Ethereum GPU mining?

[Staple]

Yeah this is my fear too. Also what are you paying for electricity?

Remember every minute of downtime is lost revenue and this scales proportionally the more Mh you are mining at. I haven't done the math but if it takes me roughly two-three days to tweak 9 miners is that worth switching over given the amount of downtime necessary to dial in the proper settings using Claymore? If on average I generate 2-3% more per miner how long would that take to even break even with the downtime. Also factor in the difficulty curve during dowtime and added power costs etc.

Paying about $.11 here. I never got the expected decred from the dual miner, and there were always a fair amount of rejected shares too. Because decred uses the GPU memory it gets REALLY hot and the power draw for me was at times 25% more than just mining ETH. Not worth killing the cards for a few dollars more especially when it didn't even look like the estimates were on target.

Here's a pic of decred mining heat generation: http://www.cryptomining-*********/7...ermal-images-for-ethererum-and-decred-mining/

 

[thilanliyan]

Paying about $.11 here. I never got the expected decred from the dual miner, and there were always a fair amount of rejected shares too. Because decred uses the GPU memory it gets REALLY hot and the power draw for me was at times 25% more than just mining ETH. Not worth killing the cards for a few dollars more especially when it didn't even look like the estimates were on target.

Here's a pic of decred mining heat generation: http://www.cryptomining-*********/7...ermal-images-for-ethererum-and-decred-mining/

I'm using dcr.suprnova and getting about 0.5% rejected shares for DCR. I seem to be getting about the expected amount of DCR. Of course, I've had to sacrifice some Eth because of the heat.

However, I'm not going to stick with the dual mining. I'm going to go for slightly less coins, but cooler cards with Eth.

 

[Madpacket]

Paying about $.11 here. I never got the expected decred from the dual miner, and there were always a fair amount of rejected shares too. Because decred uses the GPU memory it gets REALLY hot and the power draw for me was at times 25% more than just mining ETH. Not worth killing the cards for a few dollars more especially when it didn't even look like the estimates were on target.

Here's a pic of decred mining heat generation: http://www.cryptomining-*********/7...ermal-images-for-ethererum-and-decred-mining/

I've never had a rejected share AFAIK using the the standard ethminer according to the mining pools I'm at. That link you posted isn't working for me so I can't see the differences but I suspect if you're generating and storing the DAG on video memory it could add a lot of strain yeah.

Someone named "Genoil" did fork ethminer but I haven't tried it yet. Supposedly it gives better hashes for CUDA cards and has some nice features for OpenCL miners (auto DAG clean up, proper farm failover) and no fees taken.

I think we can all upgrade to the latest AMD drivers as well provided you dump the following DLL's from the 15.12 driver in the same folder as the Genoil miner directory.

"amd_opencl32.dll
amd_opencl64.dll
amdocl.dll
amdocl12cl64.dll
amdocl64.dll"

Worth checking out anyway.

https://bitcointalk.org/index.php?topic=1368785.0

 

[thilanliyan]

According to this site Ether is heavier on the memory and DCR is heavier on the core.
http://cryptomining-*********/7501-nvidia-geforce-gtx-970-thermal-images-for-ethererum-and-decred-mining/

So when we are dual mining, and up the DCR intensity, does it become less intensive on the memory and more on the core?

 

[poofyhairguy]

However, I'm not going to stick with the dual mining. I'm going to go for slightly less coins, but cooler cards with Eth.

Yeah that is the conclusion I have come to as well, it just causes too much heat to dual mine. Plus quite frankly the difficulty of Decred is rising so much that the extra 1% fee (of EVERYTHING including Eth) you get for it probably isn't worth it after power costs. I am done with Decred.

I am still a pretty big fan of the Claymore miner on my 390x though. For the 280x it seemed worthless, as my average was very close to qtminer and the effective rate seemed to have the same swings as qtminer. But holy crap, on the 390x it adds back the X! With qtminer my 390x would have effective rates as low as 27 and as as high as 38 at peak. Its average was around 31.

With Claymore miner I haven't seen anything below 30, and a good chunk of the time it is mining at WELL above 40 (like 41, 42!) which is just incredible. It reports 33, but its real average is above 35! I expected to average 50 between the two cards this month, but the Claymored 390x is pushing my average closer to 56!

