Jump to content
Search In
  • More options...
Find results that contain...
Find results in...
WhiteGoblin

Mining Monero (XMR) with my current gear? Let's look at some numbers!

Recommended Posts

Posted · Original PosterOP

Why does my network suck? Oh hello speed racer mining!

 

So I've got a Nighthawk X10 router, it's a quad 1.7ghz with active cooling, at which I've gone into the debug settings and cranked the fan to max 24/7 as I like my gear running cool. I have gig internet from Comcast, I felt this would be the appropriate non-professional router for a small home network. I was going to hand build a pfSense machine but the X10 was $400 and there was no way I was going to go that cheap on a home built. So yeah it's great for standard use, pushes 950 down all day and night unless it hiccups which it seems to occasionally do. Given it's behind a 1500PFCLCD of it's own I find that interesting though there's only a single build of DD-WRT out there for it and it's currently behind with suggestions from the community to stick on the default firmware for now. ANYWAYS, the entire home network has felt sluggish for a bit, since getting into that speed racer low difficulty work it seems.

 

This is half anecdotal from user reports & half empirical from what I've tested in a very short time period once connecting it to the mining. So my speed racer setup is doing 0.25ms on the CPU load balancer and 1ms on the GPU balancer. This is flooding shit around the local network as you can imagine. Six miners total just going nuts trying to toss around that low difficulty work and get it back to the pool asap. My primary television is used for nothing but online streaming really and it generally pulls about 200 down when it's first starting to cache up 4K. Though when I'm speedracer mining it seems to have trouble getting up past 100. It'l hang out at like 80-90 then somewhat burst. That's pretty disappointing for a flagship model even if it is a consumer level device. Though I don't have alot of options here, I'm not going to build an actual real router for six miners in my pad. 

 

Then it hit me that maybe, I could double dip on some of the processing power I've got here and divide the work load among two cat7 cables instead of one. Keep in mind I wasn't really sure this would have a positive impact at all, it was going to complicate matters across the network that's for sure. What's the haps you ask? Well, what if I turned my modem back from bridged mode into normal mode & used it's quad 1ghz processor too? They only have three modems that are authorized for comcast gig in my area and the one I've got has two eth ports on it. (lol) It's been bridged directly into the Nighthawk X10 and life has been pretty peachy minus the speed racer network spamming. Though what if I took my virtual load balancers and dedicated the second eth port on the modem to them, setup some basic security, & then DMZ'd the Nighthawk back into place on the first port. Would this actually free up my local network a bit? I mean all the clients are still coming in from the Nighthawk and being passed up through the same cable to the modem, down a new cable to the balancers, and then back down that same cable to the modem again. It seems stupid as fuck, though the idea is, that the modem's quad 1.0 ghz cpu will chew up that mining work off the load balancers since they're being routed an handled outside the NightHawk, even though they pass through the NightHawk. This includes their DNS servers, as previously they where behind the Pi-Hole, with failure fall backs of Google & OpenDNS. Though now those load balancers are just going straight to Comcast and more Comcast. Though the rest of the network would still be on the X10 & Pi-Hole, sharing files between each other, streaming, etc.

 

Results? Appear to be a win after two days of testing. The television immediately goes back up to 200ish when it wants to pickup a big chunk of content for caching. File transfers are more responsive, streaming is smoother, etc. I gotta imagine it's simply from placing the load balancers a level up on the different processor. Seems kinda counter-intuitive, I'm happy I tried this but I felt dumb while setting it up wondering what would happen. If nothing else it goes to show if you're going to get into some real equipment producing some real hash rates you're not going to want to neglect your network setup. It's as important as your miners as your backbone to the internet is how you get to and from the jobsite. Nobody going to pay you if you can't even show up on time or stay there for longer then two seconds. Though, as smooth as it appears to make things run, it does actually run the nighthawk harder, by 1% in the CPU utilization numbers presented in the debug screen. The load balancers are reporting a lower latency time as well so they appear to be functioning slightly faster this way. I mean it's six miners running around like they've smoked a bunch of crack, I can't imagine that causes any real stress on the Nighthawk but apparently it actually does. Experiment says it's a win though what I'm taking away from this is I really just gotta build a real router an stop using this consumer level crap.

Link to post
Share on other sites
Posted · Original PosterOP

Let's see those peaks baby:

nXRH1iZ.png


As the provocative image description says, this is my equipment's peak performance. Which matches up nicely with the previous image, though you can't just make your equipment peak for a screenshot & you're not always watching your stats. I was actually just walking by my phone (which I tend to leave a mining stats or market watch screen up on) when I noticed I was peaking. I grabbed this screenshot & I also grabbed another one where the CPU & GPU farms traded .1 KH/s back and forth but still maintained the same peak. We all like seeing our numbers when they're balling large & this is as good as it gets in my world right now. This is 1 KH/s faster then the same equipment averaged on it's own before the hyper-v load balancers and 500 H/s higher peaks then I ever saw before bringing everything together inside XMR-Node-Proxy. This last round of changes are reproducing that one freak time I got this high of a hash rate previously but said I could never get it back. Well, finally I have & these peaks keep returning now. Which makes me believe I've finally started to dial all this in. B|

 

Changes to the setup since the last post have been pretty slim. I installed some better security on the load balancers including putting the entire setup behind VPN. I really dislike ever having stuff coming out of my network that's not encrypted & generally limit it to televisions only & at that for netflix as explaining to users how to switch the TV's VPN address to get it to work is often more trouble then just swapping the TV off VPN while people want to watch stuff. This also gets rid of comcast's DNS servers which I was complaining about before. So now both the NightHawk and the Load Balancers are outputting 99% encrypted data again, atleast until it gets dumped out on the other side of my VPN. Which is actually pretty handy as I can now park my load balancers right next to any pool I want around the world. A feature I was really missing after making these network changes, but even better now because I don't have to route unnecessary non-miner related traffic to where ever I'm wanting to drop off the loadbalancers. Now the NightHawk can be on VPN to one location & the mining traffic off to a completely different location. Pretty happy with this actually, got some security reports getting flung back to my phone as well. Though it's not as good as just having a proper pfSense box setup as a router which clearly I need to build as I've pushed my entire household to the max over this Monero experiment.

