Hey everyone... I am planning a re-build of my first mining rig. Currently, it's 4 MSI 470's on an ASRock x99 mobo with an i7-6800k & 16GB of RAM. Obviously, those higher end components are not not being optimized by my mining opereations, so I am going to use that core Mobo/CPU/RAM to build a new personal/music production PC.
So, i have a Biostar TB85/G3258 CPU/8GB RAM + 1200W Plat PSU ready to go and 2 more MSI 470s arriving tomorrow for this re-build.
But, here's the issue... this rig with 4 GPUs is not able to dual mine ETH + DCR. Since it runs ETH only and XMR flawlessly and there is no power headroom issue (1KW PSU), it seems likely that there is a "problem" GPU that can't handle the extra stress of dual mining. I have a good idea which one is likely as it has never been able to do more than 1850 clock speed, while the other 3 can easily do 1950. What I would LIKE to do is identify for sure which one (hopefully only 1) is causing the problem and move that to my new personal/music PC since I will only be single mining ETH or XMR on that overnight or when I'm not using it for music production.
So... the question is... how best to prove which one is the "bad" one. I think the best way is to do a negative test, meaning running 3 of the 4 cards only and seeing if it can dual mine without crashing. But, I am not sure how to do this, other than literally unplugging the GPU, which I would prefer not to do if possible.
I thought I remembered that Claymore's miners allow to turn off any single (or multiple) GPUs but i looked through the instructions and didn't see this mentioned anywhere. I could have just missed it but wondering if anyone here knows how to do that? If so, is it something done within the .bat file?
yes in the .bat file let me look for it.
this post:
https://bitcointalksearch.org/topic/m.14501372-di GPU indexes, default is all available GPUs. For example, if you have four GPUs "-di 02" will enable only first and third GPUs (#0 and #2).
Use "-di detect" value to detect correct GPU order for temperatures management (requires non-zero "-tt" option); note that it will not work properly if you do not want to assign all GPUs to miner (add "-gmap" option to fix it).
You can also turn on/off cards in runtime with "0"..."9" keys and check current statistics with "s" key.so first do -di 0123
should crash
then do -di detect
figure order out.
then pull away a gpu at a time. to figure the bad one.
-di 012 no crash means 3 is bad
Hey Phil.. thanks for this! I was able to identify the "bad" card... luckily by just de-selecting it at start-up in run-time... so i didnt have to mess with the bat file at all. Sure enough, when i turned off my suspected bad card, that rig ran fine on ETH+DCR for quite some time, so that that was indeed the bad one. I have it set aside now to use in the build if my new personal/music PC and got the new Mobo/CPU/PSU/RAM up and running now with 6 MSI 470s. Still having a bit of trouble getting it to do ETH+DCR but it's been running ETH only solidly now for most of the day.
So, that now puts me at 24 GPUs with 24 pushing 640 mhz ETH, plus getting about 12 Ghz DCR with 18 cards, hoping to push that to 15 once i get things figured out with that re-built rig. I've got 1 more new 470 coming tomorrow, so will have 26 total 470s in the end. I will either mine ETH solo or XMR with those most of the time as they will be in a full tower chasis and I don't want to push them too hard and keep the heat down.
I know i've said it before, but this will really be the cap for me. I had a good test run heat wise last week and my A/C couldn't keep it comfortable with 4 rigs running... not even close! I need to get someone out to do maintenance on it and assess if it can be fixed or if I need to go with a bigger unit. Hopefully that wont be the case.
