Jump to content
WhiteGoblin

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

Recommended Posts

Posted · Original PosterOP

Okay so basically the idea was, with my gear I have laying around my place, can I actually mine? Everybody is on that Ethereum (ETH) craze but I was wanting something another step removed, something different, which landed me on Monero (XMR)! Now there's two desktops, an older laptop, and I'm considering the entire network to be a power usage for this as well. Each device has it's own 1500PFCLCD that keeps it online and running clean, though the laptop and network share one. Now I want to point out before I start getting into the stats, I just started out with XMR. I've spent a good amount of time doing little tweaks and optimizations, while watching system stability throughout the day. Though I'm positive more knowledgeable miners would be able to achieve much better results then what I'm showing here, though this is how we learn! :D


CPU #1 : i7-5930K @ 4.5ghz across 10 threads = 331.5 H/s
GPU #1 : Geforce Titan X (Pascal) @ 2ghz/5ghz = 872.6 H/s
POW #1 : 396W (monitors, speakers, mics off)
CPU #2 : i7-5820K @ 4ghz across 11 threads = 304.8 H/s
GPU #2 : Geforce GTX 1080 (Gigabyte WaterForce) @ 2ghz/5ghz = 581.7 H/s
POW #2 : 324W (monitors & speakers turned off)
CPU #3 : i7-3720QM @ 3.4ghz across 7 threads = 131.1 H/s
POW #3 : 108W (laptop lid shut) (includes modem, router, & pi-hole)

Total Hash Rate : 2221.7 H/s
Total Power : 828W


Now this returns after all the power is accounted for $4.99 a day, or $149.72 a month! That's of course at this exact moments exchange rates though I'm not complaining for power that was going unharnessed beforehand. My gig internet is $200 a month, so it's paying for about 75% of my internet bill. Though as you can see, this is not how miners build rigs! A Titan X (Pascal) on an insane custom EKWB loop? Yeah that's not helping anyone make money mining, nobody would ever purchase this gear for mining. Though like I said, they're my personal machines, so I figured why not, let's see just what they'll do! Though I know you guys are going to want screenshots on everything, some text in a box just isn't going to cut it! So I've came prepared!

 

Box #1 Mining:

n6Odj6G.png


Box #1 Power Draw:

Kom4Amd.png


Box #2: Mining

rp5iVjx.png


Box #2: Power Draw:

3xb5Ac7.png


Laptop Mining:

OxRHr9Q.png


Laptop & Network Power Draw:

LjtED8A.png


Profitability Calculator:

0LrYEk5.png


Overall this has been a fun day playing around with a newer crypto currency and looking to see what can actually be done with the gear laying around your home! I've got a couple old xeon boxes in storage that I might have to drag out, strip down, and see what they can mine in a linux environment. These two boxes still see daily use, so the miners have been dialed down a little bit to still allow for a responsive & easy to use interface. Though those boxes I could just go full tilt inside linux as they'd serve no other purpose. If I get around to this I'll be sure to post additional results! Anyways, I hope this was helpful or atleast entertaining to some of you out there. :) 

 

'Gobby

 

Update: I had already walked away from making this post when it hit me I had made a huge omission. Look at those overclocks on the graphics cards. Look a little odd? When you fire up the mining software it down clocks the memory on the cards. So once you've started mining (atleast on water), you can start pulling your clocks up again. Though, make sure you close afterburner before you close your miner if you're ever going to stop. Because once the miner closes, it'll return your clocks to normal, which means you've now got some stupid AF overclocks going on. So yeah, remember that little note. As you can see, the screenshots were taken literally like 6-7 hours into mining so the overclocks are in full force! Okay, now I'm going to walk away again. xD

Link to post
Share on other sites
Posted · Original PosterOP

Day 2 Updates! 

 

Woke up this morning and everything was still mining just fine. Though I wanted to see if I couldn't really optimize things, push them to the max, and overall improve the gains I'm seeing off this home mining setup. The first thing I did was download the newest graphics drivers and install them, then I started looking at max gpu overclocks from the standard online communities. Finally I looked at all my settings files looking for mistakes, or places I could try and push things further. After all this became the actual process of testing the new overclocks on the graphics cards, which went smoother then I expected. Both the Titan X (Pascal) and the 1080 WaterForce are running smoothly at 2113mhz on the Core though the memory on the Titan is at 5514mhz while the memory on the WaterForce is at 5556mhz. This is purely because the slider in Afterburner maxes out at +1000 which is where it's sitting on the Titan post mining downclock to achieve the 5514mhz. I also went in an enabled the network stats pages so that I didn't actually have to turn on monitors to check progress of machines & could do it easily from my phone anywhere in my home. (a nice touch!) What did all this gain me? Well a 7.57% increase in mining output! My new hash rate across the gear is 2389.9 H/s based on the 15 minute averages they record. (not the highest hit rates alot of people seem to like to post)

 

Though let's talk economics here! You know what really made my day? I called Comcast this morning, yup, on a Sunday, their call center was dead silent in the background. I spent 52 minutes on the phone chatting up a guy from Arizona. I got my bill reduced to $113 a month pre-TB cap fees. So in reality, an hour on the phone spent blind fishing a better deal has turned this whole operation into an extreme win! Now I'm completely paying for my internet bill with this setup, even on the months I have to pay heavy in extra bandwidth. ($50 extra for unlimited) Sooo yeah, Monero baby, you making me pretty happy. In comparison, I saw on reddit a guy with a new RX Vega 64 who is getting hash rates of about 1150ish while burning 300w. That makes this cracked out Titan X (Pascal) feel a bit better now that it's pushing 960.6 given it's only using 60% of it's power most the time and spiking up to 84% as you can see in previous afterburner pictures. Though let's be clear, I understand, this is a freaking titan, and not a card bought to mine on. LOL! Instead of coming back with all the same pictures this time, just with different clock speeds, I've put together all the neat little network screens showing my current hash rate.

 

Q8VjVw9.png


Once again, I hope somebody out there finds this information useful or at least entertaining! :D

 

'Gobby

Link to post
Share on other sites

Great write-up! I'm a first-time miner myself, trying to decide what currency to go for. I live in California where my electric costs are tiered ($0.19 kWh - $0.26 kWh), so total power draw calculations are critical to determining profitability. Just ordered a Kill-A-Watt meter to test my existing 1070 GTX gaming rig for total power draw before mining anything. Just out of curiosity, how much is your per kWh cost wherever you live?

Link to post
Share on other sites

I might have to try this out since my PC sits idle all day long except on the weekends. Might as well generate me some cash on the side.

 

I just do not know how to get started but im sure there are some guides here on the LTT forums.

 

Well Ill get everything setup tomorrow since I am going to be off of work and hopefully it all works out. !!!


RGB LIFE

CPU: Intel i7 5820K @ 4.5GHz - 1.29v Motherboard: ASUS X99 DELUXE/U3.1 RAM: Corsair Dominator Platinum 16GB Quad Channel (4x4GB) GPU: EVGA GTX 1080Ti FTW3  SSD: Samsung 840 EVO 250GB HDD: Seagate 1TB Barracuda Case: NZXT S340 Elite PSU: EVGA SuperNOVA 1300w G2 Cooling: Corsair H100i v2, Corsair HD120(x4) Misc: CableMod White Extensions, Corsair Lighting Node Pro

Link to post
Share on other sites
Posted · Original PosterOP

Days 3 - 6 Updates:

 

Welcome back to the mining adventure everybody! Sorry for the long gap since updates and for not responding to questions on the different foras I posted this! So as the story goes, I did infact head across town and grab a dual xeon machine out of storage, it had sat in Colorado storage for a while meaning it had seen temps as low as -15F an as high as 100+. I'd never had a problem with it before but it was over a decade old now & when I pulled it apart looked like I'd ganked all the parts out of it once before an forgotten. I headed over to the CSU Surplus Store where they sell just about anything a nerd could want at discount prices. Example? Ram there is $2.50 a gig, doesn't matter type, speed, or manufacturer. They just have bins labelled "DDR 2 ECC/REG" etc. an you dig through them. I dropped in & grabbed six sticks for the xeon box (it maxes out at 12gb so six 2gb sticks), the best video card they had in stock an old firepro V3900 for $7, & $10 for a 160GB 10K velociraptor to slap in there. Total price was $51 for all the upgrades so in terms of stepping on the mining profits, not a cheap investment for a shot in the dark.

