ZCash/ZEC GPU mining

[DrMrLordX]

Optiminer 4.0 is out which provides a small boost to performance.

 

[thilanliyan]

What's with the run-up in ZEC value?
And the dip in Eth?

 

[suklee]

8.0 Claymore https://bitcointalk.org/index.php?topic=1670733.0

 

[thilanliyan]

Decent speed bump for 290s...up to 250S/s at 1100/1250 now.

 

[Madpacket]

What's with the run-up in ZEC value?
And the dip in Eth?

No idea about ZEC but I'll take it. Ethereum down due to split bug in Geth, Parity is fine. Should be fixed soon but traders overreact.

https://gitter.im/ethcore/parity?at=5836ff18381827c24d772289

 

[Madpacket]

8.0 Claymore https://bitcointalk.org/index.php?topic=1670733.0

Thanks for the heads up. Decent gains all around. 280-285 Sol/s per Fury, stock clocks.

 

[DrMrLordX]

There's a 5.0 release of Optiminer too, so Linux junkies need to get on that. Still not as fast as Claymore but it's an improvement. It also has intensity settings.

edit: I have updated the Linux automation guide on page 5 to reflect usage of Optiminer; see the end of the directions.

 

[DrMrLordX]

Anyone noticed improvements in speed when running the latest Claymore miner on a card with 8 GB VRAM? Right now I'm running an 8 GB 390 @ 1090/1400 and (allegedly) getting ~280 H/s doing so. Changing memory speed didn't do much, I just tweaked it upwards for the hell of it since ZEC mining uses so little power. I'm using -i 4 on both cards.

Anyway a 4 GB 290 I swapped into this machine that behaves better than my 800 MHz Jerry's Kid card is currently running 1050/1400 but only putting out ~250 H/s. Very interesting.

Also I've been playing around a lot with the intensity settings under Optiminer. It's peculiar. With Hawaii, -i 6 bombs out instantly. I can get -i 5 running on some cards and -i 4 (default) on others. I'll have to play with intensity settings some more, but results seem to vary from card to card.

 

[DrMrLordX]

Sorry to keep solo posting like I'm talking to myself here, but there's more news. The Optiminer Zcash client is undergoing no future development since someone was (apparently) hacking the binaries to stop dev fee payments or . . . something.

6.0 was just released but there's no indication if that release will be hosted in the future. Some early adopters got the binaries, but now there's nothing there to download.

 

[thilanliyan]

Sorry to keep solo posting like I'm talking to myself here, but there's more news. The Optiminer Zcash client is undergoing no future development since someone was (apparently) hacking the binaries to stop dev fee payments or . . . something.

6.0 was just released but there's no indication if that release will be hosted in the future. Some early adopters got the binaries, but now there's nothing there to download.

Lol, I guess that's why Claymore didn't want to do any Linux version.

 

[DrMrLordX]

Well apparently someone coded a man-in-the-middle attack which redirects dev fee payments from the target mining addresses of Optiminer to . . . whatever address you have in mind.

So he didn't alter the binaries at all as I had suspected:

https://forum.z.cash/t/about-linux-dev-fees/9600

Theoretically someone could use the same attack on Claymore. Regardless the best Linux miner out there is now not available for download from any place that I can identify. Apparently 6.0 wasn't all that stable anyway, though I would have liked to have downloaded it before he took it offline. Linux just took another hit in the Zcash mining world, ugh.

edit: I'll edit the OP and the guide to reflect these changes when I get the chance.

 

[DrMrLordX]

Woop! Someone managed to snag Optiminer 6.0 and hosted it elsewhere:

(link removed; see later post)

Optiminer 6.0 works very well, but with the developer encouraging people NOT to distribute the software, I've removed the link. Sorry Linux miners!

 

[lienad216]

what's this man-in-the-middle attack... hm

 

[lienad216]

Thanks for the heads up. Decent gains all around. 280-285 Sol/s per Fury, stock clocks.

