of the 1060 3GB and the 1050 TI 4GB is about the same. However, you need more rigs with more fixed hardware costs (MoBos, CPU, RAM module, PSU, HDD/SSD) to get the same performance as a single 1060 rig. Those rigs also consume more power.

~$1000 = 049 mhs 2x1060 rig     0.049 mhs/$
~$1200 = 059 mhs 4x1050TI rig. 0.049 mhs/$
~$1600 = 088 mhs 6x1050TI rig. 0.055 mhs/$
~$1650 = 098 mhs 4x1060 rig.    0.059 mhs/$
~$2250 = 147 mhs 6x1060 rig.    0.065 mhs/$
In general,  the mhs/$ ratio is better with 1060, unless you get really cheap 1050 TIs and you have additional hardware just laying around.
So yes, you will get money out of it (as long as you get 1050 TIs), but you would get more out of the 1060. And even more if you went with 480s but they are pretty hard to come by lately for an acceptable price.

What type of RAM you have on your cards, Samsung or other?
For my MSI Aero 1060 6gb with Samsung, these were the afterburner settings and the averaged results. The power limit apparently differ from maker to maker. Some get good results with 50%. It seems that underpowering decrease the hash rate quite a bit with everything else the same. On the other hand, once the card reached its max, giving more power has no further effect. For me, playing with the core speed had no effet pro or con, however, I'm doing Eth only mining. ~950 mem speed, the cards go over 25. However, you can start to experience issues on the long run.

Power|	Core|	Mem|	Avg Hasrate
50	|-200|	000|	18,8
54	|-200|	000|	19,7
58	|-200|	000|	19,7
62	|-200|	000|	19,8
50	|-200|	300|	18,1
54	|-200|	300|	20,9
58	|-200|	300|	21,3
62	|-200|	300|	21,4
66	|-200|	300|	21,4
50	|-200|	600|	17,6
54	|-200|	600|	20,3
58	|-200|	600|	22,7
62	|-200|	600|	23,1
66	|-200|	600|	23,1
50	|-200|	900|	17,0
54	|-200|	900|	19,9
58	|-200|	900|	22,3
62	|-200|	900|	24,6
66	|-200|	900|	24,8
70	|-200|	900|	24,7
70	|0000|	900|	24,7
70	|+200|	900|	24,7
62	|-200|	910|	24,6
63	|-200|	910|	24,6
64	|-200|	910|	24,7
65	|-200|	910|	24,8
66	|-200|	910|	24,8

https://investoon.com/tools/dag_size
It seems they updated the number post Byzantium and now it shows next April. Last time I checked, it was showing 2018 november or something. The thing is, as the difficulty increases, the hashing power decreases. With the hasing power decreased, the block time increases. The DAG epochs are not defined by date but by block numbers. When X block gets mined, new epoch starts, DAG increases... Originally, it was estimated that each epoch (ie X number of block number mined) will take about 5 days. Now lately, epoch took as long as 10 days due to the unexpectely high difficulty ramp up. That means the originally estimated 3GB limit hit gets pushed further and further.

On the other hand, with byzantium's difficulty decrease, the hashing power of the system increased quite a lot. This means, the epoch will go fast again and the 3GB DAG limit will hit sooner.
Please note that I'm no expert. This is only my (maybe wrong) understanding of the situation, after I read through several pages and sites.

POS is Proof of Stake. At that point, the POW (Proof of Work) GPU mining finishes. No GPU mining will work after that point. If I understand correctly, you will just need to keep your ETH wallet open on your PC to get something.

I see ~20-30% increase as well.

The question is, how will this affect the ETH price. Won't this increase of ETH supply drop the price like 30%? Also, how will this affect the epoch time and the POS and the DAG? It was calculated that the DAG will not grow over 3GB for another 12-14 months and by that time, it will be POS anyway. However, the drop of diff will make epoch pass faster and it may outgrow the 3GB cards before POS.

I have 4 single fan MSI 1060 6GBs and they do just fine. All Samsung RAM. Rock stable at @64% power, -200 GPU, +900 RAM. Two runs with 65C at 58% fan and two 57C at 32% fan.  The average hash rate is 24.8 Mh/s.  If I push them to +960 RAM, they go over 25 MH/s no problem.

My latest settings with Afterburner on 4x MSI 1060 6G Aero ITX (Win7, latest drivers, Claymore 10 Eth only):
Current uptime: 120h
Power Limit: 70
Core clock: -200
Memory Clock: 950
Fan: 67
Temperature: 62C, 45C, 50C, 63C
Power: ~78% (according to Afterburner)
The last 24 hours average
Total : 99.71 (min 94.69 - max 100.41)
GPU0: 24.93 (min 22.98 - max _25.14)
GPU1: 24.92 (min 23.08 - max _25.14)
GPU2: 24.95 (min 21.60 - max _25.16)
GPU3: 24.90 (min 20.53 - max _25.14)

Mining at Nanopool:
Current Average Hashrate for last 6 hours: 107.6 Mh/s
Current ETH gain: 0.01539/day

GIGABYTE GA Z270P D3 + G4400 + HyperX 4GB Fury DDR4 2400MHz +OCZ SSD + 900W PSU
4x MSI GEFORCE GTX 1060 AERO ITX 6G OC
Win7 64bit, latest drivers, latest Claymore (Eth only, no dual)

Rate: 96.5-97 Mh/s (24-24.3 Mh/s per card)
Power limit: 63
Core: -115
Memory: +800
Fan: 68%
Power draw: ~75 per card (by Afterburner)
Temp: 60C, 43C, 47C, 54C (Ambient 24C)

On lower power limit and core speed, the hash rate was less steady. It started to drop to the 23-23.5 Mh/s territory once in a while.
The temp is not the best because 3 cards are plugged into the MoBo and only one is separated on a riser.

Rate: 97.1-97,9 Mh/s (24.25-24.5 Mh/s per card)
Power limit: 66
Core: -100
Memory: +850
Fan: 73%
Power draw: ~75 per card (by Afterburner)
Temp: 60C, 43C, 47C, 54C (Ambient 24C)