 

Anyways just wanted to show off that those peaks where real & all this effort spent putting together / tweaking this operation has paid off. Those are (to me anyways) some monster peaks an everything has the stability I crave. Users come turn on monitors, turn off monitors, watch videos, do whatever, and the system scales up & down with them perfectly. This has been a very fun educational ride down the XMR tunnel. The funny thing is, I'm still a total novice. I sit in that Support XMR chat room and talk to the developers when they're around an they're still over my head, there's just sooooo much to learn out there. I'm sure I'll make another post when the next hash improving something optimizing thing catches my eye but given all the overclocks and crazy shit happening behind the scenes I'm pretty pleased with these results. :)

Link to post
Share on other sites
Posted · Original PosterOP

Let's reinvestigate that whole statement about phones being useless:

 

Oh boi you had to know it was coming right? So... what if we spun up another load balancer, replicated the high pace low difficulty configuration (just scaled real low), and stuck say two phones on it? Would it be better then two phones running on their own? Could it do anything at all of value? How low power can we go and still flip a hash? Well like always I've came prepared with screenshots! :D

 

So yeah I did just that, I lowered the starting difficulty to 5, min to 1, max to 100, then put the shareTargetTime at 1. I'd like to go lower but given these are cell phones that tend to roam around the house as people want to watch youtube here or there I kept them at 1. (still pretty fast) If you turn on the mining without the load balancers they'll do about 1 block a minute which man, that's just moving some data... LOL. Though firing them through the speed racer config they start pushing up to a couple hundred blocks a minute. It also greatly benefits from the other benefits of the balancer, as in, somebody can take a phone & completely close the miner but the work will continue on just half the speed until I can get that device back up again. It would be really neat to get my hands on some other android devices an see what they hash given they're all powered off a 5W usb cable. I'm currently getting about 17 H/s average off them but when they peak I caught two separate screenies of them at 30 H/s. :o

 

One thing to keep in mind is that I'm using this NeoNeonMiner program that I downloaded off the Play Store for free. It's ad-supported & by that they mean full screen ads with audio will pop up disrupting whatever you're doing & it spams them at the bottom of the mining screen as well. I would not call this piece of software "friendly" at all & would gladly pay a percentage cut of mining efforts instead of deal with that type of horseshit. PLUS SIDE, I didn't even know the application did this until I took one of the phones outside the local network. This is because as you can see in the diagrams I've been drawing up, I've got a Pi-Hole sitting on this network. It absolutely roflstomps that entire programs advertising loop & I never once saw a single ad before taking it out of the pi-hole & I don't see any having returned it to the pi-hole. Fucking love that program, if you want to be a good friend / nerd then buy a bunch of raspberry pi's, preload pi-hole on them, and give them out for christmas / birthday gifts. "I give you the gift of ad-free networking across all your devices, enjoy". <-- Pretty much makes you a baller. I did a text & video tutorial to using hyper-v if you're piless & interested.

 

The stupidity of this is of course when you step back and look at it, there's eight cores all operating at or over 1ghz getting chewed up to produce these results. Now the interesting "oh my god goblin's smoked to much hash again" thought. The CSU surplus store I mentioned previously in this thread has piles and I mean piles of old phones for $10 a pop. That's to expensive, there's no money in that, it takes seven-eight years to pay off at that price and today's market rates. But what if I could get like, a bucket of phones for a hundo or something? If they'd throw in a rat's nest of old usb cables I could solder up some ghetto power web, tie them all into the balancer, & see WTF happens. It's not really about the profits at this point, I'm in full nerd mode. I wanna build it, but I don't want to spend the money it's going to take to haggle an entire bucket of phones off them. Mmmmm.. we'll see what happens.

 

One.. Two.. Three Virtual Load Balancer Proxies:

XDQyomb.png


Another Droid Joins The Mining:

MiBU5Nv.png


Watching Two Cell Phones Peak At 30 H/s via Another Cell Phone:

4e9s9dI.png

Link to post
Share on other sites

ok i have a question perhaps you might be able to help me or investigate this yourself?

I have A GTX 970 and a GT 1030  (neither of these cards are ideal, however they are producing in total a nice 600 H/s between the pair of them in using XMR-STAK )

I have undervolted the GTX 970 to 85%  and overclocked the remaining stats (See image) which has pushed my GTX 970 to 445 ish H/s  from 375 H/s average! (ID 0) 

 

My GT 1030 Isn't the best GPU in the world but is pushing 141 H/S - i don't expect much more from this however i was wondering if there is a way to increase the performance in any way as I think I have the Max settings possible)

Image description -
Top left corner - CPU (FX - 8320)  OC 4.612 Ghz
Bottom Left Corner GPU - 1 (ID 0) and 2 (ID 1)

 

Top Right - After burner
Bottom RIght - GPU Config

 

also any tips you have would be great too with this setup of an old repurposed gaming rig... 
Also the CPU isn't performing properly cos i was playing with settings - I normally have that up at about 450 H/S

 

