Advantages of k33-34 plots on cheap old used servers

Chia Plotting
1

When we built Internet, there was already microbashing. Over the 30 years, it blossomed in typical human nature to “preach about something I have absolutely no idea about, but strong desire to look smart. Any opposition to call troll or clueless.”

Study of HR IT department:

  1. all about efficiency, and price per plot - there are many crypto kids bragging with fancy HW. Simple question such as “What is your price per plot?” is followed by mental diarrhoea. If you do not know how much you spent, how do you know how much you earn?
  2. right here, right now is all you need to focus on - what is in your pocket, ready to reinvest

Our tourist, simply resalable, plotting rig spits every plot in ~19min @ 0.05$/55W/plot with capacity >78 plots/day.

BSgument #1 SIZE matters…perhaps in porno industry

From official Chia Wiki, we have found out there are no real advantages to go over k32 at the moment. We understood that larger plots are simply:

  • k33 = 2.06 x k32
  • k34 = 4.24 x k32

RESULTS of study of plotting based on this post because it really doesn’t matter. Same machine will have same multiplier no matter what.

plot min time multiplier k sizes GB k sizes multiplier advantage
k32 28.1396 108.9
k33 70.5466 2.50702213251077 224.2 2.05876951331497 -17.88%
k34 179.457 6.37738276308121 461.5 4.23783287419651 -33.55%

Result is simple, IT IS BETTER TO PLOT MORE k32 PLOTS THAT fancy >k32 plots.

We have 6 x LVM2 JBODs 100-140TB, each VG has got 1-30GB free space used for blockchain/server backups. We see no economical advantages to fill space “effectively” with ineffective plotting of >k32. We currently purchase 14.7$/TB.

BSgument #2 about “old dual CPU, high RAM DDR3” servers. We have heard kids talking about 600$ old dual CPU server with >256GB plotting ~15min/plot. We asked simple question “What is your price per plot?”

We have got typical answer “I do not know man, 450-600Wh…it is cheap initial cost…who cares” We wonder no more, why there is poverty on Earth.

If someone has got free electricity, we understand. Sadly, we paid 0.23$/kWh this year, and from next year we get lovely hike to 0.31$/kWh

We currently plot for 0.05$/plot because doesn’t exist storage system to do anything economically on such old inefficient HW. We used to have Dell T5500 dual Xeon that burnt 500Wh

We do not calculate HW costs because the used HW we bought will be resold few years later for almost same price as now.

Only consumables are energy, NVMe for hybrid plotting.

We are currently 54W/plot/0.013$/plot + latest NVMe capable of 600TWB or 1500 plots or 0.087$/plot (Popular Corsair MP 600 2 TB gives 0.03$/plot)

There are no used datacenter U2/PCIe drives available in our area, and new ones are beyond our profitable usability.

Intel Optane 900P capable of sustainable 2TBs write, still cannot beat good old Corsair MP600.

Putting storage system aside:

rig W/plot $/plot
HR 54 0.012
no idea guys 112.5-150 0.026-0.035

HR Rig:

component price $
MB B550 180
Ryzen 5950X 400
RAM 128GB 600
PSU Platinum 300
TOTAL 1480
RESELL PRICE >600

We do not feel any savings in old inefficient HW. Quicker one plots, more it all costs. We have seen some kids working on GPU plotting…allegedly 9 min record right now.

Any kind of mechanical HDD can do ~150MBs, slightly above default 1Gb ether nowadays. Each plot takes ~15min to upload.

Anything <15min plotting time needs >2.5Gb ether (while our workstation already has 2.5Gb, no server has it as standard. Our old SuperMicro mini server for farm did cost 200$) , large cache because no mechanical HDD manages to do sustained 200MBs…You will be stuck at ~10min anyway.

Bottom line is always ROI (Return On Investment). We all do it for profit.

2

You certainly do make some valid points for you situation. Everyone has goals. Yours are ROI, profit, and efficiency. Noble goals those are! Good for you… all of you! But repeated dependencies on the collective “we”, and absolutes such as “always” and others, seems to get you in a bit of a pickle here vs the common Joe or Joanne I think.

