@citronick's heat management post on @philipma1957's megathread is really useful so thought I'd share it here too: I gone through the ups and downs of managing temps in the GPU farm.
The lessons I learnt as follows, hope it helps in your environment.
1. Hot air naturally goes up - so your vent/blower out should be on top to naturally move hot air out of the farm/room.
2. Floor standing fans pointed upwards are useful to get airflow moving upwards - get rid of hot air pockets forming at rig level.
3. Static pressure VS non-Static pressure - use them to help you move hot air quicker. Avoid having outake vents/fans without intake vents/fans. Always have IN and OUT airflow so that no hot air are recycled into the farm/room.
4. The fastest rigs, ie superclocked at supersonic speed, will create lots of heat (probably reduce the lifetime of your GPUs too) and your over stressed PSUs will also produce heat - Beware that most super fast rigs are NOT stable over time. My ETH farm in the early days was a monster of a setup 30-31MHs rigs screaming away and power bill was high. However, they do need reboots due to hang etc. I was using Windows then. Currently, my farm is all Linux using smOS. Only a handful W10 for NVIDIA and test rigs. ETH mining is very very stable, cool and steady hashrate at about modest 26-28MHs rigs resulting hashrate to pool's stratum servers consistently without interruption and zero stales. I stopped ETH mining about 2 months ago. Now NH/Cryptonite and later BTG.
5. I have 2 of those air extractors/blowers - I don't use them often but on very hot days I use them and position the orange hose on top the big rigs (12/13/19 GPUs rigs) zone. My big rigs zone is in the middle of the farm. The surrounding perimeter are all the 6/7/8 GPU rigs.
6. All Deltas and server PSUs are hung like bats -- exhaust pointing upwards.
7. The farm has CCTV and I can remotely monitor environmental like temps and decibels via the CCTV.
8. AC is not required - I tried it ... AC is useless. Unless you are using Data Centre grade AC then that's a different story altogether. Its all about INTAKE and OUTAKE (top) and moving HOT AIR UPWARDS to the top as fast as possible.
|
|
|
|
Nice GPU with higher hashrate
Btw, P104-100 what is that? Nvidia Tesla? or what?
I'm sorry i dont know about that
A dedicated mining card based on 1070 but using DDR5X memory and what seems to be a mining optimised BIOS. Has no video outs and 3 month warranty e.g https://www.gigabyte.com/Graphics-Card/GV-NP104D5X-4GIf it's anything like the P106-100 (which is based on 1060) it will probably only be available directly in bulk to those with the right contacts for a while yet. |
|
|
|
Hey guys, been using my 970 to mine using nemos miner no problems. Recently picked up a gtx 1080 and the display driver crashes whenever Nemos switches to a different algorithm. Any solutions to this? Crashes regardless of overclock, underclock, or stock settings.
Try uninstalling any OC apps, doing a DDU then installing the latest nvidia drivers Thanks for the advice, tried exactly that. Got rid of GPU Tweak, used DDU, reinstalled latest Nvidia and still crashing on algorithm switch. Seems like it's only doing it when I'm using OC software though, either GPU Tweak or MSI Afterburner. I'd be fine not using either while mining, but I like to limit my TDP of my card while mining and I can't do that without OC software as far as I know. Are you the latest version e.g. Afterburner 4.4.2? It's not as convenient as AB but you could use nvidia-smi to set TDP from the commandline: http://cryptomining-blog.com/7410-monitor-and-control-your-nvidia-gpus-while-mining/ If it works you can make a bat file to run the settings. Yeah the latest Afterburner. After some more tests it seems like OC software doesn't change anything like I thought, still crash on some algorithms closing, so far: skunk, x17, lbry, lyra2, blake2s all crash when closing. For now I'm just mining one algorithm when I can't babysit the computer. From some research I've heard the same happening with other people using 1080s and other pascal cards crashing when closing certain algorithms. I don't have any 1080s but I have benchmarked and nemos can switch to those algos on all my other pascal cards: 1060s, 1070s and 1080tis. Are you using a pcie riser? They are often a weak point so try switching it out. |
|
|
|
how do I set to use only 50% of the gpu power so that I can still play basic game and hardware accelleration..
