|
Is there a trick to getting claymore's ETH miner to behave properly on nicehash? I've followed the example in the readme for ETH only mining on nicehash, but all my rigs eventually all crap out randomly when (I'm assuming) nicehash switches between different coins that require different DAGs.
What I mean by crap out, is every cards goes bad at the same time and stops mining, but without actually crashing. I can just restart claymore and be on my way again.
I've always just mined on straight ETH pools and never had an issue.
|
|
|
|
|
Is there a trick to getting this miner to behave properly on nicehash? I've followed the example in the readme for ETH only mining on nicehash, but all my rigs eventually all crap out randomly when (I'm assuming) nicehash switches between different coins that require different DAGs.
What I mean by crap out, is every cards goes bad at the same time and stops mining, but without actually crashing. I can just restart claymore and be on my way again.
I've always just mined on straight ETH pools and never had an issue.
|
|
|
|
Going to go ahead and assume Newegg Canada doesn't have this deal.  |
|
|
|
|
Just use drivers before 17.x.x or use the one button patch that enables 17.x.x drivers to work with modded RX bioses. Don't see what the issue is.
|
|
|
|
Personally, I remove the shroud leaving only the heatsink on the GPU. Than I slap a nice powerful 120mm PWM Delta (or equivalent) onto it (making sure it covers the VRMs) and then I solder the 4 fan wires to the back of the GPU's fan header. That allows the fan to be powered and controlled by the GPU its self, no headaches. GPUs run extremely cool, and you'll never have to worry about fan failures again. Also allows you to keep the original fans nice and fresh for resale.
If you're not big into soldering, I'm sure adapters exist that would allow plugging directly into the GPU's header.
I think I have pics of the process somewhere. I'll post them later if I find them.
Is this the adaptor? https://www.amazon.com/Gelid-CA-PWM-02-Adapter-Sleeved-Cooler/dp/B005ZKZEQAYep, that'll do that trick. Expensive though. |
|
|
|
|
Personally, I remove the shroud leaving only the heatsink on the GPU. Than I slap a nice powerful 120mm PWM Delta (or equivalent) onto it (making sure it covers the VRMs) and then I solder the 4 fan wires to the back of the GPU's fan header. That allows the fan to be powered and controlled by the GPU its self, no headaches. GPUs run extremely cool, and you'll never have to worry about fan failures again. Also allows you to keep the original fans nice and fresh for resale.
If you're not big into soldering, I'm sure adapters exist that would allow plugging directly into the GPU's header.
I think I have pics of the process somewhere. I'll post them later if I find them.
|
|
|
|
|
Decided to try the non US node instead of switching and so far so good. Issue seems to be with US node.
|
|
|
|
I should have a pretty interesting announcement on this front in the next few weeks ...if I were you guys don't buy anymore of these cheap switches (they suck and are not even designed properly), or even GPU risers. I will have something that will change the way we GPU mine  Same, i'm interested. maybe it's the PCIE thought USB people have discussed for a while...... I have a few boards with one 16x and one 1x slot that I plan to make into 4-5 card rigs... which would make it cheaper in the end to make the miner; because those mobo+CPU+memory are typically coming from discarded pc's and are free to me usually =) Even if I buy the $40 1-3 adaptor; it's still cheaper than sourcing a newer board, etc. If he managed to get GPUs working on USB protocols that would be quite something. If anyone can do it, it would be him. |
|
|
|
I should have a pretty interesting announcement on this front in the next few weeks ...if I were you guys don't buy anymore of these cheap switches (they suck and are not even designed properly), or even GPU risers. I will have something that will change the way we GPU mine Awesome. Can you share any more details? |
|
|
|
Well I have a Asrock H97 Anniversary with 3 Hd7990 and 3 390x. It boots fine and all 9 shows up under window 10. One of die hd7990 gpus are buggy so I disabled it. Although I have not tried the pcie switch I am more convinced that it would work on the asrock h97 anniversary. No modifications or windows registry hack required.