It kinda makes sense when you think about it. Until now 390x/290xs hash really close to 390s/390x. That is why 390xs don't have the cost markup on Newegg, miners aren't buying them. But with the Claymore miner it is like I took the chain off that full Hawaii chip, well worth the extra 1% fee. Hell it makes me want to buy another full Hawaii (or maybe even a Fiji) real bad before more people jump on this. The only cheap full Hawaii on Newegg is this one:

http://www.newegg.com/Product/Product.aspx?Item=N82E16814202042&cm_re=290x-_-14-202-042-_-Product

But that seems a lot for a reference cooler on a machine that has an open case. I might just wait until the 390x clearance deals hit after Polaris 10 releases since other miners might not pick up on the unlocked potential. At the very least I am not selling my 390x despite Polaris probably dropping its value. Claymore's miner changed the payoff math considerably, I will let it mine until we can't GPU mine anymore.

 

[thilanliyan]

^For reference, my 290 at 1100/1250 was doing 60.5MHs with QTMiner, and 62MHs with Claymore. When factoring in the 2% fee, they were basically even, so I've just stuck with QTMiner. I will however say that there is much less variance in the hashrate with the Claymore miner, although the averages for me with both are about the same.

 

[poofyhairguy]

^For reference, my 290 at 1100/1250 was doing 60.5MHs with QTMiner, and 62MHs with Claymore. When factoring in the 2% fee, they were basically even, so I've just stuck with QTMiner. I will however say that there is much less variance in the hashrate with the Claymore miner, although the averages for me with both are about the same.

Note the 2% is just in dual miner mode, its 1% just for Eth. But I will admit maybe cut Hawaii doesn't get the same benefits, it might be like the 280x where its not worth it.

I wonder what the Claymore miner does with Fiji. Anyone mining with a Fury Nano?

 

[thilanliyan]

Note the 2% is just in dual miner mode, its 1% just for Eth. But I will admit maybe cut Hawaii doesn't get the same benefits, it might be like the 280x where its not worth it.

I wonder what the Claymore miner does with Fiji. Anyone mining with a Fury Nano?

Ah okay, well I turned it on to just Eth mine...see how it goes for the next 24h.

 

[Madpacket]

Note the 2% is just in dual miner mode, its 1% just for Eth. But I will admit maybe cut Hawaii doesn't get the same benefits, it might be like the 280x where its not worth it.

I wonder what the Claymore miner does with Fiji. Anyone mining with a Fury Nano?

I have two Fury X's. I'll test it out this weekend if possible. My Fury X's average between 32-35Mh currently with just ethminer.

 

[thilanliyan]

I have two Fury X's. I'll test it out this weekend if possible. My Fury X's average between 32-35Mh currently with just ethminer.

At what speed?

 

[DrMrLordX]

So uh what happened about two hours ago? It looks like a bunch of miners went offline. Nanopool lost maybe 60 GH/s in total (down to 146.2 GH/s). My stuff went down and required a DAG rebuild when I restarted them.

 

[Hitman928]

ethereumpoool had a similar issue earlier today.

 

[Staple]

^For reference, my 290 at 1100/1250 was doing 60.5MHs with QTMiner, and 62MHs with Claymore. When factoring in the 2% fee, they were basically even, so I've just stuck with QTMiner. I will however say that there is much less variance in the hashrate with the Claymore miner, although the averages for me with both are about the same.

The claymore miner only charges 1% if you're just using it in ETH mode. You also want to look at the watt draw from the wall. The claymore miner uses slightly less power for me and mines about 1.5mh/s faster in ETH mode. For me, that makes it worth it even with the 1% fee. Of course, it's not a huge difference but why not?

Dual mining is not worth it though, decred is largely worthless especially the way the difficulty has increased in the last few weeks while the price hasn't moved.

 

[Accord99]

So uh what happened about two hours ago? It looks like a bunch of miners went offline. Nanopool lost maybe 60 GH/s in total (down to 146.2 GH/s). My stuff went down and required a DAG rebuild when I restarted them.

Every 30000 blocks, the miners are supposed to used a new DAG but frequently they'll hang up or crash and stop working.

It's very random and I've seen mining systems that works properly over one 30000 block switchover fail at the next one, and vice versa.

 

[DrMrLordX]

Hmm okay. Well a lot of miners must have choked at once then. No skin off my nose, other than the ~2 hours of lost mining time I suffered on my end.