 

I got home, attempted to put them in, and oh no I'm a dumbass. I had grabbed DDR2 F thinking for sure this old xeon server ran fully buffered. Though as soon as you try to put a stick in and see that you're a couple pins off on your notch you'll for sure feel like a very bad tech. It was already to late to head back so I just called it for the day & started playing around with the Titan more. (more on that later) The next day I dropped back in and explained to the tech there testing parts the mistake I had made. He laughed, said no big deal, let me swap my ram out for the proper model. Cool deal, get home, plug them all in. Machine beep code of doom. Look it up, memory error. Oh cool I can't use slots 5 & 6 in this configuration, okay well whatever 8gb is still better then nothing. Pull them. BEEP CODE OF DOOM. Go into bios, look it over, proceed to spend the next hour with not just these sticks but other sticks testing everything I could imagine. Turns out from ?? who knows ?? things are now really screwy and not like how I last left the box before long term storage. Memory slots 1, 3, & 5 are gonezo, no detection, nothing. Memory slots 2, 4, & 6 work but only detect 1gb sticks I pulled from another box, none of the sticks I paid $30 for work at all. 

 

Being positive minded I'm all like, we can still make this work, fuck it! Let's just put Debian on here, start mining with 3gb of ram, and see what hash rates we get. Well okay, make a usb boot drive, go to install. System Halted. Can't get past boot, always System Halted. Throws a memory config error right before. Go into bios, notice that while it's reading the 1gb sticks, the system can't tell how much memory is actually installed, it's totally fucked up. This is a lost cause, but the day was late and I had other things to do. So next morning comes, I load up all the parts I bought, plus the other ram I had pulled as I didn't need it anymore & headed back to the surplus. They don't take parts back and you have to pay to recycle there. I said well look guys, I'm going to give this to somebody on craigslist or you can just consider it a $50+ donation to the school and sell these parts to the next guy. Somebody is going to come in needing this ECC/Reg ram and I don't want to hold it for no reason at all. They said they'd make an exception and appreciated the donation. I looked around for a while and they had a tiny little box in the corner for $75 bucks, I asked them about a deal on it & they said wait a second. Went into the tech bay for a bit, then came out with a much newer version of the same box! Went from i5 to i7. They said since I had made a donation they'd just make this a really good deal if I was interested. So I snapped it up, got home, put an old Server 2012 R2 copy I had on it for giggles, spent hours running updates even though it was a blank server install with no additional roles added / installed.

 

Finally I got to pretty much a blank screen with nothing but my network adapter enabled. Fired up XMR-Stak, 225-250 H/s! Turned on the stat screen for the battery it's plugged into, 63W! If you add the laptop into this you have 375+ H/s at around 100W of use which is more hash then my cranked out i7-5930K is doing at 4.5ghz across 11 threads. (though it's also still being used as a daily machine) These are some extremely interesting numbers to me! Though for the ones you guys probably want to see more then this, I went test mode over 9000 on this EKWB loop the Titan is running on. (with plenty of screenshots) One screen I grabbed with Afterburner & Taskmanager is pretty amazing because I don't think you'll see a Titan X Pascal jacked that high again without going beyond water. If there is other overclockers out there that have three+ day long mines at higher rates then I can't imagine their setups. I'm not saying it doesn't exist, I'm just saying, IMHO, this is the limit of pascal. I was even getting my nerves shaken a bit before finding the absolute limits. Enough you say, give us the numbers!! Well 2,240mhz which was consuming 61% power draw at 36C was when the system finally crashed. Which was not instant, I had music playing on the very system as I was doing this, not just for pleasure but listening for stutter or anything to say trouble was about. After about five minutes at this speed the music stuttered, slowed, and then stopped. System locked. I also noticed that only at 2,240mhz did I see a 61% power draw on the bottom line pre and post mining jumps. Even at 2,225 it was 60% and the temp stayed 35C, but that last additional 15mhz was the straw that broke the camels back so to speak. Which is probably pretty insane to most of you out there, as it was pretty god damn shocking to me.

 

Post reboot, the card is a little jank over 2200, you can't go right to that speed, you need to ease into it after a minute of mining. Also the system response goes to shit over 2200, so I've decided it's probably wise to keep the card at 2,175 for a little bit of a buffer & keeping the responsiveness nice. That is an absolutely mine numbing pascal overclock (atleast to me personally) & keep in mind, as the screenshot shows, it was mining for three days at 2214 before I just pushed it to far with the 2240 request. Truly curiosity killing the cat. Hash wise? Didn't make a real difference compared to the stresses placed on the card. The highest I saw it mine was 979 H/s, but it wasn't consistent an the card liked to mine at 965-970 most the time. Dropping the overclock from the balls to the walls limits of the damn card results in a slightly higher hash rate an a more stable system. If this loop wasn't pretty much always supplying room temp water to the GPU across a full block I don't think you could hit these speeds anyways, they're a wasted investment in the bigger picture of mining, nobody would do this. Though it's gotta be interesting to more people out there then just me! Feel like some LTT/J2C/GN video up in here taking the card that far. So back on track, now I have, two personal machines, a laptop, & a little micro workstation mining Monero. My new combined hash rate post a week of dickery is 2,624 H/s. It's been higher, it's been lower, there's alot of drop when a user is on one of the two desktops with two 4K monitors needing to be powered by the cards as well. You'll see the big numbers after a 12 hour stretch at night when nobody has touched them at all for a long time. Though outside of adding 63W for the new i7 the power usage has pretty much stayed the same. The Titan went from 59% power usage on normal boost clocks to 60% power usage where it's at today, the 1080 hasn't changed, & the cpus don't change really. The only downside was I woke up to a windows update kick in the nuts on the laptop once as I forgot to turn automation off. :D

 

So there you have it, just about a week into mining some Monero, still figuring things out, still achieving optimizations! Pictures to follow:

 

The Xeon Box, Fresh Out Of Storage:

 

pJpUjAL.png


Inside of the Xeon:

 

vMAmDRT.png


Xeon Upgrade Supplies:

 

0LHJnbV.png


Cost of Upgrade That Didn't Happen:

 

X0car85.png


Remember Kids, Research Your DDR2 Needs Before You Buy:

 

9uAK1KE.png


The New Micro Box:

 

c5c8KBe.png

 

Lets Open Her Up & Free That Cooling:

eZwSQ6z.png


Can You Mine Micro Box? Yes You Can:

 

CkeLsLF.png


At the Brutal Cost of 63W:

 

hnK3INa.png


Cracked out Titan X Pascal Hash Rates:

 

OCN8lkR.png


Pushing Pascal to the Limits:

 

QSQ6Q2Q.png

Link to post
Share on other sites
Posted · Original PosterOP
On 8/27/2017 at 5:09 PM, TheSaint said:

Just out of curiosity, how much is your per kWh cost wherever you live?

My city works on a tiered system where the first 500 kWh is $0.09434 from Jun-Aug & $0.08893 from Sept-May. Then after that the next 500 kWh is $0.11268 - $0.09339. Then all juice after that is $0.14904 - $0.1033. There is also a fixed monthly charge of $6.14 regardless what time of the year it is. ^_^

Link to post
Share on other sites
Posted · Original PosterOP

Another question that was asked over Steam instead of a forum was, just how profitable are the laptop & mini station since they're purely CPU mining all the way here almost into September 2017. Well I went ahead and ran the calculator on that for you! A couple things to keep in mind is that this is including the donation to the developers (2%) & pool fees (0.5%) which I fully support. (I just added the dev time into the calc with more pool fees) Couldn't do it without them! Also this is at the current market rates today, live. I'm not really interested in selling this coin asap. :)

 

Ol'Laptop Profits: (i7-3720QM across 7 threads inside Win 10 Pro)

42remcg.png


Ol'Station Profits: (i7-3770S across 7 threads inside Server 2012 R2)

wt4HlIc.png

Link to post
Share on other sites
Posted · Original PosterOP

Post Week Mining Updates:

 

Well I posted a link to this in my mining pool's chat this morning, so hello everybody from US XMR! Everything is still chugging along great but it's been a pretty rough week for the pool I'm connected to. The news ticker on the front page talks about what's going on & I even saw the admin on reddit discussing the problems and how to fix them. I like that he's proactive like that & out in the community, it's a solid touch to the small pool vibe. Do not let this entire realm of troubles slip your mind if you're thinking of getting into mining. Sometimes it's not your fault! Sometimes, your entire setup is running perfectly & your internet goes down or the mining pool has an issue. So you don't always get that constant 24/7/365 profit the calculators are predicting. I've also been balancing the gear I own, how to run it most effective, and how user impact effects the machines. To put this into perspective, when nobody is on any machines at all, the pool is showing my setup as popping up to 3.3-3.5 K/s. Though during the day I tank really hard, I took some screenies to show off this effect after a couple minutes of light user work on a single 4K display.

