Question. Occasionally, my rig half-dies. I say half die, because I get a "watchdog error: GPU hangs in openCL call. Restart miner" error. But, while windows hangs, all the other cards will continue to mine.

Looking at the log when this happens, it'll give that error a few times before windows hangs. So, I'm wondering if there's something I can add to either start.bat or config that will tell claymore to restart when that error first occurs?

Hey all,

I'm in the process of pricing up a rig expansion. Currently running 10 x 580s and 2 x 560s. An interesting idea has come up, but it has pros and cons. Opinions needed and welcomed.

Right now, it doesn't take much digging to find some of the outgoing generation of GPUs at stupidly low prices. Currently, I'm finding both 1050ti's and 1060s (just the 3gb version) for sub $100. So, based on my budget for expanding, I can get a bunch of these up and running pretty rapidly. So, I'm toying with the idea of around 24 of these. My figures, based on an additional 24 cards, is a ROI of around 90 days at current prices.

Here's some of the glaring concerns. Opinions on these would be massively appreciated:
1. The 1060s are 3gb cards, so Ethash is going to be tricky. Similarly, the likelihood of it being a decent mem manufacturer like Samsung, is probably fairly low. Worst case scenario, it's a 3gb card with crap memory, what's the next best algo to mine?
2. While there's still no word on the new NVidia cards, I think we're all expecting them this year. So, is 24 x 1060s going to be better than, for example, 2 or 3 of the new gen cards?

I'm busy throwing figures into spreadsheets like it's going out of style. Ideas and opinions are welcomed.

Hello all,

I'm getting very intrigued by the whole idea of margin lending. My 9-5 job has gotten much busier, so I've had to step back from trading. My idea was that I could take my mining returns and easily put them into some sort of BTC margin lending *NOT MARGIN TRADING, JUST MARGIN LENDING*. From what I can tell, it seems to be a good way to increase returns at a reasonable rate, at least better than hodling. From what I can tell, it would increase my BTC holdings over time, and that's good enough for me at the moment.

The problem is that bitfinex registration is closed, and more importantly, has that 10k minimum registration amount thing. So I'm looking at alternatives. Cryptolend paired with poloniex and bitmex seems to be a good option. Wondering if anyone had any experience/recommendations/tips for this?

Thanks

I spoke too soon, it's down again. But, I think I've figured it out.

I was watching as it happened. When claymore switched to mine for devfee, it couldn't connect at first, so the gpus weren't running at 100%. The second it connected and the gpus went to fire up, it died.

So it's something when the gpus turn off or go from 100% to 0%. Its the whole PSU that turns off. So I'm gonna order an add2psu and see if that does it.

And to make matters worse, I can't get sgminer to work since the fork yesterday. Got the updated cryptonight.cl file, but won't start at all.

Fuck.

Yeah, it really isn't the best way of powering risers. However, I've had no problems with them so far. They don't even feel particularly hot after long bouts of running. I've a tonne of airflow across each rig. it's two rosewill 4u server racks, with 5x120mm high-airflow fans in each. They keep lovely and cool!

I'll keep an eye on things. My theory is, because ethash is more power hungry, if it's stable on this, I'm in the clear.

Thanks again for the help.

They have 4 connections. Two have 2 x 8 pin, and two have 1 x 8 pin. They power the pci connections. For the risers, they're sata type, powered in pairs.

I think I've managed to solve the problem. I swapped the GPU distribution around to 5+1 (580/560) per PSU. Mining for 8 hours on claymore ethash and no problems so far. 328Mhs with 710W per PSU. Seems about right?

I just tested this, it's very little difference. Even increasing voltage from 815 to 850 still caused the same effect, and still a good 400W below the 1000W max.

I'm trying a new power setup. Different outlets for each PSU. If I'm still having problems, I'm going to switch the two psus around, because they're the same.

I take it cables are generally viewed as an inferior solution?

Oddly enough, I had to buy 2 different sets of psu splitter cables because the first one wouldn't allow either PSU to turn on.

Out of interest, what is the best way to run two PSUs? Is add2psu as good as a decent cable or is there s better option?

I don't think it's that. I'm using overdriveN, and even if they all reverted to stock voltages, the draw on cryptonight would still be less than 1000W on the second PSU.