OC Graphics.png

Link to post
Share on other sites
Posted · Original PosterOP

Pi-Hole Mining Enters the ARMs Race:

 

Well I was having a good time late last night hanging with an Arsian when the thought occurred to me that the pi-hole install on the raspberry wasn't utilizing it much at all. Now that I had a load balancer setup for nothing but ARM connections, maybe just maybe it could lose three out of it's four threads to a little mining on the side. Which was actually pretty easily done! But only thanks to shingolavine's raspberrypimining install script which pretty much took the piss out of the whole thing. Afterwards you can watch its temps from inside pi-hole's gui itself to make sure it's in the clear, we also logged in via ssh and watched the top process list while we stress tested the network to make sure a single thread would be enough to handle pi-hole or if I needed to reduce the mining to two threads. (one thread works great even during max stress tests) Temps are great for how much stress is being placed on the device & given it's inside a case. Oh we also overclocked & volted it to the high-OC settings getting it to 1ghz across all four cores stable. What's it push out? Well about 5-6 H/s with three single ghz threads which are of course being piped through the load balancer so it's joining in on the speed racer action as well. B|

 

Here's the little Pi-Hole Miner itself:

 

iythPMw.png


Thank you to shingolavine for raspberrypimining:

 

F0kzCHv.png


Pi-Hole Temp Check / TOP During Stress Test / Seven Hour ARM Balancer:

 

0QbQepP.png

Link to post
Share on other sites
Posted · Original PosterOP
On 8/31/2017 at 12:08 PM, jwacher said:

You should really check out zcash for those nvidia cards. They would do much better there. This is about what a 1080 brings in with those power costs.

Thanks for the advice jwacher! This is a thread about Monero and investigating different ways at mining it on whatever gear I can get my hands on. Maybe in the future I'll do a zcash thread and see what happens in that scene but for now I'm an XMR man as I believe in it.

 

On 9/16/2017 at 6:41 PM, 8uhbbhu8 said:

Wow nice writeup! Been trying to get into it but can never seem to get things working right :( Mostly has to do with failing to compile xmr-stak right :P

Thank you fine sir I really appreciate it! The main problem with compiling stak is that people generally try and do it inside windows. If you're one of those folks then I'd highly recommend the following step by step tutorials: CPU & GPU. That'll walk you through getting into the dev branch of each miner. Though they're about ready to all be re released as one single package that mines across cpus & gpus regardless of brand so you might want to wait for that before going balls deep an eating up alot of your time just to have to redo it.

 

On 9/25/2017 at 6:31 AM, Metal_Kitty said:

ok i have a question perhaps you might be able to help me or investigate this yourself? (cut: whole post about his mining setup)

You look like you're doing great so far! Tips to see improvement in your hash rate though would be to pick your favorite virtualization software & get a copy of Snipa22's XMR-Node-Proxy up and running. This will combine your equipment's power into a single thread which once it's dialed in should give you some pretty decent improvements as far as stabilizing your valley to peak ratio. It's a bit of a process as already investigated in this thread pretty deeply, but if you need a hand just come on back & fire another reply off here!

 

22 hours ago, overturnedv1 said:

how do u have infinite money

Well if only this was the case, I'd really like infinite money. Even if I had infinite money you'd probably still find me just sitting here playing around with Monero. I'm really interested in this entire field. I sit in the dev chat rooms, I ask questions, I experiment and test constantly, and I'm never to scared to look like a dumbass. That's the key I think, if you want something bad enough or are interested in something strong enough you'll find your way into it no matter what. This is the gear I personally have, it's not great, it's not meant for mining, and it's certainly not giving me the best results in the world. Though I'm mining and after a month of optimizations I've gained alot of knowledge on how to be a miner, how to walk the walk, setup rigs, maximize gains, etc. This was just the dipping my toe in the water experiment I needed before getting serious. Though I guess that doesn't explain the question. Fine sir, there appears to be a misunderstanding at hand here, as I do not have infinite monies or cheeseburgers.

Link to post
Share on other sites
Posted · Original PosterOP

Don’t forget the Monero Honey!

lTKA086.jpg


He invested in crypto… He Trusted Monero!

s0YciPn.jpg


These are originally from The Monera.art which I absolutely love! Tons of great Monero themed backgrounds and artwork. I'm going to be framing some of these myself, love the oldschool themed pinup stuff. B|

Link to post
Share on other sites
3 hours ago, WhiteGoblin said:

Don’t forget the Monero Honey!

  Reveal hidden contents

lTKA086.jpg

 


He invested in crypto… He Trusted Monero!

  Reveal hidden contents

s0YciPn.jpg

 


These are originally from The Monera.art which I absolutely love! Tons of great Monero themed backgrounds and artwork. I'm going to be framing some of these myself, love the oldschool themed pinup stuff. B|

Totally agree! Oldschool style pinups are the best xD


Use this guide to fix text problems in your postGo here and here for all your power supply needs

 

New Build Currently Under Construction! See here!!!! -----> 

 

Spoiler

Deathwatch:[CPU I7 4790K @ 4.5GHz][RAM TEAM VULCAN 16 GB 1600][MB ASRock Z97 Anniversary][GPU XFX Radeon RX 480 8GB][STORAGE 250GB SAMSUNG EVO SSD Samsung 2TB HDD 2TB WD External Drive][COOLER Cooler Master Hyper 212 Evo][PSU Cooler Master 650M][Case Thermaltake Core V31]

Spoiler

Cupid:[CPU Core 2 Duo E8600 3.33GHz][RAM 3 GB DDR2][750GB Samsung 2.5" HDD/HDD Seagate 80GB SATA/Samsung 80GB IDE/WD 325GB IDE][MB Acer M1641][CASE Antec][[PSU Altec 425 Watt][GPU Radeon HD 4890 1GB][TP-Link 54MBps Wireless Card]

