Ethereum GPU mining?

[IntelUser2000]

So Tweaktown reports the Bitmain's F3 Ethereum ASIC miner(the one with 72GB DDR3 memory) gets 200-220MH/s with pricing at $2500-3000. They assume it can get better with final hardware.

Let's assume 250MH/s. Assuming it arrives tomorrow, it can make $20/day at current prices. At $2500 it can make ROI in 4.5 months. If it does 220MH/s and costs $3000, ROI would be 6 months.

That only seems attractive in light of recent events where GPUs are selling for $400-500 per card. Those that got their rigs in the months when it was at MSRP has no reason at all to consider it.

Update: Pricing seems crazier now than a month ago when the profits and value of the cryptos were higher.

 

[Charlie22911]

I don’t think this will help alleviate GPU strain at all. I doubt they will get DDR3 in enough quantity to ship significant volume.

 

[[DHT]Osiris]

I don’t think this will help alleviate GPU strain at all. I doubt they will get DDR3 in enough quantity to ship significant volume.

Yeah, this is the problem with an investment model that presents the same ROI regardless of investment input, it permits someone to spend an arbitrary amount of money and be in the green in the same amount of time, which makes people just want moar moar moar.

 

[Madpacket]

220Mh isn't that great and there's no mention of power consumption? (suspect it'll be pretty high for an ASIC given the memory bandwidth requirements) for these Bitmain machines so I'll reserve judgement on their value once we have more info.

No wonder it's taken this long for someone to bother.

Either way, hopefully this drops the price of GPU's for regular consumers but I doubt it'll happen anytime soon.

 

[IntelUser2000]

220Mh isn't that great and there's no mention of power consumption? (suspect it'll be pretty high for an ASIC given the memory bandwidth requirements) for these Bitmain machines so I'll reserve judgement on their value once we have more info.

Yea, bandwidth is a big problem. And I assume the reason they had so many memory chips were to result in enough bandwidth. 220MH/s at 90% efficiency requires 2TB/s bandwidth. We're probably talking 200W for the memory chips alone. The memory interface on the ASIC chips are going to take power too. That leaves rest for everything else.

I think doing the whole thing at 500W would be a marvelous result, just that its probably 2, or even 3 times that and even at 500W the upfront cost of the system won't be cheap.

Move to PoS is what will result in dropping of GPU prices. Scalability solutions are underway(like Sharding) and PoS is kind of a requirement to do so. I wouldn't be surprised at hybrid PoS by end of year/early next.

 

[Headfoot]

You haven't read the Ethereum whitepaper have you? You should go up a few posts and read the paragraph I quoted. I guess the community can always vote to hard fork for a new algorithm. Unlike the previous years, there's already a testnet for Casper.

There's also development going for scalability solutions. PoS is easier to make more scalable solutions because of the nature of PoS. It's easy to buy enough hashing power to do a 51% attack - its another entirely to do a 51% attack by owning 51% of total staked Ethereum. Also PoS penalizes malicious users by killing their entire stake.

I've read the part you quoted and I don't believe it. He's just guessing at human behavior in an economic sense, which is notoriously hard to predict. Plenty of people will want to "get rich quick" using ASIC miners and will have a vested interest against a poison pill contract like that. I've also read the Casper paper and I know there is a testnet for it, which is why I would not buy an ASIC miner for Ethereum even if it existed because Casper isn't academic anymore like it was previously.

I do think Casper or hybrid PoS of some flavor is close based on everything I've seen. I dont think its pie in the sky like it was 2 years ago

 

[Gellért]

Is it even worth it to mine with a small home setup with 2x 1080ti?

 

[Feld]

Is it even worth it to mine with a small home setup with 2x 1080ti?

If electric costs for 24/7 operation would be less than $150/month for you to do so (a rough estimate of the expected monthly payout from your cards at the current price and difficulty), then yes.

 

[IntelUser2000]

If electric costs for 24/7 operation would be less than $150/month for you to do so (a rough estimate of the expected monthly payout from your cards at the current price and difficulty), then yes.

Whattomine is saying $250-300 for 2 1080 Ti cards.

For new owners of 1080 Ti cards, the ROI time may approach a year, even when bought back before GPU price rise. Personally I would approach it as a gamer that can reduce the burden of buying a new card. Many gamers can use mining to upgrade to a new, "free" video card when it comes out.

 

[Feld]

Whattomine is saying $250-300 for 2 1080 Ti cards.

For new owners of 1080 Ti cards, the ROI time may approach a year, even when bought backbifore GPU price rise. Personally I would approach it as a gamer that can reduce the burden of buying a new card. Many gamers can use mining to upgrade to a new, "free" video card when it comes out.

The price of ether went up a little bit, but not nearly that much. Whattomine is showing me $5.67/day for two 1080ti cards (70 MH/s total), which is $170/month before expenses. But that's mining Ethereum, which I believe he was asking about. There are other coins those cards can mine for close to 2x the profit though, which may be what you're referring to.

 

[IntelUser2000]

There are other coins those cards can mine for close to 2x the profit though, which may be what you're referring to.

Yep. If I wanted Ether and I had 1080 Ti's I would be using NiceHash or mining ZCash with Claymore.

 

[IntelUser2000]

Technically though, it is possible from the specs they outline. They are talking about putting 576 separate DDR3 chips that potentially can deliver 12.8GB/s for a total hash of 900MH/s if each chip has its own channel. It would be an incredibly complicated board consisting of over 20,000 copper lines for the memory chips per board, but possible.

I have to correct my post, because the calculation is wrong. 220MH/s makes sense with the proper calculation.