 

I also wanted to come do a video or in failure write a section about measuring power draw appropriately and to show off the "mining spike" as I'm just calling it randomly here. (I'm completely unfamiliar enough that if there is a proper technical term for it, well, you're getting "mining spike" instead) (mmmm some sexy deep technical analyst up in these posts) Though the video I shot was pretty bad & as soon as I bust out a decent camera to do anything my OCD will kick in & you guys will get a 20 minute YouTube video on my entire setup, the ins and outs, and how it all works. For as much time as I'd put into it, I'm not really sure anyone cares! LOL! But I'm still considering doing it just for the few folks out there that are searching out more information on the subject. What I'm talking about can be seen in my screenshots above inside Afterburner, see the power draw I always pulled up before taking the pics? Now see the spikes in the power? This does not happen on CPU mining it's purely a GPU related activity. So when you're doing your power calculations, most the time you're going to see a readout of one number, the lower pre and post spike required juice. Though as it's picking up work, or doing whatever, you'll see a smaller climb, a peak, and then the climb back down.

 

This isn't my field, I don't have writers, and honestly for the places I'm posting this information there is about a thousand better people to explain this in a technical manner. Though for the novice out there, I feel like I have some decent ability to show that off to those interested. The 5820K + 1080 box currently after the week of playing around runs 333w for the baseline & spikes to a peak of 381w after a couple minutes watching it, though the vast majority of the time it's at 333w. I can't even imagine what a shitty power supply on some questionable quality power thinks of this activity for weeks, months, years on end if you plugged in say four to six graphics cards all doing this. Which is not gear I own, I'd love to show it off to you but I simple don't have it. The 5930K + Titan X (Pascal) box uses 396w all the time but spikes to a max of 477w after a couple mins spent watching the battery. Every other device I have doesn't flux at all, I plugged in each part of the network gear out of interest. The modem & router each take 20w & the pi-hole takes 5w. (just to give you all out there an idea of overhead) To be extra safe if you're going to do this for more then a giggle, I'd probably run my max numbers in my power calcs just to make sure you're not going to be sour at the end of each month thanks to the difference per series of spikes. See, you just turned that frown upside down, now you'll be pleasantly surprised each month! ;)

 

So let's talk about some additional changes. I actually changed the venting system in my place to hit the rooms with mining gear in them alot harder then the rest of the place. Do not under estimate a mining rigs ability to output heat. Sitting next to this titan box with two giant radiators pouring heat onto your legs will make you start to sweat. I also pulled the cool magnetic dust covers out to allow for greater airflow. I don't think it actually needs it, but it makes me feel better, and I'll just be more anal about cleaning it more regularly. Also I've become an absolute power nazi, I do an OCD check of my entire place to make sure things are turned off before I go anywhere or even to bed. I also turned my fridge and freezer down to 1/10 on the dial. The idea going through my head is, this is all combating the efforts of my mining rigs! The more I output the more I'm damning my efforts. Though in the big picture as the world has let me know thanks to the powers of the internet, I'm pretty much in mining heaven, my power is free compared to many places of the world. So I guess it's just a bit of a hippie'sh power nazi vibe (is that a thing? can I say that?) ontop of being in the golden land for power, that is driving my quests for squeezing more out of this operation. 

 

Because of this I overclocked my 5820K back up to 4.5ghz where it use to stand. Though for gaming, it really made no difference and you could run the fans lower 24/7 when it was only clocked at 4ghz. Though, we're mining here, so back to 4.5 it went! Difference? As seen in the screenshots below, about 10 H/s. Yeah I said that. Additional 500mhz overclock = 10 H/s gain on a i7-5820K. Whew talking big numbers here! Ultimately I hit a new high this morning on the Titan as well, I've not changed a single setting, it's still going though the temp is currently 33C across the card thanks to it just not being hot as hell outside. 982.7 H/s with much better averages, always staying in that 980-982 butterzone. Though after just a minute of being on this machine with a single 4K screen being turned on, just reading the forums, my rate dropped to 957.3 H/s! A swift kick in the mining nuts & imagine the additional stress it's placing on the card as people flip stuff on an off all day. Though it's solid & stable, they never turn off or crash, the graphics drivers never freak out, it's pretty much easy sailing if I'd just stop playing around constantly trying to squeeze every last H/s out of them. :)

 

5820K @ 4.5ghz Mining Across 11 Threads: (compared to the 4ghz screenshot from above)

sRWukuq.png


New Titan Record For This Rig:

XCuLIiu.png


Record Destroyed The Second You Turn On a Single 4K Monitor:

m3Nqfpc.png

Link to post
Share on other sites
Posted · Original PosterOP

Finding another 25 H/s via good tech work:

 

Well spite what I said about needing to stop playing around with things, here in the post week mining phase, I'm still learning! I just pulled another 25 H/s out of the same equipment through nothing more then optimization. Now the first 20 was gained through playing with the "low_power_mode" option inside XMR-STAK-CPU. I had previously played around with this and found it to be heavily favoring "false" mode on all my setups. To the point where if I turn it on the mini workstation, it'll go from 225 H/s to 150 H/s. The laptop takes a dive, as well the i7-5820K, though to be fair, it only goes down by about 5-10 H/s, not the huge dives from the other two machines.

 

I was reading the official dev boards & a guy was on there talking about how he got much better results by turning the actual cores on "false" though putting the hyper-threads on "true". This sounded interesting so I tried it.. on every machine. This resulted in the most fucked up hash rates I've gotten yet, though there was a golden nugget in the frustration. I tried all "true" mode, which from the config file: "This mode will double the cache usage, and double the single thread performance. It will consume much less power (as less cores are working), but will max out at around 80-85% of the maximum performance." Which for whatever reason, this time around, on the i7-5930K, boosted me another 20 H/s stable, all the time 24/7. So yeah, try every different machine you have across that setting, see how it treats you.

 

The next 5 H/s is really nerdy and well, we're talking finding all the nooks and crannies here. So I was looking at the threaded mining rates on the CPUs across all the systems. They all seemed really stable minus my i7-5820K that I had recently brought back up to 4.5ghz from 4ghz just for the purpose of mining. As I reported in the previous posts before this one, I only saw a 10 H/s improvement doing so. Basically there would be a really high hash rate on a single thread while other ones would be really low. There wasn't the stabilization I was seeing on the other machines. At first I wondered what it could be but then it hit me, maybe the machine wants more juice! Now this overclock passes all the tests I can throw at it, it benchmarks, torture tests, & games like a champion at 1.275 core voltage and 1.95 input voltage. Though here I am looking at the thing throwing odd numbers at me in the reports while mining. So I went into the bios and tried 1.28 on the core and found my rates stabilized a little bit an went up 2 H/s. I went up to 1.29 and found it went up 4 H/s.  So I went up to 1.3 and it seemed to greatly stabilize the threaded numbers but only went 5 H/s above, or overall now 15 H/s gained from the 500mhz overclock, instead of the original 10 H/s. Goes to show that your system might not lock up on you, or even give you any problems, but your CPU might still be very hungry for power while mining!

 

Now I'm sure I could try to hit 4.6ghz on the 5820K and probably 4.75ghz on the crazy loop the 5930K is on. Though I've always respected these processors and the boards they're on as holding up great at 4.5ghz all day and night. Anytime I've taken these past, into the realm beyond 4.5ghz, they've always gotten screwy and into power monger states where the amount of return you're getting is not justifiable by the amount of overall havoc you're causing. I mean sure, alot of people run these processors at 1.35 on the core but that makes me sweat a little to much even on high end water. The objective is make a couple bucks extra here, not toast my gear that cost an arm & leg to acquire already. If you've read this entire thread then let me just say, I'm quite pleased with that 5 H/s gained. I feel most would've missed that, or not cared, but no sir, been a nerd for way to long & those numbers just looked wrong. Even to an XMR mining noob like myself. :)

 

So there you have it, yet another 25 H/s pulled out of it. Though it's gotta be getting close to just maxed. I'll let you guys know if I come up with any other hash rate increasing tricks or tips. xD

Link to post
Share on other sites
Posted · Original PosterOP

Gaining 93.7 H/s through more overclocking!

 

Okay now first I want to say, this is something I'm comfortable with doing but I'm on premium equipment. This is not something I'd recommend everybody run out and do, specifically if you're not thermally prepared. Back when X99 launched I was hyping machines I built for their ability to hit those high cache overclocks, pushing the NB and everything far beyond what is actually required to post high overclocks & get on the leader boards. I caught ALOT of shit for doing this & most the popular guides specifically pointed out that high caches DID NOT help you score higher on the leaderboards, caused tremendous amount of stress on your system including heat through the additional juice required, and overall would not help you game / enjoy your computer anymore then avoiding these overclocks. Whatever, I did my thing for quite a long time. Though being honest, I did come across some situations where that higher cache overclocks created alot of stability issues. Specifically on lower end gear trying to achieve things out of it's class.