Spoiler

Carlile: [CPU 2x Pentium 3 1.4GHz][MB ASUS TR-DLS][RAM 2x 512MB DDR ECC Registered][GPU Nvidia TNT2 Pro][PSU Enermax][HDD 1 IDE 160GB, 4 SCSI 70GB][RAID CARD Dell Perc 3]

Spoiler

Zeonnight [CPU AMD Athlon x2 4400][GPU Sapphire Radeon 4650 1GB][RAM 2GB DDR2]

Spoiler

Server [CPU 2x Xeon L5630][PSU Dell Poweredge 850w][HDD 1 SATA 160GB, 3 SAS 146GB][RAID CARD Dell Perc 6i]

Spoiler

Kero [CPU Pentium 1 133Mhz] [GPU Cirrus Logic LCD 1MB Graphics Controller] [Ram 48MB ][HDD 1.4GB Hitachi IDE]

Spoiler

Mining Rig: [CPU Athlon 64 X2 4400+][GPUS 9 RX 560s, 2 RX 570][HDD 160GB something][RAM 8GBs DDR3][PSUs 1 Thermaltake 700w, 2 Delta 900w 120v Server modded]

RAINBOWS!!!

 

 QUOTE ME SO I CAN SEE YOUR REPLYS!!!!

Link to post
Share on other sites
7 hours ago, WhiteGoblin said:

You look like you're doing great so far! Tips to see improvement in your hash rate though would be to pick your favorite virtualization software & get a copy of Snipa22's XMR-Node-Proxy up and running. This will combine your equipment's power into a single thread which once it's dialed in should give you some pretty decent improvements as far as stabilizing your valley to peak ratio. It's a bit of a process as already investigated in this thread pretty deeply, but if you need a hand just come on back & fire another reply off here!

thank you for the tip - i will look at this now! :D 

 

Link to post
Share on other sites
Posted · Original PosterOP

Let's plug in a PowerEdge 2950! Well now that's not what I expected.

 

You ever thought to yourself, damn I bet a 2U with double xeons would be the shit at mining!? Well hold your horses captain because I'm burning 441W of power for an amazing 174 H/s inside Server 2012 R2. No, you read that right. 174 H/s @ 441W. Well this probably happens because while each CPU is rated at 80W it's meant to go into a server rack and has four 9K fans cooling the entire rig minus the fans in the double power supplies it sports. The real downfall of course is the E5450's don't support AES instructions so mining is brutally inefficient. The 56W mini-workstation is out mining these xeons by 100 H/s itself. :|

 

Actually plugging it into a calculator, this thing makes 0.30 cents a day and costs over a dollar a day to run! The only way this thing would ever remain on for this purpose would be if Monero skyrocketed in value before the difficulty curve got anymore monstrous. A four fold increase in value (lolololol) would mean this thing was finally creeping into profitability. So with that in mind, you're either turning this thing off or you're letting it waste money in the hopes that at some point the market will go up enough to justify it's constant no profit expense state now. To put this into perspective, turning old cellphones into a cluster is more efficient & profitable (when scaled) then running this beast server. :D

 

Let's also discuss the, you can hear this throughout the entire house no matter where the fuck you put it factor. These things are LOUD and I mean, there's no escaping it. I took foam studio paneling meant for audio booths and lined a box around this thing just so I could close the door to the room it's in and somewhat hear my television in the living room. I opened the window in there and you can clearly hear this thing outside from the side walk. It must sound like I've got a grow going on in here and have terribly planned my centrifugal exhaust system. It's a bit to high pitch for that to be the truth to an educated ear but for the normal person walking by I'm sure it sounds funny. Oh good times for 174 H/s. ;)

 

Meet a PowerEdge 2950:

 

wC8ZaLS.png


Let's get inside this fine lady:

 

w0yc1Ra.png


Dual Xeon E5450's Inside Server 2012 R2:

 

yua9Ymb.png

Link to post
Share on other sites
Posted · Original PosterOP

PowerEdge 2950 Update: As a continued investigation into the last post I wanted to come back and let everybody know that I managed to get 180.6 H/s out of this beast by running twelve normal mining threads across the eight cores. This is done by turning affinity off & letting your operating system manage the nightmare workload across it's cores instead of assigning specific threads. This is actually a very cool concept we're getting to see in practice here. So previously I ran four cores in low power mode to double it's cache usage from 2MB to 4MB using 16MB total.

 

Then I ran another four cores in normal mode, using 2MB each, so 8MB total. Coming together to fully utilize the available L2. Though the config file notes there will be a performance loss doing this verse just running normal mining threads. So in order to use normal ones at full efficiency we'd have to run twelve of them, which I've now tested. As crazy as it sounds to make twelve miners split eight cpu cores it utilizes the system cache more effectively resulting in a 5-6 H/s gain. BIG NUMBERS right? In perspective that's basically an entire raspberry pi being added to the ARM load balancer. Though I think this little experiment was more worth the educational value for sure. B|

Link to post
Share on other sites
Posted · Original PosterOP
On 9/25/2017 at 6:31 AM, Metal_Kitty said:

ok i have a question perhaps you might be able to help me or investigate this yourself? (cut gear part of post)

Hey Metal Kitty! I was thinking about you this morning and the way for you to achieve more hash rate is to increase your blocks on your graphics cards! I had this thought an had to come back to actually check what you reported you where using. For instance my Titan does 112 blocks and my 1080 is doing 100. The higher you raise this number you should see greater hash rates but keep in mind this is suppose to actually be an appropriate number you get from doing some math after counting off your actual card's block diagram. Though in my experience it's pretty simple to dial in; 1.) to high it crashes on open 2.) still to high after it opens you'll see poor negative hash results even at higher blocks. 3.) to low and your performance will start going down towards what you have it at now. There's a perfect Goldilocks zone in there where you're going to see your card's top performance. Hope this helps! B|

Link to post
Share on other sites
On 9/27/2017 at 9:19 AM, WhiteGoblin said:

Thanks for the advice jwacher! This is a thread about Monero and investigating different ways at mining it on whatever gear I can get my hands on. Maybe in the future I'll do a zcash thread and see what happens in that scene but for now I'm an XMR man as I believe in it.