Isnt this a bit low...? I'm getting 300H/s on a slightly overclocked 390X (1100/1500)

 

[[DHT]Osiris]

So this is more a generalized question, and I didn't want to dump yet another mining thread, so putting it in here.

I just got a shiny new gtx 1080, and wanted to get in to what all the cool kids are talking about now with mining, so did some brief research yesterday and got nicehash (running Win10), and set up a coinbase wallet. Linked nicehash to the wallet, set it to run on GPU (with the p0 checkbox) and CPU, and I'm currently just letting it do it's thing. Based on what it's reporting, it should gen about $1.70 a day, which'll probably end up at around $40? $50 a month after electricity costs (it's like $0.09/kwh here), assuming i'm mining 24/7 (which I won't be). Does that sound about right, for a single GPU/CPU NV rig under Windows? Did I do anything right/wrong/stupid? It was a lot easier to get going than I thought... I spent more time trying to figure out how much I pay per kwh from my power company than any other part of this project.

 

[DrMrLordX]

How many Sols/S are you getting with that 1080 running nicehash?

 

[[DHT]Osiris]

No idea what Sols/S are (unless that's specific for equihash?). All the coins it's mining seem to be in MH/s, which vary based on the coin being mined it seems. I think right now it's prefering Lyra2REv2 for mining (probably most profitable), and it was accepting 23.64MH/s when I refreshed the Nicehash page for my stats, but that should be with both vid card + processor. Once I get home, I can get the benchmark stats to get a better idea of what the raw numbers are for each algo/currency I guess.

EDIT: Heh, checked my nicehash again and it's showing 40MH/s now for that Lyra algo. I need to just look at the benchmarks later to get a real impression of the performance.

 

[DrMrLordX]

Okay, H/s is fine too. Sols/S and H/s are roughly equivalent. Whatever pool you're mining through should give you stats on Sols/S though.

40 H/s is not a lot. I think the 1080 is kind of gimped due to the GDDR5X memory, but I could be wrong. Have you tried the SilentArmy miner yet? You might do better with it. Honestly there isn't much support for CUDA/Nvidia in the Zcash scene right now, especially with Claymore going GCN-only.

 

[Accord99]

With the current release of Nicehash, on a GTX 1080 for ZCash I'm getting ~175 sol/s with power setting at 55%, core clock in the low 1400s and memory clock at 5216. From what I've read, the 1070s are faster with their lower latency memory, about 220-240 with tuning.

 

[lienad216]

How does GDDR5X gimp the 1080 though? Is it the latency?

 

[[DHT]Osiris]

So I grabbed the numbers for my benchmarks last night, this is what the algorithms/hash rates were as listed via the standard benchmarker in nicehash. Note that I don't know if any/all these algorithms are well known/make sense to anyone, i just copied/pasted them out. All of this is at default stocks btw, haven't gotten a chance/cared to play with clocks yet.

Algorithm Rates
x13 12.880MH/s
keccak 0.967GH/s
x15 11.350MH/s
nist5 56.790MH/s
neoscrypt 0.311MH/s
Whirlpoolx 0.43GH/s
qubit 24.31MH/s
quark 34.170MH/s
lyra2re 7.47MH/s
lyra2rev2 47.450MH/s
blake256r8 5.578GH/s
blake256r14 3.124GH/s
blake256r8vnl 5.543 GH/s
daggerhashimoto 20.712 MH/s
decred 3.305GH/s
cryptoknight 379.1 H/s
lbry 0.363 GH/s
equihash 148.96 H/s

Nicehash has been generally favoring Lyra2Rev2 so far, with about 64% of mining spent on that, next one is libry @21%, followed by cryptoknight @10% At current mining rate, it's expecting roughly .078 BTC worth of income/month.

With the current release of Nicehash, on a GTX 1080 for ZCash I'm getting ~175 sol/s with power setting at 55%, core clock in the low 1400s and memory clock at 5216. From what I've read, the 1070s are faster with their lower latency memory, about 220-240 with tuning.

Are you somehow specifying Nicehash to mine Zcash? Is this something I should do/care about? I just sort of 'set it and forget it' and it's mining whatever it auto-negotiates for, or whatever. Sorry, very new to this.

