Yes, I should clarify that my two D1800 have been working quite well - one filled out with GTX 1060's and the other getting filled out with RX 560s - but a comment made in another thread got me to do some more poking around and it turns out my RX 560s aren't working quite as well as they could (actually, I already knew that, but just assumed I got unlucky in the silicon lottery).

To make a long story short, MSI AB was claiming it was changing the clocks of my RX 560s to 1200/1900, but GPU-Z was reporting they were still at the stock settings of 1295/1750. Inserting an HDMI dummy plug on the first card caused the desired clock values to show up in GPU-Z, Cryptonight hashrate increased by 11% overall (with most from the card with the dummy plug) and Ethash hashrate increased by 8% on the card with the dummy plug while remaining the same on the other 2.

I bought the one dummy plug for my first RX 570, but it didn't appear to need it so this may be specific to Baffin cards or because of the iGPU on the Celeron or even the Onda BIOS. Dunno, and since the dummy plug is a cheap solution I'm not going to worry too much about it.

Right now I have (3) RX 560 with (1) more on the way. Given the boosts in hashrate of ~8% for Ethash to ~11% for Cryptonight per card with the dummy plug attached each plug will pay for itself in 30-50 days. Seems like a good tradeoff to me.

Half-correct. He started off by saying that the iGPU sometimes interferes with overclocking AMD cards, but he also said later on that sometimes a dummy plug needs to be used to fool the card into thinking a monitor is attached, and that turned out to be the solution for me.

I'm guessing it's a mobo-dependent issue, then, as I'm also using the latest drivers (Adrenalin 18.2.2).

Something else to consider that I just learned about myself is that AMD cards need to have a display attached to operate at maximum hashrate. It doesn't have to be a real display, though - you can get "dummy plugs" that simulate a display for $8US on Amazon: https://smile.amazon.com/Headless-Display-Emulator-Headless-1920x1080-generation/dp/B06XT1Z9TF/ref=sr_1_3?ie=UTF8&qid=1518969028&sr=8-3&keywords=hdmi+dummy+plug

Credit for this tip goes to Mattthev in this thread starting with my post #241: https://bitcointalk.org/index.php?topic=2046072.240

Holy shit, that worked! I just happened to have an HDMI dummy plug that I purchased for my RX 570 rig but didn't end up needing. I plugged it into the first RX 560 on the Onda rig and now the correct clock rates are showing up in GPU-Z while Cryptonight hashrate jumped by ~22 H/s per card (after stopping and restarting xmr-stak)! I'll try Claymore v11.0 on a <2GB DAG Ethash coin (MUSIC or UBQ) in about 15 minutes and update this post.

I have just one merit left to give out, but if solving a problem I didn't even know I have isn't worth a merit then this forum is screwed...

UPDATE - okay, looks like I need to get a dummy plug for *each* card; the hashrate jump in xmr-stak was ~66 H/s total and I *assumed* that it was spread equally per card, but when I loaded up Claymore only the first card shows a higher hashrate of 14 MH/s while the other two are still at 12.9 MH/s. Still, the dummy plugs are cheap and an easy fix!

Depends on where you are located - ahashpool is in Asia somewhere, so ping times will be long from the US - and how much hashrate you have - their 0.01 BTC minimum payout takes way too long to reach with my 6x 1060 rig.

Question for fellow Onda D1800 users: I just found out that AMD cards don't get along with the integrated GPU if the latter is enabled, but I've also read of quite a few horror stories about the external GPUs not outputting video if the iGPU is disabled on the Onda boards. Any advice on how or whether to proceed would be appreciated as my plan was to make one of these D1800 into a dedicated Cryptonight miner and I've already acquired (4) RX 560 for that purpose.

I haven't seen a batch file this complicated in 20+ years... Not since the bad old days of having to tweak himem and qemm to get DESQview to work (in other words, pre-Windows 3.0)!

My first suggestion would be to provide some explanation as to what this batch file does* if for no other reason than so we can know what to expect from its operation, and when to intervene if something seems amiss. I'm not saying write a detailed "click on this button; press this key" type of description - mining is not for the moronic, after all** - but an overview would be nice.

My second suggestion upon cursory reading of the batch file is to ditch lbry - it is no longer practical to mine with a GPU due to the Baikal ASIC miner. Pascal, Sia, & Decred have also been ASIC'ed.


* - the readme.txt file included in the zip is empty, so perhaps there already is an explanation, it just got lost in transit.

** - except for NiceHash; it doesn't take much brains to get it running. No offense to NH users; I occasionally use it myself, too.  

Thanks for the clue on what to check and you are indeed correct - GPU-Z 2.7.0 reports the stock and current clock values are the same (1295/1750, vs. 1200/1900 reported by MSI AB) so something buggy going on for sure, and you get a merit! Disabling the iGPU on the Onda mobos has led to a lot of problems with NO video showing up, and it not being recoverable without removing the CMOS backup battery.

EDIT - so the plan, I guess, is to order a DVI-D to VGA adapter (so I can continue using my KVM switch with my other Onda board that has all NVIDIA cards) and see if I can get the mobo to disable the onboard VGA *and* enable one of the PCIe GPUs for video out.


I'm interested... but I need to sort out this iGPU mess, first.

Well, the PS script crashed just after attempting to load miners. Here are the last few lines from the miner.log:


I did not press Ctrl-C, even though that is what it looks like. I did type in 'exit' when I noticed it was sitting at the PS command prompt.

EDIT - I should also note that ccminer-polytimos was still plugging away on X17, so this wasn't a fatal error.