 

[thilanliyan]

The claymore miner only charges 1% if you're just using it in ETH mode. You also want to look at the watt draw from the wall. The claymore miner uses slightly less power for me and mines about 1.5mh/s faster in ETH mode. For me, that makes it worth it even with the 1% fee. Of course, it's not a huge difference but why not?

Dual mining is not worth it though, decred is largely worthless especially the way the difficulty has increased in the last few weeks while the price hasn't moved.

Since it creates the DAG on the GPU does that mean the claymore miner is harder on the vram? I'd rather just have the DAG on the hdd if it's less stress on the card.

 

[reb0rn]

Claymore mines is not harder on ram, it works just fine on 2GB cards... for me reson to switch was it does not crash on DAG update like the original eth miner, and looks like its easy on RAM usage also

 

[Shmee]

I just tried the 15.12 DLLs in the same folder with the new Genoil-ethminer, and can confirm reported hashrates have gone back up. Currently on 16.4.2

 

[DrMrLordX]

Hey you Linux miners, I have some questions here that you might be able to answer.

Currently I execute the miner (still using ethminer, not sure about the newer software but I'll wait-and-see) with a line like this:

ethminer -F http://us-east1.nanopool.org:8888/address/miner3 -G --cl-local-work 256 --cl-global-work 16384 -t 4 --farm-recheck 100

Yes I'm still using nanopool, and no I haven't bothered to set up a Stratum proxy yet but that'll come later. That particular machine has 1x 390, 2x 290s, and a 270. Works pretty well. But I got a crash the other day and I wanted to narrow down which card had problems, so I wanted to separate the mining into separate processes executed by separate users. So I was going to do something like this:

echo password | sudo -S -u miner3a ethminer -F http://us-east1.nanopool.org:8888/address/miner3 -G --cl-local-work 256 --cl-global-work 16384 --opencl-device 0 --farm-recheck 100

That line returns three protocol errors (huh?) and then one more error indicating that a device with 2Gb memory could not be found please remove -G option. If I try this line:

echo password | sudo -S -u miner3a ethminer --help

It executes ethminer and shows the help as normal without any obvious errors. So something is going wrong when I try to pass multiple arguments to ethminer at the end of my sudo invocation, and I'm not sure why?

edit: I ran into the same protocol errors when manually switching a terminal to another user (su - miner3a). What I had to do was change the user to an admin type, switch user with su, and then use the sudo command to invoke ethminer. Anything else and it would not access GPUs for mining. So I can do it manually . . . now I need to tool around with sudo -S -u to see if I can do it in a single script invocation.

Another issue: whenever I get a driver crash from unstable card settings or whatever, I get a situation where one core of my quad (Phenom 9750, whee) is pegged at 100% while all operating processes show 0% CPU time. Attempts to reboot the machine just cause it to lock up while exiting the desktop environment, forcing a hard reboot/power cycle. Any way around that?

 

[RussianSensation]

Prices on Maxwell cards are dropping a lot.
http://slickdeals.net/f/8731567-evg...icro-center?src=SiteSearchV2_SearchBarV2Algo1

I have a spare Core 2 Duo & Asus 650i SLI board that hates AMD cards. Was thinking of picking up a pair of NV cards for ether.

Can someone with NV's GTX970, 980, 980Ti please provide rough estimates of their hashing rates so I can confirm my estimates (17-19MH/sec for 970, 19-24MH/sec for 980, 22-24MH/sec for 980Ti)?

This is what I found so far.

970 ~ 18 & 970 OC ~ 19 Mh/sec

1399mhz Galax 970 with Windows 10 Pro and 347.52 drivers shows 21.6MH/sec
vs.
1417mhz EVGA 970 with Windows 7 and 361.75 drivers shows only 17.5MH/sec

Gigabyte G1 980 has 24MH/sec (!) but 980Tis have 23-24MH/sec? No scaling?
The rates are all over the place in the database here.

 

[DrMrLordX]

I'm not sure if Maxwell is worth it yet at those prices? They'll chew up less power than Hawaii I'd guess, but I'm not sure power/MH/s will wind up favoring Maxwell?

 

[poofyhairguy]

I have seen 970s are 17 and 980s around 21 at their best.