thanks
I don't know if this will work for gaming but you can try reducing the intensity of the miners. In the miners folder edit the .ps1 file for each miner and find the -i parameter and lower it. e.g. in ccMinerAlexis78.ps1 change '-i 21' or similar to something like '-i 10', do this all the active algos (lines which aren't commented out with a # at the start) and see how it works. Just guessing a value you'll have to experiment to find one that works. |
|
|
|
All the hot air is probably stuck in top of the room. While the fan is pulling fresh air in from outside it will stay near the floor and not forcing out the hot air trapped above. So you're heating the room slowly from the top down. The rigs mostly being higher up in the the shelving will increase this effect.  You could probably improve things a bit by moving the rigs as low on the shelves as possible. But that will make access harder and they'll get more dust near the floor too. First I'd say try extending the ducting and relocating the fan to extract the stale hot air from the top of the room as quickly as possible:  If that works you could even make a rough 'hood' to increase the exhaust area and catch as much of the air rising directly above the shelving as possible. You could also try changing the location of the fan along the ducting to see where it pulls the best.  If you can add another exterior vent for the exhaust then you could keep the existing intake and might get it working like this:  |
|
|
|
This probably isn't the right place for this but I'm having a hell of a time moving my 4x1070's into one rig.
I set everything up on a fresh install of windows 10 and I can get it to detect 3 cards but when I add the 4th nothing boots I just get a black screen and can't even get to the bios.
It's not the risers because I can make any combination of the three cards work and it's not the PSU's because I, again, can make any combination of the three work. It's not the PCI slots because I can make any combination of the three work as well out of the 6 slots. As soon as the 4th card is added with power to it it just won't start.
I've read that maybe my 4gb of ram isn't enough, but I've also read that you really don't need more than 4gb of ram. I'm really stuck here and hoping that one of you might have had this same issue and know a way to correct it.
Have you changed BIOS settings for mining? e.g. 'BIOS Above 4G Decoding' is enabled, PCIE speed set to 'Gen 1' You are an awesome person. Thanks for this. All good, glad I could help  |
|
|
|
Hey guys, been using my 970 to mine using nemos miner no problems. Recently picked up a gtx 1080 and the display driver crashes whenever Nemos switches to a different algorithm. Any solutions to this? Crashes regardless of overclock, underclock, or stock settings.
Try uninstalling any OC apps, doing a DDU then installing the latest nvidia drivers Thanks for the advice, tried exactly that. Got rid of GPU Tweak, used DDU, reinstalled latest Nvidia and still crashing on algorithm switch. Seems like it's only doing it when I'm using OC software though, either GPU Tweak or MSI Afterburner. I'd be fine not using either while mining, but I like to limit my TDP of my card while mining and I can't do that without OC software as far as I know. Are you the latest version e.g. Afterburner 4.4.2? It's not as convenient as AB but you could use nvidia-smi to set TDP from the commandline: http://cryptomining-blog.com/7410-monitor-and-control-your-nvidia-gpus-while-mining/ If it works you can make a bat file to run the settings. |
|
|
|
This probably isn't the right place for this but I'm having a hell of a time moving my 4x1070's into one rig.
I set everything up on a fresh install of windows 10 and I can get it to detect 3 cards but when I add the 4th nothing boots I just get a black screen and can't even get to the bios.
It's not the risers because I can make any combination of the three cards work and it's not the PSU's because I, again, can make any combination of the three work. It's not the PCI slots because I can make any combination of the three work as well out of the 6 slots. As soon as the 4th card is added with power to it it just won't start.
I've read that maybe my 4gb of ram isn't enough, but I've also read that you really don't need more than 4gb of ram. I'm really stuck here and hoping that one of you might have had this same issue and know a way to correct it.