 

Though after my last post I left my place & grabbed some coffee. I've not had caffeine in a hot minute so while it was hitting me I kept thinking about the day & my results I've been getting. Mining is a different scenario on equipment then gaming by a long shot. The cache bragging I was doing back in the day, might hold true here. Maybe hopping the whole system up and eating the increased power it takes would improve my hash rates. Finally I could point at the overclockers fan bois who just read forums all the time with no technical expertise and say, look, there is a reason for this guys, it's not just about benchmark software & will it run crisis. (crysis? whatever) So yeah, I headed home, turned off my 5820K rig, jacked it's cache up to 4ghz and turned it back on. CPU hash rates through the roof! I wish you guys could've seen my smile. Turned it off, jacked it up to 4.5ghz, which is mind you, a pretty ridiculous overclock, I went to stress this, and my machine booted into Windows 10 Pro. Then as soon as I double-clicked the miner it locked up. I rebooted, went in and changed my cache voltage from 1.20 to 1.25. So to be clear, I'm running core 1.3, cache 1.25, with an input voltage of 1.95. Fired the machine up, wham  363.6 H/s. Now that is an improvement. Keep in mind I've got one thread that is just running windows in the background while managing the system & all the networking, vpn, security, etc. It's not like I can really give that thread up, but if I could, this 5820K would be pushing 390ish H/s. I think that's pretty major, I looked at that & thought, damn that's kicking my 5930K's ass now!!

 

Which leads to my next move, running over, shutting down the Titan rig, and booting it up with the same exact voltages, same cache overclocks, just a slightly higher system agent voltage. Get into Windows 10 Pro, flip on the mining after everything else boots up, and what do you know it's pushing 382.8 H/s!! Once again, one thread is running the system but if it wasn't this CPU would be doing 410ish H/s. Also pretty impressive to me, though I'm just a novice in this area. For those wanting easier stats, that's a 45.2 H/s gain on the 5820K and a 48.5 H/s gain on the 5930K! Given I just wrote a lengthy post about how I squeezed 25 extra H/s out of things, this is a pretty massive improvement. Both systems have been up for a little over an hour now with the new overclocks in place, mining their ass off. Though once again, I want to stress, this is some shit to do to a computer. Be prepared for this not to go so easy or friendly. I've got everything on batteries, there's no quality lacking, and this is not a traditional mining setup, this is a gaming and studio workstation that I'm playing around with. :)

 

5820K @ 4.5ghz Core / 4.5ghz Cache Mining w/ 11 Threads: (inside Windows 10 Pro)

9BzbX9M.png

 

5930K @ 4.5ghz Core / 4.5ghz Cache Mining w/ 11 Threads:  (inside Windows 10 Pro)

qClho4G.png

Link to post
Share on other sites
Posted · Original PosterOP

Post Cache Overclock Notes:

 

Well gents (and ladies!) I did say, and I quote (myself); "Though once again, I want to stress, this is some shit to do to a computer. Be prepared for this not to go so easy or friendly."  Even on what I consider to be pretty decent gear the jump up to dual 4.5ghz overclocks on each machine did result in mass downtime. Not because anything broke or was damaged, but because I'm not 24/7 watching the entire setup & haven't scripted email alerts into all this yet. The 5820K would mine just fine for 2-3 hours then out of no where just lockup, same type of lockup you see when your processor is starving for more juice & just can't get enough to do what it wants. I went in and changed the cache voltage up to 1.275 and it once again, after a couple hours in, froze. I was actually sleeping when this one hit so an email wouldn't have waken me up anyways. As soon as I noticed I rebooted the box, increased it's cache voltage up to 1.3 and then also increased it's system agent voltage by 0.05. This has resulted in a completely stable machine for the last eight hours or so mining at the higher hash rates though the heat output has also gone up (as expected) and I can't imagine the system loves that 1.3 cache voltage, that's as far as I'm concerned, some saucy shit. The CPU clock was stable before playing with the cache, as well the graphics card, so I didn't want to touch much else in an attempt to stabilize it. Overall, unless something changes, that's going to be where it sits as I'm completely out of ideas on how to improve that hashing rate anymore without just going to a linux environment. If anyone out there has ideas, hit me up & I'll research them.

 

The 5930K on the other hand experienced a completely different problem! All a sudden the Titan didn't want to stay mining, it would just randomly after 3-4-5 hours turn off, which then leaves afterburner with some jacked settings which confuses the piss out of the drivers, and system stability goes down. Given the only thing we've changed here was the meaty cache overclock I went into the bios and actually decreased the system agent voltage as I was fearing it might have been a bit to high. This did not fix the issue but it not harm the system either so anytime you can turn down voltage, that's pretty solid. I was watching afterburner for a while and noticed it had naturally turned it's boost clock down another 25 mhz on it's own after the cache overclock. I went ahead and did both, boosting it another 25 & dropping the overclock by 25 so I could observe the system's characteristics each way. +25 and the miner would close, so driving it harder was not the key here. -25 and it's never crashed since then. Showing that at the higher overclocks we've got going on, the titan is responding by saying, nope, things where fine but now they're a little different. The funny thing is, the only difference I'm seeing in this trade off is around 0.5 H/s. I mean, yeah I could've went in and played with my bios and tried juicing things some more until it stayed at it's previous setting but we've got a stable double 4.5ghz overclock going on here, losing 25 mhz of overclock on a Titan that's off in the land of "what the fuck is going on" really doesn't mean anything when placed next to achieving long term stability. I mean really, it's 0.5 H/s. Even a guy like me who just wrote a damn book on getting into XMR isn't sweating half a hash a second in this type of sitch.

 

Okay so there you have it. I hope how I handled those issues & how I've handled this thread in general helps some of you out there. I'd love to hear that some of these more specific tech oriented tricks where used to help somebody else out there improve their own mining efforts. If anything, I hope this thread serves as like an "Introduction to Mining" guide or maybe more an "XMR Mining 101" micro blog for new comers or even outsiders to better understand the struggles, background activities, and in general knowledge you might consider useful to have a grasp on before moving into this industry. I'm not even making more then $150-200 a month with my play toys here, imagine real miners stacking rooms of gpus or high end server farms with outdoor water tower cooling that's fins get automatically sprayed. I'd love to set that type of stuff up and play around all day, but think about the technical knowledge required. Sometimes I'm very conflicted while thinking about the mining craze. It has the image that people all over the place are just slapping their gaming boxes up on the pools and making millions, no technical skill required, or effort really! Though look what I've just created, if this thread's wall of text factor hit any harder your eyes would be red! (that's a post legalization cannabis joke, was cached but I still tried) When I was at the surplus store there was a kid in there who had his arms full of $2-3 graphics cards from waaaay back, I said jokingly, "what are you mining or something?" and he said "yeah man uhh 2-3 bucks a pop, going to earn me a free beer every once in a while" and he walked off. These where all different types of cards, a bunch of PCI, a ton of AGP, a couple PCI Express, fuck he might have had an ISA in there that wasn't even video related, this guy had no idea wtf he had. He walked right past the FirePro cards in the glass case that where $7 each and bought out like twenty cards that are tits useless. I mean this is my competition? No, this is a common misunderstanding, yet another public relations problem the crypto currency industry faces, specifically related to the mining side of it.

 

Okay okay I'm sure nobody came here to read my thoughts on mining PR issues. Just going to keep it semi-technical, long winded, & as educational as possible. That's how you do it. B|

Link to post
Share on other sites
Posted · Original PosterOP

You Stuck Your Hacker In My Mining Pool:

 

Where's all the profits guys?! Remember a couple posts ago where I was talking about how sometimes things go wrong that are simply out of your control? My pool I'm connected to, along with a bunch of others, have been under constant attack from DDOS to customized exploits. The difficulty curve on XMR has grown alot in a very short period of time as it's achieved popularity. This means that massive mining farms that are now coping with the new difficulty are resorting to other techniques for achieving profits. At one point I saw the global mining rate had tanked and looking for explanation basically went to the chat rooms and forums for the pool operators. The big pools, that charge a much higher fee for operation are chugging along just fine, actually, making a killing because the climbing difficulty curve just smoothed out as all the other pools around them stopped operating. Not only that, it's a domino effect, the longer they can keep other pools from working the more and more miners will swap fully over to the larger pools in order to continue making money, while just eating the large fees they're paying to the pool owners.

