I've also seen lower power consumption (but still 100% GPU utilization) when mining XMR.  Right now, nvidia-smi says my three 1070s are burning 250W, and KTccminer-cryptonight is reporting about 1.98 kH/s.  I normally have power consumption throttled to 325W (2 at 115W, 1 at 95W), and other algos bump up against that limit.

I'm not sure if it's optimal for the purpose, but I've been running this one for a few weeks now:

https://github.com/KlausT/ccminer-cryptonight

Building from source is fairly straightforward:

One more observation...have a look here:

https://blockchain.info/stats

As of now, they're reporting 98 blocks found in the last 24 hours.  The normal pace would be somewhere around 144...thought I'd heard some pools were shifting mining power off of Bitcoin, so that would account for that.

More interesting than that, though, is the transaction-fee total: about BTC421.97.  Divide that evenly across 98 blocks and you get about BTC4.31 per block.  That's about a third of the value of the current block reward.  At some point, transaction fees are supposed to surpass the block reward, and for that to happen, they'll first have to approach the block reward...but did anyone think it might happen this soon?

I've had transactions go through in 1-5 hours with 25-50 satoshis/byte.  Earlier today, I swept the last $3 worth out of Bitcoin Core onto my Trezor, and set the fee at 5 satoshis/byte:

https://blockchain.info/tx/fefae7ed63cb5171f11ba32c82d1d68f313b8e6da34578eb478be145c82bde6a

Let's see if this goes through, and if it does, how long it takes.  One transaction a while back took a day and a half to confirm...don't recall how low I went with the fee, though.

I'm not, but for a data point that's probably not on there, I'm seeing about 28 Mkeys/s from a Radeon RX460.  (Host system is Gentoo Linux with the AMDGPU driver and AMDGPU-Pro OpenCL layer.  I had to tweak calc_addrs.cl as described in https://github.com/samr7/vanitygen/issues/38#issuecomment-31248556 to get it to work.)

That'd definitely be advisable for the bugfixes, but since newer versions of OpenSSL don't install certificates, there should be another package that restores the current certificates.  On Gentoo, that would be app-misc/ca-certificates, which was last revised in November 2016.  I'm not sure what Ubuntu calls its equivalent package, as I'm not as familiar with it (the only Ubuntu box I have is my nvOC rig).

That's an ancient version of OpenSSL.  If you have it installed on your rig (as opposed to just extracting a header file from the source tarball), it may have overwritten your root certificates with older, possibly no-longer-valid certificates.  That could cause the problem you're seeing with the switcher querying Coinbase.

Newer versions (such as 1.0.2k, which is what my Gentoo boxes run) don't include certificates; you obtain and install them separately.

You don't choose which currencies to mine; the whole point of the switcher is that it automatically selects whatever's most profitable from among the choices the pool offers.

...and now it's up:

https://alfter.us/wp/wp-content/uploads/2017/07/nvOC_v0017.zip

I tested this direct link; it should work.  If it doesn't, follow the link from here:

https://alfter.us/wp/nvoc_v0017/

...and this one blocks adblockers. 

I'm uploading a copy I'd originally downloaded from Google Drive (or was it Mega?) to my VPS.  I'll post a link once it's up.  SHA256 hash on the file is d11929368d186acf73651be1b9d90234cfe6ef8a097eff56668dee4cbbe69e8c, as announced here.

I thought I read somewhere that running the fans much past 85% won't do much in the way of additional cooling, but it will run up additional wear on the motors.  

My somewhat limited experience is that it only squeezes out a few more cents, not dollars.  I'd also need to redo miner benchmarks, as I've been using Genoil's ethminer.

Commit some Bitcoin to these and you can trade them offline:

https://gitlab.com/salfter/btcnotes

* You can scan the address any time to validate the value within. 
* Don't accept if the private key is visible.
* To spend it offline, just pass it along without opening up the private key.  It handles like cash.
* To spend it online, reveal the private key and sweep it into your wallet.

FWIW, I was sometimes seeing similar errors when I had my 1070s connected through risers.  How's your rig set up?

In other news, the risers I ordered nearly a month ago finally arrived earlier this week.  After fixing an unrelated power-supply issue, I rearranged my rig to use the risers, powered it up...and saw only about half of the hashrate I had previously been getting, with nvidia-smi indicating per-card power consumption fluctuating all over the place.  BIOS settings are as recommended.  I've updated the BIOS and redid the settings.  I've tried slowing down the bus.  I've tried plugging the risers into 1x slots instead of 16x.  PCIe spread spectrum is still enabled; would disabling it likely help or hurt?  Beyond that, the only thing I can think to try is to install Windows and see if it'll work any better, but then I'd have to port my mining switchers to it (if that'e even possible...are the command-line overclocking tools used by nvOC supported on Windows?).  I got so fed up with it that I just put everything aside late yesterday evening and have been mining sweet bugger-all since.