 

Thank you fine sir I really appreciate it! The main problem with compiling stak is that people generally try and do it inside windows. If you're one of those folks then I'd highly recommend the following step by step tutorials: CPU & GPU. That'll walk you through getting into the dev branch of each miner. Though they're about ready to all be re released as one single package that mines across cpus & gpus regardless of brand so you might want to wait for that before going balls deep an eating up alot of your time just to have to redo it.

 

You look like you're doing great so far! Tips to see improvement in your hash rate though would be to pick your favorite virtualization software & get a copy of Snipa22's XMR-Node-Proxy up and running. This will combine your equipment's power into a single thread which once it's dialed in should give you some pretty decent improvements as far as stabilizing your valley to peak ratio. It's a bit of a process as already investigated in this thread pretty deeply, but if you need a hand just come on back & fire another reply off here!

 

Well if only this was the case, I'd really like infinite money. Even if I had infinite money you'd probably still find me just sitting here playing around with Monero. I'm really interested in this entire field. I sit in the dev chat rooms, I ask questions, I experiment and test constantly, and I'm never to scared to look like a dumbass. That's the key I think, if you want something bad enough or are interested in something strong enough you'll find your way into it no matter what. This is the gear I personally have, it's not great, it's not meant for mining, and it's certainly not giving me the best results in the world. Though I'm mining and after a month of optimizations I've gained alot of knowledge on how to be a miner, how to walk the walk, setup rigs, maximize gains, etc. This was just the dipping my toe in the water experiment I needed before getting serious. Though I guess that doesn't explain the question. Fine sir, there appears to be a misunderstanding at hand here, as I do not have infinite monies or cheeseburgers.

i mean u talk about how your going to just go ahead and grab a titan x or something so like u kinda must have infinite money

Link to post
Share on other sites
5 hours ago, WhiteGoblin said:

Hey Metal Kitty! I was thinking about you this morning and the way for you to achieve more hash rate is to increase your blocks on your graphics cards! I had this thought an had to come back to actually check what you reported you where using. For instance my Titan does 112 blocks and my 1080 is doing 100. The higher you raise this number you should see greater hash rates but keep in mind this is suppose to actually be an appropriate number you get from doing some math after counting off your actual card's block diagram. Though in my experience it's pretty simple to dial in; 1.) to high it crashes on open 2.) still to high after it opens you'll see poor negative hash results even at higher blocks. 3.) to low and your performance will start going down towards what you have it at now. There's a perfect Goldilocks zone in there where you're going to see your card's top performance. Hope this helps! B|

Thanks - managed to push an extra 38 Hashes out of my Main 970 and an extra 10 out of my GT 1030  - I am only doing this Because it's fun, and electric isn't an object to me... 
Soon I will be running a mining rig 24/7 out of a nice Air conditioned room kept at 18 degrees, and on multiple GTX 1080's SO :D 

Thanks for the help and a great write up 

I will get back to you on the Hosted Proxy soon though. Because I need to look at that idea.. but for another reason!!!! 

 

Link to post
Share on other sites
Posted · Original PosterOP

Questions Round Threadripper: How do I update XMR-Node-Proxy? Can I do it without reinstalling everything?

 

Saw this pop up in the supportXMR chatango room earlier today. It's actually rather easy and nothing to be worried about! Also, no you do not need to redo everything. Simply enough, bring up a terminal & navigate into your xmr-node-proxy folder. If you're lost in Linux or just need a little help getting your nixfu going again, I've included a picture below that has white boxes around the exact commands I've used to go through this process.

 

After you're inside simply issue a "git pull && pm2 restart proxy" command where the word proxy is to be replaced with the actual name you used while setting it up. As you can see in my example below, this was my CPUs balancer and it was located in the xmr-node-proxy2 directory since it's my second balancer I'm running. To those following along that would mean the normal directory is my GPUs and the proxy3 directory is my ARMs. Roll through all your different installations you're running using the same command and you'll be up to date without even rebooting your machine. Dead simple! B|

 

Example screenshot of me updating my second load balancer:

H9LOD0h.png

Link to post
Share on other sites
Posted · Original PosterOP

Properly Understanding Difficulty & 51% Attacks: (with special developer guests)

 

Alright so I've been sitting in the SupportXMR chatango room pretty much 24/7/365 since swapping pools. So much so I've been turned into a moderator so if you come in there spamming a bunch of nonsense you've already been warned the hammer will drop. (though seriously we're actually super friendly just don't be an asshole) The community is intelligent and extremely helpful not to mention quite global so if you don't speak english well don't worry about it & please still come on in if you have questions. Anyways I try an interact with the community as much as I can an be helpful but sometimes I gotta take a back seat when guest stars show up.