 

So if you can drain a pool of it's money, equipment cycles, admin time, and finally the actual miners themselves who fundamentally build the pool then you'll crash them. It brings alot of control, profits, and dominance across XMR back into the hands of the few. When I first started getting into this and started this thread in order to hopefully inspire other people to jump into mining, the pools where all under DDOS. There's a note on the front page of US XMR about it but you can find much better discussion elsewhere. Then as of the last day or so, there's an exploit ramping the actual pool server software to 100% CPU usage without needing all the machines targeting them like in a ddos. This is actually working great because admins are racking their brains out trying to figure out how it's happening meanwhile all of us miners have boxes sitting around just throwing errors because even if it can connect to the pool, it's to busy to give us any work or for us to return work. Don't underestimate the volatility of working with crypto currency. Three hours into this new attack people where already calling the smaller pools dead and a waste of everybodies time to continue on them, citing the ddos attacks before these as proof people needed to step up to bigger and badder networked pools that can handle such attacks. No doubt most of this was just posted by the people actually doing the attack themselves as a way of getting more damage out of their efforts, but it works. 

 

Keep in mind, I'm some guy with two single gpu desktops, a laptop, and a micro station. To other people this is a huuuge business and cheap darkweb hackers are a damn effective tool. What are all these pool operators around the world going to do? Trace them across the darknet and then sue them in real life? No you just fix the exploits, beef up your gear, prepare yourself as much as you can, an keep going. Which means this is problem that I'm predicting won't go away anytime soon. Which out of my short time playing around in the industry has actually resulted in a very considerable amount of down time. If I had more gear or even a light investment into mining I'd have left my pool minutes after the attacks started. That's what makes them so effective, people who do this for a living or who are serious about it just go "oh well that pool/coin sucks, on to the next one so I can keep my money train going" which leaves smaller pool operators in a hell of a spot. I've been reading what Frost Monkey the admin of the US XMR pool has been going through and it's a nightmare when you have a bunch of people on the other side of the attack going "hey what's going on, why aren't we making money, hey tell me stuff, like now or i'm leaving, okay i'm gone, peace out bitches" all before anyone really has time to even read it over the more serious issues that are occurring. Anyways, there's some perspective for you this day on the mining industry! Everybody go thank and tip their pool operators for keeping things going for us! :D

Link to post
Share on other sites
Posted · Original PosterOP

Hash Rates Via Pool Numbers:

 

I got a question over steam today asking to show off my actual hash rate from the pool software's history graph. I'd be more then happy too! Here in the following two screenshots you can see the peaks I mine at night when nobody is on the machines & you can see the lows during the day as people rotate on & off of them. :)

 

Peak From Last Night:

 

mI2ll34.png


Low From Today:

 

ausU1lI.png

Link to post
Share on other sites
Posted · Original PosterOP

Gaining 100+ H/s Through Prefetch Optimization:

 

I think on the developer forums I might have misread what a guy was getting at the other day because now I'm seeing gains through playing with the prefetch settings inside XMR-STAK-CPU. Previously I've said that the 5930K gained 20 H/s changing everything to "low_power_mode" "true", though it only worked on this one box, the rest of them did not see a gain but infact a loss by changing that. Though tonight while playing around I found that across all my mining rigs, no exceptions, it was more valuable to actually change all the threads to "low_power_mode" "false" then go into the "no_prefetch" settings and change it from the default of "true" to "false" across all the real cores while leaving it as "true" on the hypers.

 

This actually gives a huge performance gain! I got around a 25 H/s gain across all the machines! For instance, the 5820K went from 363.5 to 388.8 or a 25.3 H/s improvement! What happens now when I watch the threads is the hypers get a much lower hash rate but the original cores go higher, enough that we see a net gain through the trade. I had a good friend of mine duplicate this on his machines and he had almost the exact same results. We found this very interesting. Anyways thought I would share the discovery so other people could also test these results on their rigs. B|

 

Example Of Config File Post Change:

bOiTgMH.png

Link to post
Share on other sites
Posted · Original PosterOP

Building a Mining Load Balancer & Proxy through Virtualization w/ Hyper-V:

 

I've been trying to give you guys little factoids or interesting nuggets every day over the last two weeks, though there's a limited amount of expansion I can do with the amount of equipment I actually own. Without starting a pool, working for one, or finding a job somewhere in the industry what's really left is just buying more equipment & configuring it in new ways. Which is why I've now setup XMR-Node-Proxy which; "allows you to manage and direct your workers in an efficient manner. The proxy will provide several improvements over currently available solutions like atrides proxy or HAproxy deployments. Biggest improvement will be without a doubt the ability to split the requested work to individual miners. That means unlike other solutions, it will appear to the pool as a single miner requesting work at normal rate, but at the cumulative difficulty the miners connected to the backend are capable of. This not only potentially reduces the load on the pool servers dramatically, but also allows for more efficient mining. First tests have shown a dramatic increase of hashrate especially with low hashrate miners."

 

Everything I've read about this says it's designed to be used with like 10+ low end mining threads that you need to boost into one acceptable thread. I may only have six threads here (2 gpu + 4 cpu) but I really love this idea completely. So I grabbed a copy of Ubuntu 16.04 LTS minimal install an off to the races I went. Now there's decent guides out there for building this into a free or $5 a year amazon cloud but I really dislike getting away from my own gear that I have access to here locally. Soooo I thought you know I have that single hyper thread (not even a real core, lol) left that is running Windows 10 Pro, a VPN, Steam, and generally a user throughout the day using a browser or listening to music. That really only takes about half of that hyper thread.. can we sneak an entire linux OS ontop of that in the half a hyper thread remaining? In short, yes. Yes we can. :ph34r: 

 

So I opened up Hyper-V Manager, started a new machine up, gave it one core, 10gigs of ram, and created a network switch that I dedicated the second gigabit adapter on my motherboard to specifically. Ran another cat6 cable over from the router & plugged it in, which after a pretty lengthy install for a basic linux box was downloading all the newest updates and installing the mining proxy. What I really like about this situation is I now have a single status screen that I can watch which will tell me my current global hash rate. It also allows me to manually dial in minimum and maximum difficulty rates that I'm wanting to push across my devices. I can swap pools automatically, open different ports for different pools for different difficulty levels, I mean there's some neat shit going on here. They ask for a 1% donation rate themselves, so between the 2% on your CPUs, 1% on your GPUs, 0.5-1.5% pool fee, this is yet another 1% to eat. Though it's pretty cool software, so it's got that going on for it. ;)

 

I spent a long time trying to get it to work with my current pool that I've been mining on for the last two weeks but nothing I did would jive. All I'd get is a shitload of errors and problems. I swapped over to the supportXMR pool and it worked first try. I took some screenshots along the way though I lost one really badass multiple day uptime shot with steam in the background installing a game as well. I needed to do some maintenance on this water loop & had taken the screenshot, pasted it into irfanview, then forgotten to actually save it before shutting down. Sigh. Anyways, I really wanted to show off just how much you can get away with on a single hyper thread!! So in another one of the pics I even dragged my mp3 player over to show off how much was being done on thread 11 without any issues, stalls, or even hiccups. (both to the mining and my general use of the machine) System stability is always the greatest importance, never forget that when setting up your machines and getting into those overclocks! 

 

Installing Ubuntu While Mining & Multitasking Across a Single Thread:

lxaoFRn.png


Up & Running, a Single Status Screen To Rule Them All:

XjzU3m8.png

Link to post
Share on other sites
Posted · Original PosterOP

Optimal Load Balancer Configuration Woes + Gaining Hash Rate Through Difficulty Curving: (aka: setting up speed racer mining)

 

Well with this morning's stunning market rates I'm sure everybody is interested in picking up some new ideas on improving your hash rate. Thankfully I've been spending an embarrassing amount of time playing around with XMR-Node-Proxy and finally have alot to say about it. Now once again, I want to state, I'm not a programmer here, I don't understand mining at some intense level, and what you're getting is a guy utilizing his available time to fumble fuck results into place. Anyone more technically capable of laying down a proper and I do mean in-depth understanding of everything going on is MORE THEN WELCOME TO (and I do encourage it) come in and lay down some major information on not only what I'm saying but their own optimizations. What I really care about is bringing you all empirical data showing success, proving their is optimization to be had. Honestly for most of the last five days I've been doubting that was the case with my small amount of power I have vs. what XMR-Node-Proxy was designed for. Though I can finally say, yes, it is worth your time using, even on the home mining front!