For the sake of a good argument and to end the year the sucked bad for most, some comments:

  1. “Bottom line is ROI” - something of utmost importance to you, for legit reasons, but that does not apply to everyone (or you would not be here arguing the point lol).

  2. “It’s all about efficiency” - Straight off -“HR IT” dept (link goes to some cars site - don’t understand?)- Many of us focus on things like plotting speed, farmer efficiency and longevity, simple enjoyment and certain technical achievements, etc. as their goals. And yes, efficiency may play some part in that, and people do strive for it, but it may not be of paramount importance for a variety of reasons both practical and financial.

  3. “It’s all about price per plot” - Really, this is one metric I’'ve never been concerned about. Sure it’s a nice idea, but in practice it is so vague as there are a multitude of costs and factors that make calculating this nearly nonsensical. How do you add up > time spent, electricity used, multiple hardware component cost and time/wear depreciation, just to name a few items, of the myriad of component in a plotter/farmer node over a long period, currently more than a year? Quite the project that, and can be quite variable depending on what is included in the calculation! Further, each node is unique in different ways, making comparisons justifiably inaccurate.

  4. " …how do you know how much you earn?" Apparently for you…but how much one can earn is perhaps not the 1st thought that many have…I mean…earnings can be something in the future…if you are not happy with what your ‘investment’ is making right now. So unless you sell immediately to support your node, not so much of an issue you can calculate.

  5. “Result is simple, IT IS BETTER TO PLOT MORE k32 PLOTS THAT fancy >k32 plots.” Better? Because someone says so somewhere in the Chia docs? There is, IMHO, misinformation/old information/wrong information in Chia docs spread around the internets. Do you believe some factoid because you read it on the internets (from so-called experts)? I’ve learned over the years to make up my own mind, based on my own experience, about what is my truth and what is not, regarding what I have in front of me that I experience.

  6. K33s and larger plots are not fancy whatever “fancy” plots mean to you, for starters. But, to use your guiding light, I would posit that >K32 plots are more efficient for the farmer. Less lookups, less HD wear and tear, less energy used. And the farmer’s use of plots is far, far more comprehensive, over time, than what little the plotter does…which is a once and done proposition.

  7. “Each plot takes ~15min to upload.” - Perhaps on your system, but not mine or many others Again, you use of absolutes is misguided as you are not everywhere to experience what others do. 160-220 MB/sec are my norm, for what it’s worth to tell you. And that doesn’t translate to ~15 min/plot.

  8. “Our tourist, simply resalable, plotting rig spits every plot in ~19min” That’s OK, but fairly slow actually. My k32 are on average plotted in ~ 15min. So if you make 10,000 plots, for example, saving 4min/plot is 667 hours, or almost ~28 days, almost a full month of 7day/week…24 hr/day work (of losing efficiency and money) due to you relying on that “tourist” PC over a humble workstation plotter. I can’t comment on other’s ‘old hdw’, but some have got pretty impressive results with them.

  9. “We all do it for profit.” - I get that from what you say your post. But who is the 'We" you use throughout? Is 'we" a for profit company? Your workplace? Perhaps all the numbers and calculation you make justify to whomever that you are 100% spot on. That’s a good thing for you all. But “here” you must realize that all your justification may simply not be relevant or do not apply to many, don’t you?

Happy New Year!! :sparkler: :tada: :piñata: :clinking_glasses: And hope it’s a better one this time around lol!

1 Like
3

what you are trying to say is “we do it just as hobby, and do not care about costs?”

  1. what is point of paying 30k $ for something with no returns? It is same as throwing those resources into bin. Amsterdam/Prague is better way to waste money with long term memories :stuck_out_tongue:

  2. yet, most have no idea about price/plot or profit/losses on running farm. Weird claim. We have 600TB @ 260Wh or 0.044Wh/plot or -0.01$/plot. Each plot makes 0.03$. What is your efficiency if you mentioned it? There is difference between plotting speed and efficient plotting speed. What longevity? HDDs fail, you have no control over it. NVMe wear off, you can decide how much you pay per plot. Plotting speed is interesting concept if you can keep same wattage and reduce time. Kids tend to OC CPUs, hoping they get free stuff. AMD Ryzen 3000-5000 series is most efficient around 4GHz…if you go beyond, your gains will be little at brutal energy costs. It is called 1% tuning in automotive, where one gains 1% speed for -30% efficiency.