 

This morning I was just reading chat casually in the background when a really good discussion about difficulty and the possibility of 51% attacks just came out of no where when a user logged in with a great question. That's always how the best stuff starts, I've learned alot lurking in the background. Anyways I took the chat log, cut out all the ip addresses, cut out all the discussion that wasn't directly related to this exchange (there was a ton), and created the following for everybody to follow along with. It's important to note, this is not an original document, I have edited out entire parts of sentences from people when it was additional content directed at other personal questions or comments. I want to make that clear that this is not an exact c/v of the entire conversation. It's purely a creation from that to help people as an educational study tool.

 



FireiceUk: @m5m400 have you noticed that Monero blocks come in threes or fours nowadays?

M5M400: @FireiceUk I have noticed. I have also noticed that there a days with unreal lack of luck... like yesterday, where we found a 1000+% effort block

M5M400: @FireiceUk looking at the pool hashrate stats, there are pools that could do such an attack.

OpticFlow: @FireiceUk what does that mean, to have blocks in a row?

FireiceUk: @OpticFlow one way you would see it is if someone was holding onto blocks before broadcasting them

OpticFlow: @FireiceUk interesting

OpticFlow: @FireiceUk can you mine blocks in advance and use them later?

FireiceUk: @OpticFlow that's how you do the majority (51%) and related attacks

miziel0: but if you withhold the info about block... wouldn't it become invalid?

FireiceUk: @miziel0 nope, unless someone creates a longer fork in the meantime

OpticFlow: @FireiceUk so you make a hidden branch of the chain and only show one block at a time?

FireiceUk: @OpticFlow yes, 51% works like that, A finds a block, pockets, A finds another block pockets, B finds a block, A publishes two blocks, B's block is orphaned

miziel0: hmm... but we are finding consecutive blocks and I doubt m5m is keeping them secret 

GaryO22: So I'm reading the pastebin that someone linked but I'm stuck at this point "So the network consensus algorithm will choose the sub branch of blocks whith the heaviest added difficulty." (https://pastebin.com/uCjLtqqa) Can someone explain to me, hummm, what's the consensus algo ? How can two block with the same height have different difficulties ? What's a sub-branch ?

FireiceUk: @GaryO22 basics - you understand how difficulty works? if not let me know if yes then i will take it from there

GaryO22: Hum it's kind of fuzzy actually.

FireiceUk: @GaryO22 ok, imagine you are rolling a 100 sided dice

FireiceUk: @GaryO22 every ten rolls you will get something starting with 0, right?

FireiceUk: 09, 08, 07, .. 00

GaryO22: Statistically yes

FireiceUk: @GaryO22 ok so let's say network difficulty is 10

FireiceUk: @GaryO22 so you you need to roll 100 / x = 10

FireiceUk: @GaryO22 to get a block

FireiceUk: @GaryO22 100 / x >= 10 to be precise

FireiceUk: @GaryO22 your roll

FireiceUk: @GaryO22 but you can roll 04 or 07, either will get you a block, with me on that5?

OpticFlow: @FireiceUk so you stash those until the diff is <= 4?

GaryO22: Yep, 100/04 is 25 which is > 10

GaryO22: same for any roll < 10

FireiceUk: @OpticFlow depends on the hashpower, but yes the higher difficulty result the longer you can keep it

GaryO22: Why lol ?

FireiceUk: @GaryO22 correct, the rule of blockchain is, "the chain with the highest total difficulty counts"

OpticFlow: @FireiceUk yes, because it may drop to 7, but might never get as low as 4 again?

FireiceUk: @OpticFlow because it is less likely that somebody will come up with a better result before you can get yours

FireiceUk: @GaryO22 so if one person rolls 7 (diff 14) and the other one 4 (diff 25)

GaryO22: The one with 4 is accepted 

GaryO22: I think

FireiceUk: @GaryO22 then you have two chains, assuming they started on the same block, let's say total so far was 123

FireiceUk: @GaryO22 one chain has 123+14, the other one has 123+25

FireiceUk: @GaryO22 so the 4 gets accepted

GaryO22: Okay, thanks  !

FireiceUk: @GaryO22 now let's say that one person is rolling twice as fast as everyone else

FireiceUk: @GaryO22 he gets 123+25 roll, but doesn't broadcast it

FireiceUk: @GaryO22 everyone else it still working from 123

FireiceUk: @GaryO22 in a minute he has 123+25+11 chain, and someone comes up with 123+20

GaryO22: 123+20 gets rejected ?

FireiceUk: @GaryO22 he sees the 123+20 broadcast and pushes the work so far which is 123+25+11

FireiceUk: and the 123+20 gets orphaned

FireiceUk: hence two symptoms of those kind of attacks are a) long strings of consecutive blocks coming quickly b) high orphan rate

GaryO22: It's a type of attack ? really ?

FireiceUk: yup, it's called 51% attack

OpticFlow: @FireiceUk so if one obtains 51% of hashrate could get monopoly on the longest branch?

OpticFlow: and always one-up everyone that finds a block

GaryO22: Okay; well thanks for taking the time to explain this to me 

FireiceUk: @OpticFlow yes, he will be the only person that can mine a block

GaryO22: What if you get 123+11+20 and someone push 123+25 ? Does the longest is accepted even though 11 < 25 ?

FireiceUk: yes, 123+11+20 wins over 123+25

Goblin420: @FireiceUk That's pretty neat shit you two have been going on about. Thanks for sharing.

Goblin420: @FireiceUk please keep talking though, it's a good read for us noobs.

anon6291: @Goblin420 Agreed.

