same here run first in win 10 but moved to linux after some overclocking crashes; I have more stability in linux using the same settings. Currently I’m using 1080 ti asus strix and zotac amp extreme. I’m planning to build a 7 to 9 rig but wanted to check about the 20amp circuits. Do you have more info about 20amp circuits? Why the limitation or if anyway you can check your circuits capacity. Anything can help. Thanks
I have dedicated 15 and 20 amp mains lines to my rigs (GTX 1080 20 amp, AMD 15 amp). If you live in the US or a country with 110-120V mains then 15, 20, and 30 amp are standard breakers with 14, 12, and 10 gage wire respectively. A 7 GPU GTX 1080 Ti rig at full power will draw about 1900 watts (measured at the wall), or 15.8 amps with a mains voltage of 120. This is a 79% load on a 20 amp 12 gage circuit (so all is good, below 80%). For a 8 GPU rig its 2160 watts, 18 amps, and 89.9% load (NOT SAFE).
Do NOT believe what the GPU driver tells you the load is, its NOT correct, and can be a very costly mistake. When you are running circuits within 1% of maximum continuous load then its important do do things correctly and use a wall watt meter. I like being this close to the trip point of the breaker on my big rigs as (4.2 amps) a short or other fault is more likely to trip the breaker before anything bad happens. If I am using a 30 amp line for an 8 GPU rig then I need short or fault of 12 amps to trip the breaker.
Having said that, the breaker is not meant to protect anything you plug into the circuit. Its sole function is to protect the wire in the wall from starting a fire. 10 gage wire will safely carry 30 amps, regardless if the rig plugged into it is on fire or not. That is why I use 15 amp lines on my AMD rigs since they only draw 1200 watts.
Also, plugging in two PSU’s from one rig, into two different 15amp circuits is not a good idea. If one breaker trips and the other stays on then what will happen if its the secondary (Slave PSU) that loses power and not the master. The master PSU will stay powered and the slave will go off and half your rig will be powered. Depending on how you have your rig wired this could take out your mobo, GPU’s or worse.
I think it may have to do with Pcie lane utilization. If I plug one GTX 1080 TI into a x16 slot on the mobo I can get 770+ Sol/s easy (Gen3 speed). If I add a second GPU that number goes down and a third and so on (all depends on the mobo and Pcie speeds). By using the CPU to mine as well, it is somehow making better use of the lanes available to it? The chip set also provides indirect Pcie lanes but they are not the same as the direct CPU Pcie lanes. Perhaps mining with the CPU forces the use of the direct CPU lanes instead of the indirect chip set lanes. Just a guess, but its worth further exploration.
I really like that the Claymore miners include a watchdog and will call a reboot.sh file when the miner gets stuck. It’s a real pity that the EWBF miner doesn’t offer the same.
BTW, if anybody is wandering what @ZC93 is talking about, there are a set of low-level kernel commands that you can use on a Linux box to reboot in an emergency i.e. when the OS itself can no longer reboot the machine. The commands can be typed or called non-interactively (which is most useful for mining rigs).
Below is the script I use. It’s a cutdown version of the classic REISUB sequence. The SIGTERM and SIGKILL have been removed as killing the parent process (the script itself) before rebooting would be a Bad Idea™. The script sleeps for a few seconds to ensure that the syslog has had enough time to flush to a graylog server. I like capturing the reboot event and system state and, if necessary, triggering a text message alert.
I’ve mentioned above that my fav 1080 ti cards are the Corsair Hydro all-in-one cards. Just be aware that some of these going out as “new” cards might be missing the plastic sticker covers over the glass and the water-cool connector to the card may be slightly damaged. The first few batches of cards were fine but in my latest batch they all had these issues. One had a nice big fingerprint on it also. Ironically the one card that did have the new-card sticker had an issue with the blower fan.
Overall I’m pretty disappointed. Worth noting that Corsair customer service was easy to work with and is exchanging them/offered refund without any hassle.
These are expensive cards and I did expect more. Perhaps to keep up with demand they are cutting corners on quality. Just keep an eye out if you’re getting any of these.