Have you changed BIOS settings for mining? e.g. 'BIOS Above 4G Decoding' is enabled, PCIE speed set to 'Gen 1' |
|
|
|
 And how are you guys doin?  Wha? So much going on. First up you're comparing different pools? And then that X17 amount… my fastest rig is 105Mh/s and it's showing 57.663USD/day on ahashpool24hr… please explain! |
|
|
|
Guys, total noob to mining but I've been linked here by Hardforum. Will be mining with a single Titan V.
Sorry for the noob questions:
1. Downloaded NemosMiner-v2.4.zip and followed the instructions in original post. Says: Running Failed #. Any thoughts?
2. Should I put my Titan V back to stock clocks or even under-clock? Any prevention measures?
Send me one and I'll help debug. (Sorry not constructive but I couldn't resist  ) Definitely try using stock clocks initially. Does a miner window flash up quickly then disappear? Is there anything in miner.log in the logs folder? |
|
|
|
Hey guys, been using my 970 to mine using nemos miner no problems. Recently picked up a gtx 1080 and the display driver crashes whenever Nemos switches to a different algorithm. Any solutions to this? Crashes regardless of overclock, underclock, or stock settings.
Try uninstalling any OC apps, doing a DDU then installing the latest nvidia drivers |
|
|
|
NemosMiner-v2.4zpool hashrefinery ahashpool miningpoolhub nicehash https://github.com/nemosminer/NemosMiner-v2.4/releases/download/v2.4/NemosMiner-v2.4.zipAdded: XMRig (cryptonight) mBTC/perday along side BTC/perday -workername (nicehash) Lyra2z --submit-stale (miningpoolhub) updated: ccminers (Tpruvot & KlausT) Increased Benchmark time 8min per Algo (takes awhile but very Accurate) Added: -ActiveMinerGainPct (percent of advantage that active miner has over candidates in term of profit (default 3%) this will help reduce unnecessary switching bug fix: for 8 card Rigs (not closing miner) Tested with 6x10603gb & 6x1080ti Awesome thanks Nemo! Installing now |
|
|
|
The problem continues, does not leave the BENCHMARCHING I hope to answer because I am writing the second topic with the same problem, there are more people with the same problem and Nemos does not respond or even gives importance attached image with problems. https://prnt.sc/i1krwyAlso for me, i have a problem with Palgin benchmark ! Everyone with problems and no solution, complicated because it put new version but did not advance anything. its not complicated there is no new version... im working on a new version and have let you test it... the latest stable release is still : NemosMiner-v2.3c-beta : https://github.com/nemosminer/NemosMiner-v2.3/releases/download/NemosMiner-v2.3c-beta/NemosMiner-v2.3c-beta.zipBest Regards Nemo Even with the version that you passed continues same problem, they are 3 GTX 1080 TI, I already did everything and I am also developer ... but until now it continues on this screen. https://prnt.sc/i1nap8ahashpool does not support equihash disable equihash and the rest of the algos will bench |
|
|
|
I am having problem with crashes.
Every 6/8 hours the system crash. I have just one gtx 1060. I use dual claymore.
I lowered the overclock but it didn't help.
In windows with the same overclock settings I never have problems.
How can I find out what's the problem?
Thanks!
I have a similiar problem. I use 8x gtx 1060 cards and when I run ETH miner (solo) it seems stable. But when I run claymore dual (with siacoin) it keeps on crashing that even reboots doesn't help. I need to fisicaly shut down the rig and put usb out/in. My friend runs an identical (literally) rig with also hive OS and it works for him running claymore dual. What can be the problem? Electricity plug? USB maybe? It is driving me insane. Dual mining will draw more current so might be taxing your PSU. 1060s aren't very strong at ETH let alone dual ETH+SIA, have you tested out equihash algos? |
|
|
|
Hi to all
I decided to expose my prob here after so many searches and unsuccessful attempts...