Link to post
Share on other sites
Posted · Original PosterOP

Goblin show us your rigs & setup, get personal with us:

 

Well outside of watching 300+ people sit in the chatango room because the stats screen is down & they're all freaking out, there has been some decent discussion this morning as well! Everybody wants to see my mining setup and honestly it's not that cool. The desk I sit at, that's pretty PCMR, I'm not going to lie, makes my epeen feel about 1000x large but in reality I'd gladly trade my cool 4K proart monitors and updesk for some rx vegas to put in my rigs. ;) I also decided I was tired of hearing the mining fan load throughout the house so I centralized all the equipment I could into a corner of one bedroom next to an open window. My desk isn't going anywhere so you still hear that but that's to be expected. This is nice because it also gives a location for everybody to return their mobile devices (laptop+phones) to for me to make sure mining is still running on them & they're being properly cooled. Speaking of properly cooled, whoa buddy that room got toasty fucking hot real quick. Even with it wide open and 57F outside, there needs to be a fan in that window asap. I've taken a picture of the temp in the room over the last 24 hours & it's pretty brutal. Thank goodness we're going into winter as this will heat my home no problem, just put a fan in the doorway instead of the window. :D

 

Looking out my window while I'm reading chat & studying XMR:

Bycq6Wf.png


The i7-5820K's kraken x61 needs cleaned after mining nonstop without the prefilter installed:

WDuW0B9.png


Network starts here with it's own dedicated battery:

J6oH435.png


It then runs into the mining corner inside the next room over:

pW3RhPg.png


Mining room over 24 hours with the window wide open to the outside:

dVK8Fkd.png


The PCMR desk I moderate your chat from & the i7-5930k+TitanXP:

Ebg3x99.png

Link to post
Share on other sites
Posted · Original PosterOP

Speed Racer 2.0: (lets go faaaaaaaaast)


 The other day I was sitting in the SupportXMR channel reading through everybody's discussion. Some users where debating the differences between mining software and it was noted that SGMiner uses two threads per GPU while everybody else uses one. The channel debated this for a while before moving on in topic. This stuck with me for the rest of the day and I started testing it out of interest to what would happen. Can you run multiple miners across your single devices to any advantage? The answer to this is yes, yes you can.
 
 Though it's a process, just like about anything I've discussed in this thread so far. I saw other people say they had tried this to no success, most said they couldn't even get XMR-STAK to turn on with multiple threads per device. Release or Dev branch, just crashes for days. The secret is not trying to add in a second miner to the configuration file like you would additional cards. The key is running the mining application TWICE, as in, two separate executable launches. THIS will connect to the same device as many times as you want! Though hard crashes, system reboots, and a whole lot of driver hell will follow if you start playing around without carefully considering everything.


 Now how to dial in this process is simply by dividing your graphics card's power into smaller but equal separate threads to run. (please turn off your overclocks for the original configuration part of this, we will turn them back up later don't worry.) Example my 1080 Waterforce has twenty streaming multiprocessors, so I could run four miners at five threads each instead of one at twenty. The card at 2.1/5.5ghz overclock can handle 100 blocks across 20 threads. Though if you put a miner on 2 threads it can still push the same 100 blocks, actually it can push way more blocks then 100 before crashing. So if I open up two two-threaded miners both at say 150 blocks each now I'm getting 300 blocks out of four threads instead of the 100 blocks I was seeing before.
 
 I've spend DAYS testing this theory and there is a breaking point where you're just maxing your gear out and causing issues. Example, 14x2 threaded workers on my titan just bogs it to the ground, not that you can't do it successfully, it's just not optimal. Though on my 1080 I saw a 50 H/s improvement splitting it into ten two threaded miners. There's just alot of cool numbers going on the background and it all does actually make sense once you dig deep enough into it. Through the division of power you can achieve some pretty radical results depending on how you've setup your load balancers. (which is an important element to this)
 
 I would highly recommend starting out by dividing your graphic card's power in four. While I've seen success dividing it more times then this it's a really minor gain for an incredible amount of system stress and additional network load. Example: My 1080 has 20 SM's so now I'm running four five threaded miners across it. My Titan XP has 28 SM's so it's now running four seven threaded miners across it. It's really important you measure your pre-split pre-load balancer hash rates, both max and averages. Then compare these numbers to your split numbers. If you're not achieving atleast the same total level of performance as before the split then you've configured things wrong. 
 
 Like setting up a single thread, to many blocks is a bad thing! Once you've got them fired up it's important to watch all your miners work at once and then occasionally test their hashing rates (this can be done easily by enabling verbose level 4 automatic output reporting) then seeing if everybody is working at the same rate relative to their difficult they've been dynamically assigned. If you're not seeing an equal division of your processing power then your system is having issues keeping up with the division of the work, you're bottlenecking somewhere. Lower your blocks by values of two until you achieve equality across all the miners. Or on the other side, if you turn it on and everything is equal right away but really low hashing rate numbers then you should increase it by two until it starts crashing.

 

 If everything is stable & the rates look good across your split up device load then you can kick up your overclocks again. What you're going to see is a CUDA crash on your last opened miner if you've over-overlocked your device. If you've out in space overclocked it then it'll start crashing on multiple mining threads instead of just the last one. There's two ways to handle this situation 1.) step your overclocks down a little, bit by bit, until it's stable with your current setup across all the mining threads 2.) go into your config file and slowly step down the block rate by values of two until it no longer crashes when your normal pre-workload division overclocks turn on. This can be really handy because maintaining the same clocks you had running beforehand makes it easier to analyze your results. (plus to me, it confirms you're achieving a different configuration at the same stability and speeds)

 

 Finally once you've got equally divided workloads (once again, taking variable difficulty into account) running without problem & they're all overclocked then you're pretty close to dialing this in. You might still notice your hash rate isn't what it use to be or that your improvement was pretty small overall. Scratching my head I fired up alot of tests to try and pinpoint my error in understanding. Then it hit me while watching the graphics card's power consumption rates/levels in afterburner, what if the reduced threads weren't asking for enough juice? It seemed like I wasn't seeing those spikes on the power graph anymore. I thought that would be nice as it was a more stabilized & efficient way to mine plus both systems seemed to exhibit much faster response times now!