I suspect the surge protector thing could be the culprit too. It's a slightly beefier version of just a box standard one. The tricky thing is that even though the rig is beside a 2-outlet socket on the wall, space is tricky. I have a powerline plug for the network cable that won't work on a powerbar, and the next nearest outlet is quite a bit away. I think you might be right though that the one powerbar is essentially getting overloaded. I'll test it tomorrow.


I think it's the first option. The PSUs are 2xRosewill Hive 1000W. I think it's the powerbar thing though.

Cheers for the ideas, both of you.

I've had this problem before, it fixed itself for a while, but now it's back with a vengeance.

Currently running 10x580s and 2x560s, W10 Pro, 2x1000W PSUs, latest amd drivers, sgminer or claymore depending on the algorithm. On ethash, using around 600W on 1 PSU, 510W on second, measured at the wall.

Mining runs fine for a varied amount of time, anything from 1hr to 2 days. Then randomly, it's as if the 2nd PSU just shuts off. The fans etc stop, and wattage measures 0. I've tried checking risers, cables etc and I can't seem to identify the problem. The only thing I can think of it being is some sort of surge or overload at the wall or powerbar, but I'm using a 20A 2000W surge protector.

Any ideas at all, either on how to narrow down causes or find a solution? Happening very frequently now.
 
Thanks all

I don't think that's going to happen. I like the flexibility/familiarity of W10. It's been stable up until recently.

but thanks for the suggestion

Hey all,

Having a very strange problem with my rig. Currently running 10 x RX580 and 2 x RX560, W10Pro, 2 x 1000W PSU, H110 mobo. The problem is that the rig will run stable for a while - anything from an hour to a week. Then half the rig will crash, but it's an unusual crash. It's as if the second PSU overloads and shuts itself off (all the fans are off, cards are off etc.) The odd thing is, there's less load on that PSU. The first PSU runs 6 x RX580s, and the mobo. The second PSU just powers 4 580s and 2 560s. Both PSUs are identical, and connected by one of those 24-pin Y cables.

So far, in an effort to eliminate variables, this is what I've tried:
- Tried various algos, Ethash being most demanding (in terms of Wattage) and Cryptonight being the lowest. Currently on CN, which also has 0 HW errors on all cards EXCEPT ONE.
- Tried various drivers, 3 or 4 from Blockchain drivers to current adrenaline ones. Currently running Crimson 17.11.4, which has proven to be the most reliable for my setup
- Measured Wattage at the wall. Second PSU is around 450-500W draw.

Things I haven't tried:
- Replacing risers (one of the 580s on the 2nd PSU throws up a number of HW errors, not sure if this is the cause)
- Getting a new PSU (reluctant to do this)
- Crying

Any ideas?

Hey all,

Having a strange situation where half my rig crashes. I'm running two PSUs, and after mining for a while, maybe 8-10 hours, the second PSU just switches off. I've checked wattage at the wall, and it's WAAAAY below the 1000W rating for the PSU. So, in eliminating variables, I've had difficulty with Radeon Settings/Wattman reverting to stock voltages, pushing wattage up higher. So I want to install drivers without settings, but still switch to compute mode.

I have a vague recollection of someone having a clever little piece of software that would do that, without having to re-install stupid Radeon Settings. Anyone know what it is? Google has failed me on this one.

Thanks,

I am considering that. But I wonder if failover only would work better, or would it still only try a few times

No problem.

I think there's something with either failover or no-client-reconnect, but I'm just struggling a bit to understand the commands.

I didn't think that was in sgminer, and isn't that the opposite of what I'm looking for? I thought no-retry was that if it couldn't reconnect, it just went to the next pool?

Afaik, in cgminer, there's --r 0, which means it'll attempt a reconnect once. I essentially need that, but --r 9999999

Sgminer has a habit of quitting after one or two attempts to reconnect to the stratum, in the event of a connection loss for example. I've gone through the readme, but I can't seem to find a command to either tell it how many times to retry, or to disable the whole quitting thing.

Any ideas?

I had success with those before, but I think they only support 8 or 10, I need 12. I know there was one shortly after that worked for me before, I think it was the November one, but I'm hoping someone here might know which one. I remember there was a few that if I didn't install amd settings, it would assume it was doing graphics and my hashrate would suck on anything it cryptonight. Ideally, I need a stable pre-adrenalin one, where I don't need to install amd settings.

I just wanna go back to ethash on sgminer

Edit: now that I think about it, maybe those blockchain drivers were multi gpu. I'll give them a go!