I assumed each 1Gbit chip would achieve 12.8GB/s, which requires 64-bit width running at 1600MT/s. However, DDR ICs that are sold individually only have 16-bit width. So the bandwidth is quarter of what I said earlier.

16-bit/DDR3 IC x 1.866GT/s x 32 DDR3 ICs/ASIC x 6 ASICs/board x 3 boards
=2.15TB/s maximum bandwidth.

Ethash, assuming 100% efficiency needs 8GB/s bandwidth for 1MH/s. That turns it into 269MH/s, theoretical.

In practice, utilization of memory controllers aren't perfect, and top of the line mining software has efficiency in the range of 90%. At 90% efficiency for software and 90% for hardware, we get 220MH/s.

 

[Headfoot]

so really it'll boil down to what sort of power consumption we're looking at for 220 mh/s, since you can easily get to that kind of throughput using modern GPUs. Interesting calculations though, thx for the info

 

[Red Squirrel]

Doing rough calculations I pretty much figured that it's almost always going to be worth it to mine even with power consumption. Keep in mind that more than half the hydro bill is fixed charges so you're paying at least over $100/mo no matter what. The usage is pretty much the amount after that and is a smaller portion of the bill. If you figure a VGA uses 200watts (which is generous but I like to always think worse case scenario) then that's 4.8kwh in a day per card. At 13.2c/kwh which is my peak rate, that comes up to 63c per day. In reality it's lower, because the card uses less, and the rate is not the same throughout the day. But as long as I can make more than 63c per day per card then I know I'm profiting.

Basically even if you double your household electricity usage, your bill is not actually going to double, because the majority of your bill is fixed charges.

 

[Markfw]

Question, do you actually get a check or a deposit every month for your mining ? Not just credits somewhere, real CASH

 

[Red Squirrel]

Mining itself, you get crypto currency. So for Ethereum mining you get Ethereum. It goes in the wallet address you specify when mining. You can then move that currency to an exchange and then buy other crypto or dollars. I officially made around $25 so far but I have around $50-$100 worth sitting in ethereum. Price is kinda low right now so holding on to it.

 

[VirtualLarry]

Question, do you actually get a check or a deposit every month for your mining ? Not just credits somewhere, real CASH

Yeah!!!

I get like $20-25 every few days deposited into my bank account.

 

[Feld]

Question, do you actually get a check or a deposit every month for your mining ? Not just credits somewhere, real CASH

If you want to, you can do it that way, yes. Or if you think the value of the coins you mined will go up in the future, you can choose to hold onto some or all of them instead. The choice is up to the miner. Though regardless of what you choose, you pay income tax on what you mine according to its value at the time you mine it, and if there's any difference in value between when you mine it and when you sell it then you pay capitol gains tax (or take a capitol loss deduction) on the difference (at least in the US).

 

[TempAcc99]

Is it even worth it to mine with a small home setup with 2x 1080ti?

Depends on what is "worth it". my personal opinion is no due to the hassle. With 2 1080Tis you make with current prices and minus power cost maybe $500 per year. But greatly depends on your local power costs and of course the value of ETH. You for sure aren't getting rich of it.

 

[ultimatebob]

Yeah!!!

I get like $20-25 every few days deposited into my bank account.

How does that work? Are you sending your crypto directly to an exchange that's cashing it out for you? Mine just sends the piddly amount of crypto I mined to my Ethereum wallet. I needed to send it to an exchange like Coinbase if I want to cash it out.

 

[VirtualLarry]

How does that work? Are you sending your crypto directly to an exchange that's cashing it out for you? Mine just sends the piddly amount of crypto I mined to my Ethereum wallet. I needed to send it to an exchange like Coinbase if I want to cash it out.

NiceHash -> NH internal wallet -> CoinBase -> Sell -> Bank Account

Rinse, repeat, every few days.

Edit: Yeah, CoinBase $1.99 "sell" fee sucks.

 

[[DHT]Osiris]

Depends on what is "worth it". my personal opinion is no due to the hassle. With 2 1080Tis you make with current prices and minus power cost maybe $500 per year. But greatly depends on your local power costs and of course the value of ETH. You for sure aren't getting rich of it.

Those numbers are a little off, depending on electricity costs I guess. I'm doing ~$700/mo at minimum with 5x 1080ti's, so 2x should gross him ~$250/mo or so at absolute minimum min. Power costs aren't that outrageous unless you're paying like, >$.15/kwh or something.

 

[[DHT]Osiris]

Question, do you actually get a check or a deposit every month for your mining ? Not just credits somewhere, real CASH

What others have said, you either a) keep the raw coins you mine, b) have coin proceeds coverted into a 'higher level' coin, like BTC or Eth and delivered to a wallet of your choice (I use coinbase at the moment) and either HODL that until it goes to the moon/craters, convert it to some other coin, or just sell it off for $.

 

[IntelUser2000]

Geth 1.8 is out, and it says it has done a lot to fix light sync problems. Should try it out.

https://blog.ethereum.org/2018/02/14/geth-1-8-iceberg¹/

The problem with disk space taking up an amazing amount over time is significantly reduced too.

 

[thilanliyan]

Geth 1.8 is out, and it says it has done a lot to fix light sync problems. Should try it out.

https://blog.ethereum.org/2018/02/14/geth-1-8-iceberg¹/

The problem with disk space taking up an amazing amount over time is significantly reduced too.

I finally "upgraded" my storage drives to WD 2TB Black drives, from my nearly 10 yr old 640GB Black drives, so I will finally run a node with one of the 640GB drives for both Eth and XMR.