 

"Step 1" so to speak would be figuring out your load balancer's "shareTargetTime" optimal setting for the port difficulty you're connecting too. This is done by dividing the difficulty rating by your peak hash rate across your equipment. That's essentially what it considers to be a broad but semi-accurate starting value for the load balancing effect to start working with. EXAMPLE: You connect to a port of 25000 difficulty, you hash at 1.3 KH/s across your home, 25000/1300=19.2308 shareTargetTime. Through doing this and just connecting my entire home's mining operation to the balancer I saw a peak hate rate (reported on pool and verified) of 3.8 KH/s though it would still ball out low to 2 KH/s at times. So the high's where better, but the bouncing between the peaks was still rough & very much present all the time. The stabilization I was hoping for was nonexistent. The problem you'll run into as soon as you look at your default config file is it's designed to be setup where you can open multiple difficulty ports on your own proxy & it'll balance the unit it's grabbing from the pool based on what you connect. Well the question is then, like, there's only one "shareTargetTime" so which port and which difficulty do I optimize my timing for? 

 

You can't. So some of you might do what I did for a couple days, which was really dumb, an that was try every possible combination of difficulty changing on my side, then stacking different share times on top of whatever seemed to make a difference after 15 minutes of being left alone. (this is where the hours and hours and hours have been wasted) Also let me step back a second in case somebody missed the boat on finding what type of hash rate to expect to even type into your load balancer in the first place. Fire up WHATEVER mining program you like, as you can read I'm big on using STAK's stuff right now. Make all your overclocks, system tweaks, etc. there, before you ever even have the idea of XMR-Node-Proxy in your mind. Once you've got a setup that is stable 24/7, no crashes, no issues, look at your peak & 15 minute average hash rates. Write them down, add them up across your gear, now you have your normal rate to expect (the 15 min average) and the peak your gear is going to pull when everything's coming up milhouse.

 

Okay so now back on track, you've got a load balancer that is giving you new high peaks but brutal lows, and you can't figure out how to optimize your shareTargetTime for the different difficulty ports you're setting up. Well from what I can find online, most people do not care about it. They leave it at the default setting of 15 or even increase it to 30 depending on the latency timing of all their remote miners who are connecting to the proxy. Though as I'm complaining about, the peaks to valley ratio sucks doing this on my personal gear. I think it's because it was designed to run 10-2500 low end cpu connections and get a better rate through distributing the work load better. Not four mid range to high end CPUs & two graphics cards, one being an unreasonable beast like the titan. It throws the whole balancing element out of wack, or atleast, it appears to me it does, I'm guessing there is some trouble always lining up the process for this unique split. Solution? "Step 2" is creating multiple load balancers depending on how far off your gear is from each other. I've simply made two balancers running, one for my CPUs, another for my GPUs. Then the difficulty & port settings are stripped down to a single port, a single difficulty, and a single shareTargetTime for each load balancer, no flipping around between them, no splitting loads between pools or ports, just straight up one road back and forth.

 

To do this, go into your hyper-ubuntu, open up terminal where you run pm2 monit to watch it work, hit ctrl-c to close it if you still have it up and running. Then type "pm2 stop proxy", followed by "pm2 delete proxy" to get rid of your current running setup. Then, reconfigure your config.json file to reflect the equipment you're trying to separate out. (cpus or gpus in this case) Once you've got a single port config ready to go head back to your terminal and repeat the process you did for firing it up the first time. This is "pm2 start proxy.js --name=NAME OF YOUR FARM --log-date-format="DD-MM-YYYY HH:mm Z"" for anyone who might currently have some wings like I do. Though you can see the new "NAME OF YOUR FARM" part. Here is where you're going to want to say "PimpDaddy_GPUs_Only" or whatever you like. It's gotta be different then the next one we're setting up though. Follow this up with the same old "pm2 save" afterwards to make sure we're good. Next copy and past the xmr-node-proxy folder in it's same location, it'll give you "xmr-node-proxy (copy)" as a folder. You can rename this "xmr-node-proxy2" or whatever you like. Now go into this folder and change the config.json to reflect your other shareTargetTime you'd like to achieve with the other side of the now separated hardware. Once you've got it ready to go head back to your terminal and "cd ~/xmr-node-proxy2/" into your new folder. (obviously, change name to reflect whatever you named it). Here we repeat the same process from above but give it a different name, so another "pm2 start proxy.js --name=NAME OF YOUR SECOND FARM --log-date-format="DD-MM-YYYY HH:mm Z"" followed by another "pm2 save".

 

Boom, type "pm2 monit" to get your back into your basic monitoring screen & now you'll have two load balancers running but reporting to the same global logs screen. You can see this in my screenshots below. Which leads us to "Step 3" which is another very long round of optimization of your two load balancers to achieve a sustained mining rate higher then what your gear would do on it's own without all this fuss. So I'm sure you're asking, how is this done? Well now that you've got two balancers running on your half a hyper thread (lulz) you can watch the stats popping up inside pm2 monit's dashboard. Once it's dialed in (let it run for 15-30-60 min tests) on an average difficulty that is achieving your equipments hash rate pre-loadbalancer look at what it's working with for difficulty across your machines. Go back to your actual mining software prompts and watch for it's difficulty to change. When you look at all your machines you'll see the lowest difficulty and the highest that it's using to achieve that average in the load balancer's dashboard. Then you can go into your config files and change the min and max difficulty to be a couple hundred in either direction. With your advertised port difficulty to your machines starting where the load balancer has finally taken you after letting it sit for a while figuring out what's best. Now you can fire your load balancers up again & see that it's taking your gear almost right to the hash rates you where getting an hour into your testing instead of the way off hash rates you got the first time you fired it up. (pre-dialing in phase)

 

The next step is really, and I'm sorry I don't have a better explanation or technical breakdown for you, but is simply playing with your shareTargetTime, VERY SLIGHTLY, and letting it sit for 15-30-60 min stretches while you record the results. Eventually you'll see going one way will help, one way will hurt, and going to far either way will wreck your rates. Finally you'll get to a place where your gear is actually running above the hash rate it would achieve on it's own through the load balancers act of splitting blocks down into individual segments for your gear to rip through. That's called success baby and we all fucking love it. Though you're going to notice the same peak to valley system happening still, even post all this nonsense. That's the nature of the beast with mining, but what you're trying to do here is pull the valley's up to form a new higher minimum hash rate over time & to increase the peaks, climbing higher & faster through this splitting of the work. Let's talk some real numbers, I'm still testing many theories I have on how this all works, I could've held back this post for another week easily. (but I wanted to get something out there for people to absorb and possibly bounce ideas back at me) I've currently seeing a minimum hash rate improvement of 500 H/s average during daily rate with people flipping on and off the machines. The dips use to shoot down to 2 KH/s at the worse times & now they only go down to around 2.5 KH/s during the falls. (once again, normally, on average, we're not talking always, it's not an absolute) While my peaks, they've gone mad. I hit 3.3 KH/s alot more often then I use to and I even saw a peak of 3.8 KH/s yesterday while I was playing around. (though I've not been able to reproduce this, ever) My equipment does a 15 min combined average of 2.65-2.7 KH/s when it's not on the load balancer so the objective is to tweak it to do better then that across the same 15-30-60m measurements. 

 

This also has a couple nice benefits, example would be when your miners go to their donation cycle. Instead of just dropping the work all together, then reconnecting to it later, or just getting new work later, the mining you're doing doesn't stop as the other miners connected are going to immediately shift over to pickup the leftover work. So your average goes down a little bit on the charts (dips) but you quickly recover when whatever miner left, reconnects to your node again. It's fairly notable on the charts when the Titan X (P) drops off the GPU load balancer and just leaves the 1080 hanging yet the average hash over the day is still higher then just letting the two gpu's do their own thing connecting to the pool directly. Which after you've finally done all this, if you're like me, just won't be enough and you won't stop tinkering around. Which takes us to "Step 4" noting that when you play around with the settings of the load balancer & get some craaazy results, they do not report to your mining pool accurately at all! So example, I can get this whole place up to a reported 8 KH/s in the load balancer. All the machines are working hard as ever, heat pouring out of them, energy getting used, nothing changes on my mining clients, but the load balancer is going bonkers. (try dialing in high difficulties) THIS DOES NOT MEAN your pool is accepting all those results, or any of them. Often when I get crazy high numbers in the load balancer I'm getting terrible numbers reported as accepted by the actual pool. Given those are the people who pay you, if they're not counting it, it doesn't matter.