Plenty of reports of 3GB 7950/7970 cards working, the issue is 4GB + cards. Your setup would be 30GB of VRAM reported to the motherboard's BIOS. 8x 4GB cards would be 32GB. Than again, I'm running rigs with 7x 8GB GPUs without issue, so 56GB reported. So that's probably not the actual problem. I've also tried running 2x 390 Devil 13 + 4x 290 for a total of 8 GPUs which also didn't work. Tried with both W10 Pro and W10 Enterprise. |
|
|
|
People don't want to reveal the secret to getting more than 7 gpus working. It gives them an edge that they don't want to share. I remember reading that the motherboard's bios needs to be modded for 8+ to work.
For what it's worth, I had a go at 8 and 9 gpus on an MSI Z87. Board would boot woth 8 but only detect 7. With 9, the board just wouldn't boot.
i think win10 works with 8gpu without any mods - just rumor, need to confirm. I think it would work with 8, but the issue is getting the motherboard to work with 8. |
|
|
|
|
People don't want to reveal the secret to getting more than 7 gpus working. It gives them an edge that they don't want to share. I remember reading that the motherboard's bios needs to be modded for 8+ to work.
For what it's worth, I had a go at 8 and 9 gpus on an MSI Z87. Board would boot woth 8 but only detect 7. With 9, the board just wouldn't boot.
|
|
|
|
Fantastic thread, going to start reading through it. I wonder if I should use my sheds to mine.
I would need to trench a line to do a 30 amp pdu.
My gpus pull just about 24 amps.
I could lay two ethernet cables and two ten gauge 30 amp lines.
Most likely I would never use two pdus just one.
Just a thought .
Why 2 ethernets? I'd run a single 60A 6AWG cable instead, would surely be cheaper than 2x 10AWG. |
|
|
|
What clocks? Got 930H/s on 1140@1625 on 1125 strap.
Wonder what kind of speeds wolf0 could get out of these. 1240/1625 1125 strap hynix:  Damn, that's impressive. I'm guessing hawaii BIOS needs to be edited in HEX to change the straps? I was told by Eliovp that changing straps on hawaii doesn't help. Clearly he's wrong! |
|
|
|
regarding this 100% fan issue....
well... I don't want to be overly optimistic but it does look like 16.11.4 fixes the problem! (DDU and manual drv install) Thanks @Prelude!
Really weird, cause I'm 100% sure I tested 16.11.5 and the bug was there.
Anyway running since yesterday GG fork on several machines and it's all good.
EDIT: now when I'm finally able to run xmr on SGminer I'm sure Claymore would soon update his miner...
Awesome, glad to hear it. Both my RX rigs are working perfectly with 16.11.4/5 |
|
|
|
|
It's still happening. At this point I've burned a whole lot of kilowatts on rejected shares and lost a bunch of revenue so I'm going to have to change pools.
That said, your pools are my favourite. If you figure out what is wrong with the pool eventually, let me know and I'll switch back over to suprnova.
|
|
|
|
i've reloaded the US node, looks like it was a bit overloaded, should be fine now
Cool. I'll zero statistics across all rigs and keep an eye on them. |
|
|
|
Rejects are still happening:   Any ideas? |
|
|
|
No, that's not the issue. It's a bug with sgminer and some driver versions. I say the bug is with sgminer because claymore doesn't have issues with the fans.
I'm over 24h without fan issues now, problem solved. Use 16.11.4 or 16.11.5. Hopefully the sgminer bug can get sorted, I'd like to use the latest drivers on my gaming rig.
Nope, installed 16.11.4 this morning, and it was pretty stable for the whole day, but just a few minutes ago got 100% fan with one of my rx 480/4 nitro cards.  There might be some correlation between the driver version and how often this 100% fan thing happens, but it's not just the driver. I still think it's card(s) being not 100% stable. I'm running HWiNFO now in the background, looking for "GPU memory errors". Had to actually drop my memclocks quite significantly before those errors stopped occurring. I thought I was 100% stable at 2150 memclock (1750 strap timings, Samsung chips) for most of the cards, had absolutely zero issues with other miners, but I never checked that counter in HWiNFO before. This sgminer-gm problem made me look for solutions and I've just discovered this HWiNFO thing only today, so not sure yet how it works and whether it's normal to have some errors or there shouldn't be any at all. Weird, mine has been perfect since changing. Maybe the Nitro is different than the reference card in some regard. Consider using claymore until this gets sorted. You'll be a bit slower, but avoiding downtime will more than make up for that. |
|
|
|
|
Show Posts