Same here - yesterday was a total disaster with all sorts of problems plaguing me from the moment I got up until I finally called it quits to watch a movie around 8p and by the time JK told me about his script I was in no mood for it. 

But the MrPlus fork of NemosMiner just crashed after spending another 8 hours stuck on X17 so why not try out some new and exciting ways to be vexed!?!

Haha, very funny. In restrospect this is precisely the behavior I should have expected, yes. But consider the confluence of factors here: I installed a new wallet; it did not appear to automatically recognize my existing keystore; it also prompted me to either import an old account or create a new one, but selecting import appeared to do both, while also zeroing out the old account; there was no message about syncing the blockchain and the old 0.7 wallet did not show your wallet balance until it was done. The last issue was the complicating factor here - something funky happened with ethernet connectivity with the one port on the switch this computer was connected to and I ended up having to cycle power to the switch and the router to fix it. Frankly, I've never seen just one port stop working so learn something new - usually unpleasant - every day, I guess.

Was this on Zpool, too, or another one? I switched over to 1.2.1 last night and in the ensuing 8 hours it still has only mined X17 (according to the entries in miner.log).

This seems exceptionally unusual, though I'll admit I've only been testing out multi-algo pools for about 3 weeks now so I'm inclined to let it slide for now.

Otherwise, the main PS script window is much more Spartan now and I kind of miss the ordered list of algos with their (hilariously wrong) projected earnings; if nothing else seeing that list - especially if values change, if not the order - would give me more confidence that the averaging function is working correctly (or to judge when the pool data is messed up, as has occurred with SHA256 and Lyra2v2 on Zpool recently).

In contrast, I really like the average earnings stats and projected pay date that you've added. It seems like you are taking this in the right direction overall.

Because image embedding often doesn't work here, I made it a link.

https://imgur.com/a/zeuOW


Yeah, I didn't think so, but there was a reason I asked - the forked version of NemosMiner I was trying out was stuck on X17 for 3 hours before I closed it and started up an older version of the same to see what it selected as most profitable; for about 20 minutes it was X17, but now it has switched to Tribus. 


I normally love tinkering with this sort of thing, but I am approaching burnout today as a result of hunting down numerous issues that decided to happen all at once.

Yes, I'm confused because what you describe does not match what I experienced/am experiencing.

In fact, I just reopened the v1.0 wallet and now it is showing two addresses, one of which is my original address and which now has *Zero F'ing Coins*.

So thanks, wallet devs, you just managed to wipe out 120 MUSIC because your old wallet wouldn't back up properly, both wallets store keys in some ridiculous secret Windows folder rather than in the same directory as the wallet, and upon opening the new wallet for the first time it presented a bunch of choices that somehow zeroed out my balance without ever warning me this could happen.

Oh, and I can't manually open up the original wallet's UTC/JSON file file because, once again, an explorer window pops up when I select that option and then notepad automatically opens the file when I double click on it. FFS, this is moronic.

EDIT - I finally got it working, but I don't really know how. I rebooted the computer and after the entire blockchain synced my old balance showed up. Had I known what to expect I wouldn't have gotten so vexed, however.

Okay, so I'm testing this out using the startzpoolplus batch file - with my BTC address, of course, -interval set to 120 and -ActiveMinerGainPct set to 4. Benchmarking on a 6x GTX 1060 system took around 2 hours with all the default algos selected for Zpool except for LBRY (it's been ASIC'ed, so no point mining that with a GPU rig anymore).

2.5 hours after normal auto-switch mining began and the only algo selected has been X17, which seems awfully suspicious. I expected less switching, sure, but not this much less. Where should I look to troubleshoot, or is this entirely the result of the pool reporting bad/stuck data?

Does the magenta highlighting of an algo indicate it is the one that is most profitable at that time?

It took about 2 hours to benchmark all the algos selected by default for Zpool (except for LBRY, as that algo has been ASIC'ed so no point mining it with a GPU) and it has been running in auto-switching mode for about an hour. So far it has remained on X17 the entire time, which is definitely unusual. Neither Nemos nor MrPlus do a very good job of explaining what their mining automation scripts do so I am inferring much from observation, but it appears that the MrPlus version needs to run for many hours to build up a decent average for the profitability stats, hence why I thought it was unusual it was sticking with X17 for so long.

That said, NONE of the multi-algo pools I've tested over the last few weeks - using NemosMiner, in all cases - beat out NiceHash when run concurrently on similar hardware. Two pools got on the wrong side of a hard fork which wiped out ~35% of my earnings so those were clearly outliers, but MiningPoolHub and Zpool both came in around 10% less than NH and so I have begun to suspect NemosMiner is the culprit, especially since I verified that both pools pay fairly/honestly when run in single-algo mode (Neoscrypt on MPH and Lyra2v2 on Zpool).

I'm *hoping* that MrPlus' fork of NemosMiner will help, but if not I will use NiceHash for auto-convert to BTC and Zpool or MiningPoolHub for "single-algo, convert to whatever I want."

That's what I was using for interval, along with 5% for the threshold to switch, but I just downloaded and am running the benchmarks for the MrPlus version that blends together short and long term profitability to reduce chasing brief profitability spikes and have set interval to 120 seconds and threshold to 4%. See the ANN thread here: https://bitcointalk.org/index.php?topic=2965976

I can't say whether it is better or not since it is still benchmarking, but it makes more sense to me than merely polling the API every x seconds and then switching even if that algo is getting pumped by some low-float shitcoin.