To start here my config :
-Win 7 pro X64
-i7 3930k with asus P9X79
-32 Go DDR3
-128 GB SSD
-2x Titan X (pascal) 12Go DDR5x
I tried to mine eth with Claymore and ethminer genoil and i had always the same result ; only 6.5 mhs for each cards... I know i can have really better hashrate but it seems impossible !
I tried so many things found on every forum but without results...
I tried other coins:
- Zcash (with zec miner) -> 400 sol/s for each
- SiaCoin (with marlin) -> 1500 mh/s for each
I don't know if those hashrate are normal or not, if my problem concerns only eth or all coins
Please help me !!!
(sry for my english, not my native language..)
Try Win 10 with the latest nvidia drivers |
|
|
|
All my Colorful mobos have Colorful markings with serial number
is this the one with K17 (without CPU) and K35 variant (with CPU) ?
Colorful is totally different from K17 Colorful have 8x 6 pin at the GPU side , 3 fan headers , single ram slot , 8 pin CPU in with special out cable from board , Sata power out on board with special cable. K17 have 6x 6 pin at the CPU from , 6 fan headers , no CPU in , or out or sata out . Also colorful board have larger GPU spacing My K17 died and killed 8 RX570 GPUs + Msata HD after running 8 GPUs for few minuets , on the other hand the Colorful been running for like two weeks now . My colorful doenst have any colorful markings and no serial number . Crap. I have had a K17 based rig running for a week and have a K35 on the bench to be built but now I'm wondering if it's a good idea to use them at all. I'm running 1060s at 65% PL which shouldn't be stressing them too much… Do you know what failed on the mb? |
|
|
|
Hi
I've just joined the forums as I have just got a Gigabyte 1060 windforce to add to my MSI gtx 970 and I'm not exactly impressed. I'm using my PC as I've just started mining in the last few weeks. I'm just testing at the moment, hashrates seem ok but it's the temps and more worryingly I think the voltages that I'm not happy with. If i'm mining anything on equihash it goes up to about 70 degrees and the fan noise is pretty loud, I have a custom fan curve in afterburner 70% at that temp. I've just tried mining electroneum as a test which my gtx 970 doesn't go above 40. The 1060 is at 56 but worringly it's at it's performance limit for power, thermal and reliability voltage according to hwinfo64. GPU-Z seems to be saying the same thing, it says under perfcap reason m vRel Vop. The 970 just says PWR so obviously it's having no issues. The power consumption on the 1060 while this is happening is about 68%. Sorry I was trying to add a screenshot but it's not workng. And this is with -100 on the core and no overclock on the memory, power limit 85 and the voltage not touched!
Maybe it just needs a lot of configuring but hoping you guys can help. Is this normal for a 1060 to do this? I have got one with Hynix memory unfortunately By the way I'm using Awesome Miner.
Thanks
Hynix is a performance constraint from the outset. If you are using a case you'll always run into cooling issues. Ideally you want them in the open air with a constant fan speed. That said 70˚ shouldn't hurt them. Are you running equihash with -100 core? It's a core dependent algo (a little ram boost helps too) so you want to boost that. I don't know exactly about vRel Vop but I'd guess it's because that underclock is starving the GPU of voltage it wants! My 1060s run best at 65% PL, +110 core, +250 mem. They're in the open air and sit around 60˚ (20˚ ambient) with 60% fan. |
|
|
|
Thanks for the replies guys and thanks for the info about the ONDA board, but the shipping and VAT most likely kills that idea unless she writes a low value for the customs. And I already went ahead and bought four pci-e risers. I was thinking about getting this mb, which lets me to expand up to 6 GPU's if needed. ASUS PRIME H270-PLUShttps://www.asus.com/Motherboards/PRIME-H270-PLUS/I have five of those boards and they work very well and are good value (at least in my location). You can expand it to 8 GPU using two M.2 to PCIE adapters ( these ones are cool because they connect directly to the riser usb) That said I recently setup my first riserless rig running linux off a usb stick and if you're expanding it might be worth the shipping + vat once you factor in buying a CPU, SSD, risers and time to set it up. |
|
|
|
|
Show Posts