 

So I just open up my account inside my pool & watch the double load balancers report in there. I only use their graphs, their charts, their reports, to verify if what I'm doing in the load balancers makes any dick of a difference at all. Because the numbers in the load balancer are either 1.) lying 2.) hashing an incredible fuck rate of bad or unsuable data. You absolutely have to use your pool as your measurement for success. I basically just carry my phone around with me everywhere and every time I report back a block it tells me my current hash rate and my averages. I can tell pretty quickly if I've missed the mark with some radical setting I'm trying out or not but I like to let them run for a while to make sure it's not just some fluke. (proper testing time guys, let them run) So then you gotta be asking me, what has made a difference? Well it's speed racer low difficulty all day. I don't know if that's because my whole place is wired for gig and I have actual gig internet with an average latency of 5-10ms but for me, SPEED RACER ALL DAY. I've got my shareTargetTime's cranked down to 1.5 & 3 seconds.  o_O  They crank ultra low difficulty work all day but they fly through it just constantly spamming results back and forth between the pool & my load balancers. This seems to do about the best of everything I want, it minimizes dips when they occur, it brings the lower floor (valleys) up a considerable amount, & it tends to consistently provide hash rates that are higher then my gear does on it's own without the balancers. PERSPECTIVE, my titan was pulling like 50,000 difficulty work off the pool all the time on it's own, and crunching through it no problem. (though it took some time) Though now at this moment of typing it's speed racer burning through 1450 difficulty results like a crack addict with a brand new eight ball. Shit is on fire the console log is just a spam of results getting reported.

 

THIS is what is getting me better averages then anything else I've tried. The global logs in the load balancer dashboard is actually accurately reporting hash rates that are right next to what I'd get without the balancers (so I know that's working) but the rate I'm getting from the pool on my results is better overall. So once again, don't rely on the numbers inside your load balancer for anything but a basic check to see if they're in the correct ball park at all. Fine tuning needs to be done from your pool back to your balancers, not your balancers to the pool. Here's another BIG TIP, no not make a hyper-v checkpoint in the middle of mining & then revert to it later on with your internet still enabled. It'll pick back up where it was when you saved (obviously), error the fuck out for a bit, throw some flags, and you'll get an invalid share from the pool over your previous work. I've never had an invalid share in my short but now gaining some time mining career / life, so I was really upset when I saw one pop up. So yeah, do not do that, going to cost you a share! (currently on this new pool, I have 83,982 valid shares to my 1 invalid share, but still, pisses me off to see the big red X on the stats screen with a 1 next to it)

 

I'm not giving up yet, I'm convinced I can fine tune this thing myself into better hash rates. I think I've just not done a proper job testing this in a scientific manner nor applied enough time across each measurement to really find optimal results over luck & the natural mining cycle. I'll be sure to let you guys know what I find in the future as I'm clearly not going to stop playing around with it until I'm pretty dead nuts that I'm getting the best out of my gear that's possible. Though let's keep in mind, this software isn't even developed for this use. From the author himself, it's not worth running if you don't have atleast 10 miners connected to it. Though I'd like to say I've provided some interesting results showing it is actually still handy to home miners like myself. 99% of people probably don't give two shits for how long this explanation has been & the required dialing in time. Though I know there's going to be somebody out there who is going to read this and be like FUCK YEAH IMPROVED RESULTS! and will join me on the load balancer quest for optimization and efficiency.  xD


Load Balancer Failure! AKA: when things go wrong & you're not around:

aYZd0qy.png


Double Balancers sharing info in the Global Logs:

ZYY6nM1.png


Success! Double Virtual Balancers running inside half a Hyper thread providing better hash rates then the gear does alone:

WoP277e.png

Link to post
Share on other sites
Posted · Original PosterOP

Bro, you been mining wrong for weeks:

 

So this morning I joined the SupportXMR chat room, dropped the link to this thread, and said, hey new comers this might be helpful / if you're an expert please come school me. At which within about three minutes, the admin endor came in & said "well you've got your mining configs all jacked up so that's a good place to start". Which after this much time made me go lol wut. Then he gave me a very polite & technical breakdown of why I'm a god damn noob. To sum-up what I've learned this morning: each mining thread you launch is going to consume 2MB of cache. My i7-5930K has 15MB L3 Cache, so it can handle 7 mining threads before only having a single meg left. Which isn't enough for another mining thread without gimping it, but could be left to help out with the OS overhead / load balancers. Which means, I've been getting poor hash rates out of my CPUs this entire time because they're dicking each other for cache no matter how fast I've got them going. It also means, that I see vastly improved hash rates using 7 threads (6c/1h) then I do using 11 threads with my previous (6c/5h) setup. This is running them all in "low_power_mode" : false & "no_prefetch" : false. We discussed the differences in utilizing the cache through the low_power_mode since it doubles down on it. This would be handy if you needed to save threads for other activities but still wanted to use the cache. Though big picture time, when people sit down to use these machines during the day they now have five open threads instead of whatever is left of the remaining one post double load balancers. This has been a wonderful morning for my own advancement in understanding & has shown immediate positive results. So yeah, get educated. xD

 

I'd like to toss a big thank you out to endor for all his time this morning, plus all the other folks who helped out inside https://supportxmr.chatango.com. Very friendly place, great community, with an active chat! B|

Link to post
Share on other sites
Posted · Original PosterOP

Questions Round II? III? Round Something! How do I properly setup my GPU inside XMR-STAK-NVIDIA?

 

Well after getting wrecked this morning on my CPU config this is a great question. (asked via steam) So when you look at your config file you're going to see; index, threads, blocks, bfactor, bsleep, & affine_to_cpu. Now the first one index, that's simply which GPU you're configuring. Also the last one, affine_to_cpu you'll want to turn on if you're also CPU mining on the same machine. Now it's the next four inbetween that are fun. To figure out your threads, what you're actually looking for is your graphics card SM value or the number of "streaming multiprocessors" it has. Now this can be done simply by going to google and typing in your card and bam you're going to find reviews talking about how many SMs it's packing. OR if you want to be technical and not trust people, then go to google and search for your graphics card's "block diagram" where you can physically count them. Which is not a bad idea because outside of testing random variables (which I do actually suggest next) you're going to need this to find your card's specific "block" value now that you know your thread value. This is done by looking at your block diagram & counting the clusters per streaming multiprocessor. Though this is a bitch & you can type in 60, 100, 150, 200, etc. and quickly find a few things. 1.) what crashes the miner asap 2.) what gives you the highest 10 second hash rate for the ones that do fire up. It'll take no time at all to dial this in & the real number has never once matched my most optimal setting. So a little bit of experimentation goes a long way here! ALSO NOTE, this will greatly change how your card is using power during the mining. Kick open afterburner or your favorite tool & watch your power consumption charts as you change these & play around. NEATO! :D

 

The next two settings, are simply for how much system lag you want if you're using an OS with a GUI installed like Windows. Bfactor of 8 and bsleep of 100 is the recommended settings for keeping a somewhat usable machine. Though I think alot of people would prefer a much more responsive system & would recommend dialing bfactor back to 10, or even more, until you find the level of responsiveness you can tolerate while it mines in the background. In my experience, bfactor is much more taxing then bsleep, all my machines can run bsleep at 0 and it's the same as if I run it at 100. Though that's just my equipment, so your mileage may vary. I know on my buddies 3770K/660 system he runs bfactor of 14, sleep of 100, before it feels normal to his family. Anyways, that's how you get the proper settings for your graphics cards for XMR-STAK-NVIDIA. Which is handy information to have & I hope that helped some of you out there. :)

Link to post
Share on other sites

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


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]

RAINBOWS!!!

 

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

Link to post
Share on other sites
Posted · Original PosterOP

Power Usage Vs. Temperature:

 

Hey everybody! I've been sitting in the SupportXMR chatango pretty much 24/7 chatting up the good folks over there while attempting to help other newcomers as well. Gotta say once again very good community there if you need a hand & can be a polite individual. Anyways, I grabbed a screenshot this morning while I was enjoying my coffee that I found really interesting & thought I'd share. So once again on the "nobody would buy this gear for mining" front, I'm running this Titan across an EKWB 480mm radiator from their extreme line that is extra thick. It has eight vardar 3K's in push/pull on it & an extra high performance pump to keep those flow rates up. That's how I've been posting these insane temp & overclock screenshots for you all. Though I had left my window open over the night to let out the mining heat, and as it's getting towards late September it got chilly last night. So much so I was sipping my coffee & noticed as I was going through my morning checklist that my titan was mining at 27C!! That's uuhh 80F!! :ph34r: 

 

I could've turned up the overclocks at this point but given it's already daylight out it's just going to get warmer. Though the thing that caught my eye was the power consumption dropped by 3% on the baseline from 60% while mining to 57%. That's ridiculous & very very cool to notice. If only I was up in the middle of the night when it got the coldest out, I gotta wonder what it was crunching at in terms of efficiency. I'll be sure to keep an eye on this in the future & report back any other interesting observations. Also I want to note, have more posts coming, just want to make sure I'm technically correct before misleading anyone at all. I worry anytime I advise people in this field that I've not got the credentials behind my name. So taking a bit of a gap between posts to learn myself, get better educated, and bring back more valuable content for everybody. :)

 

3% Power Reduction while Mining by going from 35C to 27C:

6BVEkxw.png

Link to post
Share on other sites
Posted · Original PosterOP

Gaining Additional Hash Rate & Stability Through Affinity Configuration:

 

So I was sitting in the Support XMR chat room lurking as people discussed various tricks they where using to mine on a range of different equipment. I was also chatting over steam with an Arsian who has been bouncing configs back and forth with me through this whole dive into mining. I was telling him this morning how it was really interesting you could make a single core use twice the cache through low_power_mode but I couldn't double dip on my threads per cached miner. If you haven't laughed already don't feel bad because it took me a minute + reading other peoples random chat to figure out I'm a big dumbass. Of course I can do that, though I have to trust my operating system to see how well it handles it. This would be done by removing affinity to any specific thread. Which mind you, most the guides and notes I've found have advised against doing. Even going as far as saying you need to affine directly per mining thread if you where going to toss gpus into the mix as well since they don't play nicely.