40 H/s is not a lot. I think the 1080 is kind of gimped due to the GDDR5X memory, but I could be wrong. Have you tried the SilentArmy miner yet? You might do better with it. Honestly there isn't much support for CUDA/Nvidia in the Zcash scene right now, especially with Claymore going GCN-only.

Yeah, noticed that support for nvidia is pretty slim. This one seemed like the easiest and most idiot proof I came across in an afternoon. No idea what silent army is, how is it different from Nicehash? Also, I'm noticing there's vastly different rates for different algorithms, I'm assuming some algos just process faster, therefore need more mh/s to be profitable? The aforementioned Lyra2rev2 seems to be getting accepted by nicehash at around the same rate that I benchmarked (45ish MH/s vs 47ish MH/s bench) so that's good I guess. 45MH/s doesn't seem like a large number but the profitability seems appropriate, Nicehash is currently quoting a smidge under $60/mo (.078BTC) which doesn't sound terrible considering this is a non-dedicated rig with a single card and what I think is a very simple setup. Do AMD cards scale that much higher in similar rigs? Quicker ROI notwithstanding, I wouldn't expect them to be pulling massive numbers or anything.

 

[Madpacket]

what's this man-in-the-middle attack... hm

MiTM attacks (I don't know the specifics of this particular one) are pretty simple but a little hard to explain.

This site explains it better than I can. https://www.owasp.org/index.php/Man-in-the-middle_attack

I'm sure people are doing similar things to Claymore miner. Unless Claymore starts using TLS and file level encryption there's no easy way to prevent these types of attacks.

 

[DrMrLordX]

Yeah, noticed that support for nvidia is pretty slim. This one seemed like the easiest and most idiot proof I came across in an afternoon. No idea what silent army is, how is it different from Nicehash?

Nicehash, the way you're running it, basically rents out your computing resources to the people who provide the Nicehash client. Nicehash uses your hardware to do the work, and they give you a cut of the mining proceeds. Or at least that's my understanding of it. It is possible to get a version of the Nicehash client that mines only Zcash though I think there is still a dev fee to use it.

SilentArmy v5 is a dedicated Zcash mining client that mines only Zcash, and you keep all the profits. It has fairly good CUDA support.

MiTM attacks (I don't know the specifics of this particular one) are pretty simple but a little hard to explain.

I'm sure people are doing similar things to Claymore miner. Unless Claymore starts using TLS and file level encryption there's no easy way to prevent these types of attacks.

It's my understanding that the attack involves running a separate piece of software on the mining computer to monitor the network behavior of the Optiminer client. It knows which t-addresses are used by the Optiminer dev for dev fee mining, so it can intercept outbound packets from Optiminer, spoof replies to the mining client so it thinks everything went okay, and then reshape the packets to use a different target mining address for the dev fee mining.

Either that or its software running on a separate hardware device.

 

[Accord99]

Are you somehow specifying Nicehash to mine Zcash? Is this something I should do/care about? I just sort of 'set it and forget it' and it's mining whatever it auto-negotiates for, or whatever. Sorry, very new to this.

Yeah, in the settings I turned everything off except Zcash. I didn't want to go through the trouble of trying out different core and memory clock settings for the various different algorithms.

 

[Madpacket]

It's my understanding that the attack involves running a separate piece of software on the mining computer to monitor the network behavior of the Optiminer client. It knows which t-addresses are used by the Optiminer dev for dev fee mining, so it can intercept outbound packets from Optiminer, spoof replies to the mining client so it thinks everything went okay, and then reshape the packets to use a different target mining address for the dev fee mining.

Either that or its software running on a separate hardware device.

A small program makes sense. A simple pcap proxy or script that waits for the expected "T" address, replace it with your own T address and then respond to miner with expected "ack" packet.

If the devs used proper TLS (Transport Layer Security) they could probably avoid this attack altogether. They would only have to switch to TLS for the fee mining and then switch back to non TLS for everything else to reduce overhead.