But I wouldn't know how that is worth it. My 280x does better than 970 for about the same power when OCed, except the 280x can be found on Newegg for $180 Canadian:

http://www.newegg.ca/Product/Product.aspx?Item=N82E16814202138&cm_re=280x-_-14-202-138-_-Product

I don't see how any Maxwell card is even close to that value.

 

[poofyhairguy]

Hey you Linux miners, I have some questions here that you might be able to answer.

Currently I execute the miner (still using ethminer, not sure about the newer software but I'll wait-and-see) with a line like this:

ethminer -F http://us-east1.nanopool.org:8888/address/miner3 -G --cl-local-work 256 --cl-global-work 16384 -t 4 --farm-recheck 100

Yes I'm still using nanopool, and no I haven't bothered to set up a Stratum proxy yet but that'll come later. That particular machine has 1x 390, 2x 290s, and a 270. Works pretty well. But I got a crash the other day and I wanted to narrow down which card had problems, so I wanted to separate the mining into separate processes executed by separate users. So I was going to do something like this:

echo password | sudo -S -u miner3a ethminer -F http://us-east1.nanopool.org:8888/address/miner3 -G --cl-local-work 256 --cl-global-work 16384 --opencl-device 0 --farm-recheck 100

That line returns three protocol errors (huh?) and then one more error indicating that a device with 2Gb memory could not be found please remove -G option. If I try this line:

echo password | sudo -S -u miner3a ethminer --help

It executes ethminer and shows the help as normal without any obvious errors. So something is going wrong when I try to pass multiple arguments to ethminer at the end of my sudo invocation, and I'm not sure why?

edit: I ran into the same protocol errors when manually switching a terminal to another user (su - miner3a). What I had to do was change the user to an admin type, switch user with su, and then use the sudo command to invoke ethminer. Anything else and it would not access GPUs for mining. So I can do it manually . . . now I need to tool around with sudo -S -u to see if I can do it in a single script invocation.

Another issue: whenever I get a driver crash from unstable card settings or whatever, I get a situation where one core of my quad (Phenom 9750, whee) is pegged at 100% while all operating processes show 0% CPU time. Attempts to reboot the machine just cause it to lock up while exiting the desktop environment, forcing a hard reboot/power cycle. Any way around that?

I would suggest using qtminer. I got it to work VERY easily on Linux.

 

[RussianSensation]

I'm not sure if Maxwell is worth it yet at those prices? They'll chew up less power than Hawaii I'd guess, but I'm not sure power/MH/s will wind up favoring Maxwell?

I am not buying this week but just want to see what my options are. If prices fall a lot over the next month, I'd like to be able to throw in 2 NV cards into this LGA 775 SLI board.

I have seen 970s are 17 and 980s around 21 at their best.

But I wouldn't know how that is worth it. My 280x does better than 970 for about the same power when OCed, except the 280x can be found on Newegg for $180 Canadian:

http://www.newegg.ca/Product/Product.aspx?Item=N82E16814202138&cm_re=280x-_-14-202-138-_-Product

I don't see how any Maxwell card is even close to that value.

Ya I get it. I have a spare 7970 that's just sitting there and doing nothing. I'd sell it and offset the cost of the Maxwell cards. It seems a single 970 earns roughly $50 USD a month after electricity. If I can score 970/980 for $250 USD, even if it takes 5-6 months to pay off 2x 970s or 2x 980s, I'd sell them later and put that $ towards next gen cards. I have the CPU, RAM, SLI mobo, heatsink. The problem is this SLI board doesn't work with my R9 295X2 or my 7970. I tried everything from flashing new BIOS, etc.

I also may put together another Skylake or Broadwell-E rig. That one can house Polaris 10, R9 390s or GTX1070s if they are any good. Wish we had a better idea of the launch date of Polaris 10 but it will take until the end of the month to find out. R9 390s are still holding their prices.

 

[poofyhairguy]

The problem is this SLI board doesn't work with my R9 295X2 or my 7970. I tried everything from flashing new BIOS, etc.

I see. So it has the 390, some other machine (gaming box I assume) has the 295x2 and the 7970 is wasted. That is a shame, Tahiti is a mining champion. I get not wanting to spend more, I have three rigs running now to avoid having to upgrade a PSU to consolidate two into one. All those extra costs beyond the GPU screw up the payoff equation.

Good thing about Nvidia cards is their value bottoms out at a higher place on average.