In other, other news, as an alternative to flaky USB PCIe risers that never seem to work right, I stumbled across this:

http://shop.dmp.com.tw/INT/products/23

Depending on how the module's configured, it might need a jumper removed to change the PCIe 1x connector at the end from a target to a host, but once that's done, I'm thinking a simple adapter board might be possible that would hold a GPU in one slot, this card in another, and some power-supply circuitry as appropriate.  Add a MicroSD card and an Ethernet jack, and it'd potentially turn a GPU into a standalone miner.  The module's also available by itself for inclusion in your own designs, which might be more appropriate in the long run...but just to see if the idea would work?  If 128MB RAM isn't likely to be enough, the 1GB version is $15 more.  It should run Linux without issue, and as long as nVidia and AMD avoid certain unsupported instructions (someone said CMOV isn't supported), their drivers ought to work on it.  It's not particularly speedy, but it ought to be fast enough to keep a GPU fed with mining data.

The NiceHash and MiningPoolHub switchers base their decisions on information provided by the respective pools.  MiningPoolHub doesn't support LBRY, so they provide no information on its profitability compared to the coins that they do support.

I also wanted to support pools that will exchange mined altcoins for Bitcoin automatically.  I wrote another switcher previously (https://gitlab.com/salfter/CryptoSwitcher) that worked with any pool, though I don't think I had exchanges fully automated (and Cryptsy and BTC-e have both fallen by the wayside) and I don't recall how well it would've handled multi-algorithm mining as I was using SHA256 and Scrypt ASIC miners at the time (this was back when an Antminer S1 was useful as more than just a space heater ).  Also, CryptoSwitcher used full-node coin daemons (bitcoind, litecoind, etc.) as an independent source of mining stats (and I might've had pools paying out to local wallets); having a bunch of those running chews up lots of RAM and disk I/O.


The MPH switcher builds miner commands from information provided by their API, including host and port numbers.  It should automatically pick normal ports (such as 20535 for Ethereum).  The only configuration you should need to do is in the first few lines of the config file...things like your username and miner name.  If you benchmark the different algorithms on your cards, you could tweak the speed and power-consumption figures to match your system, though (especially if you're running 1070s) the numbers I put in are probably a good start.


I'd then need to dig into the current exchanges' APIs and figure out how best to automate their usage in the current environment.  I don't want a bunch of different altcoins hanging around.  (...though I did find a non-trivial amount of NeosCoin in my wallet the other day that has shot up in value over the past couple of years or so since it was mined...might need to go trade it in.)

That was one of the pools they had been saying for some time they would shut down on or about the 29th.  You were supposed to quit mining on it and withdraw whatever funds you had in it (couldn't exchange it...I suspect that none of the three exchanges they support had it listed).

Higher difficulty makes it harder for small rigs to find shares.  If it's high enough and you're switching between coins to mine the most profitable, it's possible you might find no shares in 10 minutes, 30 minutes, or more.  Without shares, you don't get paid and you're just wasting electricity.

Lower difficulty, OTOH, means that larger rigs spend more time communicating with the pool and less time mining.

Ideally, the stratum server would adjust each miner's difficulty to something lower for smaller rigs and something higher for larger rigs, so that everybody's finding shares at about the same rate (I've heard one share every 10 seconds as a reasonable target, though you could maybe go as high as a minute).  This also tends to improve the accuracy of whatever statistics the pool supports.

In the same vein as the NiceHash profit switcher I released a while back, I now have another switcher available that works with MiningPoolHub:

https://gitlab.com/salfter/mph_switch

Out of the box, it will mine all the algorithms supported by the miners bundled with nvOC.  Instructions are included to add miners for Sia and CryptoNight.  The contents of the config file are a bit different than for the NiceHash switcher, so there's no sharing the two.  Integration with 1bash, if desired, would likely be similar at this point to integrating the NiceHash switcher in its current form.

As for NiceHash vs. MiningPoolHub, that's something I'm figuring out for myself right now. MPH takes 0.9% on all coins mined and imposes a BTC0.0003 miner fee on withdrawals, so by raising the auto-withdraw threshold to BTC0.03 (NiceHash is fixed at BTC0.01), the miner fee ends up being only 1% of the total.  As for which is more productive (read: produces more BTC per week), that's something I hope to get a handle on over the next couple or so weeks.

Is this still the CryptoNight miner for CUDA?  I noticed that it's not been updated in a while, but I just built it under nvOC v17.  With three 1070s underclocked a bit, I'm seeing about 1.35 kH/s.  Is this about in the ballpark of what I should expect?  Is there a newer miner for nVidia GPUs that I should be using instead?

Edit: This thread refers to this fork of ccminer-cryptonight, which is pulling 2 kH/s across three 1070s.