Discussion in 'Blockchain & Cryptocurrency Mining' started by DarkStyle, Dec 11, 2017.
Play a game /benchmark
Edit: What Myne said...
I see Bitmain released a Zcash ASIC miner. Good specs, 10k Sols/sec. 300W power consumption.
However it's $1,999US which means it will take months to make even a cent profit.
If you keep your $1,999 in BTC you would have made 4% profit today alone.
I do around 400-500w per 1200w HP PSU, with very little noise. It's more about getting the number of plugs needed. At 75W a card, that is around 5-6 cards per PSU. I think if you need quiet PSUs and more power the Corsair HX1200s are good.
Sounds great. I used to use Linux on USB when I mined. It was the most efficient way. Brings down rig cost (don't even need a drive) and does exactly what a miner needs.
Would you guys rather mine with 1x Vega or 3x RX560? With a view to adding more gpus afterwards. Im currently mining c7 to nicehash with my AMD gpus.
Whats the best value eth mining setup?
If the cost for both alternatives is the same, Vega. If its "close" Vega.
Can tell you that a ref Vega will do 1950-2050 h/s on c7 for about 140-160w...a Hynix mem v56 will do about 1850hs with same power draw.
Pretty sure that's at 4x 560s (@450-550 h/s depending on memory)
Have read that pretty much all 64s have Samsung mem, so those might be what you'd want to look out for if you can only get aftermarket units. If you end up with Vegas that is.
Testing this out after windows update breaking stuff and the rigs needing more resets lately. Seems awesomely simple and effective, so i'm hoping it proves itself with the uptime.
My only gripe is i can't get sub 900mV on my rx570's. I used to use OverdriveNTool in windows to force it down to the 850mV range. So im stuck with the power draw up ~15%
Any one have any great info on undervolting in linux OS's?
Not me, I just used to lock in the desired values in the cards' BIOS. Not sure if that's still a viable option.
I've tried to, but from playing around in windows, any bios setting such as 850mV just ups and reverts itself to an actual 900mV in gpu-z. The clocks, fan speed etc work as desired, it's just the mV that seems to have a minimum 900mV threshold
Could always just up the MHz to backfill the power increase if it can't be lowered. May as well if you're already stuck with higher power usage.
Or if it bugs you more than windows did, theres windows?
Haven't had much luck getting over 1930mhz, mostly elpida ram in the later batches of cards. Gets 27Mh/s so can't complain too much
I'm sick of having to babysit them periodically, so i'll probably just bite my tongue and live with linux, even if it means dropping from 8 cards to 7 where need be
babysit? Why do you babysit?
I check nanopool maybe twice a day, and then usually every 2-3 days one of the rigs (a different one every time) will have dropped out, so i have to go find it and reboot it, and more recently even have to reinstall drivers and/or windows.
Seems to be getting worse of late, i used to get 2 weeks+ uptime easily.
Another example, if i need to go to do some welding in the shed, then one particular rig will always freeze up (despite a few of them being on the same circuit).
Ironically my desktop has a simple batch script to run dstm's miner on boot, and that's ran 2+ months without a single issue, and doesn't interfere with everyday use
Weird. Mine's pretty reliable. It only plays up when I play with it (bios, drivers, cards etc) But I use w7.
Try underclocking the CPUs to prevent freezes. Drop the core and ram to the lowest you can manage.
It could be crappy ram, or something getting too hot or the power draw causing slight voltage changes. Best to reduce the unimportant load to the minimum.
my Elpida cards went from my worst, to my best with a memory strap mod recently.
I run them @ 1150/880mv 1975/880mv
They can do 1250/906 2050/930, but the gain is roughly 70h/s but the power increase is not acceptable.
Basically each machine "gains" half a cards worth of performance, for the power cost of a full card.
Both settings are stable however.
Perhaps bump your current settings 0.05mv on core/mem to see if stability increases.
Regarding your welder, perhaps try a power meter somewhere on the circuit ($20 unit from Bunnings are pretty good) and check your voltage when your gear is all turned on. You may have issues if it gets much lower than 220v (230v is nominal).
Your issues could also be attributed to dodgy risers/usb cables
Welding?! I didn't see that.
Welders blast EMI.
I'd suggest your boxes aren't as tolerant as a nuclear reactor's management.
426: Interested in more on your cards/ROM. Mine get a bit over 27 with elpida on 1900mhz/850mv
I had a few reboot issues with new builds but most were solved with powering each riser and GPU off the same pay psu cable, (found with the help of the great crew here).
Other than that I had some reboots due to pausing win update instead of disabling it, sorted now.
This is the strap I'm using:
Just copy it over the 2000/1750 straps of your original bios and you're good.
Going from the "one click" timings on PBE, I went from 860 h/s to 930 h/s at the same timings. (1133/1955)
1233/2050 gives 1020 h/s, using intensity 54 in SRBminerCN. it goes to 1030 at intensity 58, but wasnt stable compared to the 54 setting.
The cards are Asus STRIX 570s. Some guys are getting higher hash rates with their own straps, but nobody is giving those away, so the above was a nice gift.