 

 Though if the systems weren't getting wrecked as hard then it seemed logical I wasn't getting the max out of them through this division of mining capability. How do I get the juice back up & fire these smaller work loads up a notch? Bfactor! Previously single threaded I ran at a bfactor of 8 on systems that nobody was touching often and all the way back to 14 on systems people used all day. It made the automated scaling up and down of the mining processes in the background less noticeable. (even though it did mine less) Anything lower then 8 and you're in for a bad time inside environments meant to be used by users all day. Though since we're doing little miners maybe they could handle a lower factor?

 

 YES! Yes they can! I'm able to get them all the way down to 4, half the bfactor of before! Though for responsiveness I pushed them back up one to 5, then up again another to 6 on the most heavily used systems. You can go into afterburner and watch the power consumption graph, once you've got everything dialed in properly it should be jumping up to those peaks from your old valley but at a much more rapid rate now. You're not asking one guy to climb to the top anymore you're asking four people to, so you can literally watch four spikes in your power chart now instead of one.

 

 Once you see your baseline power numbers return to normal and you see your peaks return to around where they where before then you're back to dialed in mining on that card. Depending on how far you push your bfactor you'll get that full on system lag where you wait five seconds after clicking something. Even at dialed back rates making the systems MORE usable throughout the day then they use to be I'm seeing the benefits. Now even the titan is surpassing it's previous max hash rates. I saw it peak up in the low 1100 H/s when previously the highest I'd been able to get out of it was around 990 H/s.

 

 Inside the SupportXMR android application when I'd be watching for peaks I'd generally be stoked to see 2 KH/s on any given proxy but when I kicked up this new concept I saw a 2.2 KH/s peak within a couple minutes! New record! By 200 H/s even! Not only this, but our block counts are through the roof now! Previously the 1080 did 100 blocks, now it does 660!! The Titan does 704 now!! So while we're talking the same exact gear, we're seeing 660-704% increases in block rate when configured in this manner.

 

 Now talking CPUs instead of GPUs, to achieve this effect with your processor all you have to do is run as many whole copies of the miner as you can fit into your cache. Example: My i7-5820K & 5930K both have 15MB L3 cache. I can get 7x2MB miners into that 15MB fully. So I remove affinity and fire up seven individual copies of a single threaded full power miner. This saw another jump in performance! I was actually surprised that the CPUs could yield even more hash rate then I had them at already as when compared with others they're already ahead of the crowd.

 

 Doing all of this division work also has the additional benefit of working with xmr-node-proxy on levels it was more intended for. The dev has said 10+ devices is a good point to start, but normally a non-server related box would only have 1 cpu miner and 1-2 gpu miners. So per computer people would only have 2-3 devices to toss at the that basic entry point of ten. Though you split up your devices into smaller wokers and all a sudden you've got like 12 devices per machine! Your load balancer is going nuts by the time you distribute out your entire network of machines in this new manner. (and the amount of data flying around your network increases proportionally as well, just a heads up!)

 

 This of course dramatically lowers the difficulty of what you're pushing. (keeping with the high speed low difficulty concept) As you're dividing the same equipment's power, it's not like you added any new hardware, you're just running many more miners across it now. We're all carrying tiny little bricks to the job site and back instead of lugging boulders around. Load balancer wise I've had to change how things are running to maximize these new changes. I've lowered my starting advertised difficulties considerably.

 

 Since we're dividing the work up so many more times now it's going to take the load balancer forever to automatically adjust to an optimal rate if I'm asking everybody to still do the heavy lifting from before. It's important to note, this will fix itself overtime, but not telling a bunch of little guys to lift a couple tons helps in the grand scheme of getting things working quickly. My shareTargetTime is 0.5 on my CPUs and ARMs while that is just a bit to fast for GPUs to really extend their legs & dig in. Because of this the GPU proxy doubles down & runs on a shareTargetTime of 1.0. Anything beyond these numbers starts to show a loss of hashing power throughout all of my tests & believe me I've been testing this for a while now. (LOL look at the thread you're reading)

 

 So there you have it, that's a crash course to Speed Racer 2.0. You can now duplicate this across your systems and see what results you get! ^_^


Speed Racer 2.0 Running On i7-5930+TitanXP Box:

ZSKdq9l.png

 

Load Balancers Showing New Miners: (gpu and cpus running within 10 hash of each other!)

yDeSzd6.png

 

A New Individual Proxy (CPU) Record Post Changes:

wb5FYKv.png

Link to post
Share on other sites
Posted · Original PosterOP

Man, I haven't been in a BestBuy for a minute. I shrugged my shoulders and figured how bad could it be? They had to have some type of decent graphics cards right? Maybe a little over priced? Let's just walk in and find out. Finally found the graphics card isle an it's empty, all out of stock signs. Three employees where all standing one row over circle jerking so I yelled over, guys, what, does best buy not sell graphics cards anymore? They all laughed and said "bro, bitcoin, I don't know if you've ever heard of it? it's a crypto-currency thing, don't worry about it, basically, been sold out for the last six months straight, still sold out, never lasts longer then a day when we get them in, you can drink beer for free until you pass out, every day, so all the kids grab them!" at which I just laughed an walked out. xD

 

Yup, there's a mining craze still going on:

wcUZyuL.png

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

Newegg

×