 

PERSPECTIVE, if you've not joined along down the entire journey of this thread so far. I originally was splitting my CPU's L3 cache among all my threads but one left for OS overhead, meaning for example on my i7-5930k I was using 15 MB of L3 across 11 4.6ghz threads. I even configured it a bunch of different ways and tested them extensively as you can read above. Then I got educated that given each thread required 2MB of L3 you would benefit from only running however many you can stack fully into your L3. Which in this case would be 15/2=7 fully enabled mining threads. Which is great because I did actually see an improvement in hash rate configuring the CPU correctly. (lulz gents) It also left five threads open for people to use as they sat down at the computer throughout the day instead of just being left with one. Good deal all around. Though with the thought to remove affinity that would hopefully spread the same mining load using 14MB of L3 across 12 4.6ghz threads instead of the 7 I'm currently limiting them too. Would it work this way? I sure as fuck hoped so but I wasn't sure Windows was up to the task.

 

In short? Yes!! It seems to work that way atleast within my limited testing with Windows 10 Pro and Server 2012 R2. So, once again, this will divide your mining load across your leftover cores which will shred your machine. Just like when I was originally running it into the ground with to many mining threads. The difference is the L3 cache is now getting used properly this time around. Well what happens? My CPU mining rates are instantly almost up to their peaks and respond to userloads on the machines extremely quickly. The latency between bad performance when they're watching youtube and good performance the second they stop and are reading their email is much much shorter then before. I mean clearly you can go into resource monitor and watch it working all your threads again, it's a big difference. Though this also works for the video card, removing it's affinity and letting it get managed by windows is also greatly enhancing their response timing to outside stress factors. (though I only have single gpu boxes here, I have no idea what this would do across a quad or more gpu setup)

 

Both my i7-5820K @ 4.5ghz and my i7-5930K @ 4.6ghz are pulling 500+ H/s in this setup using seven mining threads utilizing 14MB of the cache but spread across the twelve threads of the processor. I mean huge thumbs up on the continued tweaking of the same exact gear from when I started. I'm really learning alot here, most of which should have been obvious, though if anything this thread can stand as a proving grounds for others to come & not feel as bad. If I'm making these mistakes there must be others out there doing the same. I also want to note that I fixed a configuration error today when I was testing this concept. My laptop while being the same i7-3770 series as the mini workstation actually has 2MB less L3, so I've been over working it this entire time still!! I dropped it to three mining threads instead of the four I had turned on & it's hashing rate immediately went up again, yeah because surprise like the others it's now properly configured. Then once again, flipping it's three mining threads to be spread across it's eight total threads through lack of affinity settings saw another improvement in hashrate. So yes, you can do that, you always could, I'm just a noob who looked right past it in the settings.

 

 :o  ...  :|  ...  xD  ...  B|

Link to post
Share on other sites
Posted · Original PosterOP

Getting Your Feet Wet In Monero, Starter Edition:

 

Okay everybody so there's alot of sources I've been using to get through my introduction to Monero. I thought instead of just tossing helpful advice in this thread I'd also give you guys a breakdown of the places I've been using. Keep in mind there is ALOT MORE OUT THERE then this, but these are the places & pieces of software I particularly like or have found useful. :)

 

Official Website: https://getmonero.org/
 - This is the starting grounds for all things Monero. If you want to learn more, find software, or just see who's behind things.

 

Official Monero Forums: https://forum.getmonero.org/
 - The natural extension of the above website, these are the official community forums for everything from PR to Dev.

 

Official Offline Wallet Installers: https://getmonero.org/downloads/
 - Download a wallet to receive/hold/spend your XMR from that is stored & protected locally on your machine.

 

3rd Party Online Wallet: https://mymonero.com/
 - An online only wallet to receive/hold/spend your XMR from that is stored & protected in the cloud.


Monero Coin & Market News: https://www.monero.how/
 - This is a great source of news for all things Monero! Recommended for getting your dive into the culture on.

 

XMR Profitability Calculator: http://whattomine.com/coins/101-xmr-cryptonight
 - Put in your equipment's hash rate, your local power cost, and what your pool fees are to see hour, day, week, month, & year long estimations of profits.

 

XMR Hardware Benchmarks: http://monerobenchmarks.info/
 - Not sure what you'll earn? Here's a wide list of broadly configured equipment hashing Monero.

 

PPLNS Explanation: http://give-me-coins.com/support/faq/what-is-pplns/
 - Thinking of joining a pool instead of solo mining? This explains how you get paid vs everybody else in the pool.

 

XMR Pool List: http://moneropools.com/
 - A giant world wide list of pools you could join, what they charge, how they work, & where to find them.

 

Support XMR: https://supportxmr.com/
 - This is the pool I'm currently running in & I mention it out of respect because it's extremely well maintained with intelligent & helpful admins.

 

SupportXMR Droid App: https://play.google.com/store/apps/details?id=de.blindjerobine.supportxmrpool&hl=en
 - If you happen to join our pool then this is a cool android application you can download to watch your mining & the pool from afar via your phone.

 

Droid Market Management: https://www.blockfolio.com/
 - If you're larger then one pool or really into day trading across multiple coins this is another great application to watch all the markets & your investments.

 

XMR-STAK-CPU: https://github.com/fireice-uk/xmr-stak-cpu
 - This is the leading CPU mining program, it's the one I've been using throughout this entire thread so far.

 

Windows Compile Notes: https://github.com/fireice-uk/xmr-stak-cpu/blob/master/WINCOMPILE.md
 - If you're looking to get onto the Dev branch or change variables with a self compile, here's some great instructions!

 

XMR-STAK-Dependencies: https://github.com/fireice-uk/xmr-stak-dep
 - You're going to need these as well if you're looking to start compiling your own versions.

 

XMR-STAK-AMD: https://github.com/fireice-uk/xmr-stak-amd
 - This is the leading AMD based GPU mining program, it's made by the developer of the one I've been using throughout this entire thread so far.

 

XMR-STAK-Nvidia: https://github.com/fireice-uk/xmr-stak-nvidia
 - This is the leading Nvidia GPU mining program, it's the one I've been using throughout this entire thread so far.
 
 Nvidia Config Breakdown: https://drive.google.com/file/d/0B6LV7PJ6boeBVGRZWTdLRkZKeUU/view
- This is a Step By Step breakdown of how to properly config your Nvidia graphics cards inside STAK w/ pics.

 

Windows Compile Notes: https://github.com/fireice-uk/xmr-stak-nvidia/blob/af2ac2dd064774e1e3825eb4c928490abfcb3f9b/WINCOMPILE.md
 - Same idea as above, here's some great instructions for nvidia compiles. NOTE: this process is quite a bit longer to setup then the CPU one, but just as easy so nothing to fear.

 

XMR-Node-Proxy: https://github.com/Snipa22/xmr-node-proxy
 - This is the load balancer and proxy I've been using to split workloads across my equipment & configure difficulty curving.

 

SilentXMR Win: https://gitlab.com/Elycin/SilentXMRWin
 - Monero (XMR) Wrapper to mine when a computer is idle. Will utilize STAK then claymore's as a failover.

 

Goblin's ARS Thread: https://arstechnica.com/civis/viewtopic.php?f=18&t=1396927
 - This is literally the same thread from here on Linus Tech TIps but running over at Ars Technica. (vastly different communities)

 

Goblin's LTT Thread: https://linustechtips.com/main/topic/827432-mining-monero-xmr-with-my-current-gear-lets-look-at-some-numbers/

 - This is the very thread you're reading right now if for some reason you wanted to know that.

 

Anyways that'll probably do it for helpful places to get started that don't require to much from you. Once you've got that all conquered and you've read this thread a couple times learning from my mistakes, well, you'll be a god damn coin wizard. xD

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


  • Recently Browsing   0 members

    No registered users viewing this page.


×