  3. And that is your point of doom if you don’t know what you are worth per hour, or how much resources you wasted…you will always be poor. We usually work for 150$/h. We wanna know exact costs such as time/energy, HW, and ROI. What multitude of costs? It is very simple. It took at least 50’000 hours to learn all you need to know for IT. You know how much you paid for HW, and how many hours. Basic bussines 101. If you do not log your work, and resources…it is just a fun without any goals or objectives. Escape from your rat race at work. If you wasted 20k on “fast” plotting HW, that will be not possible resale. Your plot price must be 5$/plot that earns something like 0.03$/plot righ tnow. How much you earn? Of course, it may pay off if you plan to plot 1m of plots. There is zero variables, simple math and discpiline.

  4. so why you do it? have you no life that you waste 30k $ on something that will never pay off? Unless, you can predict future (as Wall Street tried for 100+ years, and failed horribly) then I understand your business plan. We use what is on table now. Speculation that XCH will be maybe 100$ in next halving is drunken dreaming or wishful thinking of Lambo-crypto-kids. I recall a guy who bought BTC for 55k, and ETH rig in the middle in frenzy…I warned him about a half year ago…“I keep it till next halving” So you lost about 50k $ on some imaginary new tech, your wife is about to kill you, and you hope if you keep your losses next 4 years, it will come back to 55k…and then to 200k as S2F BTC guy precisely predicted, and not happened ROFL We are always amazed by strategies of poor.

  5. What you are saying is that sunrise/sunset is fake, and the whole scientific world is bollocks right? We believe no one, we do analytics and decide if it makes money. It has been shared with a confused public. It is up to you whether you wanna be right, or do it right. Numbers don’t lie.

  6. You remind me my photo career - there is still people who do all “manual” because it looks cool. We do it simple, we learn limits of tech, and use it. Why waste time by ego issues? Whether farmer processes 1000 x 101GB or 480 x 224GB…Still same amount of data to hash. I doubt there is any efficiency hit. What efficiency you talking about? Only issue we see is that you waste resources on dream instead of earn more.

  7. Our uploads are “free”, already part of any system. How much did you pay for your improved speed? We have some old SuperMicro 10Gb fibre…3 cards + 30m cable…200$ gone. How do you pay for it with such a speed “improvement”? Of course we can move all faster, but plotting speed drops. How about that? Oh wait, you guys buy another NVMe to cache all plots so you can transfer everything fast. How do you pay for all this unnecessary crap with XCH <30$ ?

  8. Yes, it is slow but once again how much do you pay per plot? ROFL 0.05$/plot is 6 years on 18TB right now. With your super-mega-fast system, you pay something like 0.2$/plot??? or even more??? that is 24 years to repay just 18TB HDD (I hardly doubtmodern HDDs survive 10 years as the old ones). How do you repay your 30k$ plotting HW that is useless for most, and has zero resell value. We currently have around 7000 plots, and there is no future plans for expansion at current prices. After halving, income drops to almost zero unless price correct itself. You guys think like Chia is sure think that will last forever. It is experimental startup with big names on account. Question is, can they deliver or will they be regulated to death as BTC or various clones as dying ETH aka Ponzi 2.0

  9. “we” is rich people that work smart, efficient, getting filthy rich and also make lots of money from our investment. Apparently, your vision is to waste money on some experimental stuff with questionable future. What if next year, Chia dies. What do you do? We cut the losses, sell all stuff for at least 50%, and we got free experience/knowledge along the way…moving to another project. What will you do? We see, you have most likely paid 50k $ for HW, and it will take 30 years to repay if Chia survives, and price keeps rising to at least 50$

Weird thinking guys. Be rich, and also earn lots of money.

1 Like
4

Gee, that person has a real issue and far too much time on their hands.

5

says guy with “free” energy, and advanced math skills.

6

Lotz, lots of conjecture, I didn’t tell you much of anything. You have no idea, but please march on :hugs: You do you. It’s is the only path you know.

1 Like
7

And another one ignored!

2 Likes
8

Not sure if you have been part of the forum before with a different username or if this is your first “experience” with the…regulars or long timers. I get the gist of what you’re saying. You seem more than sensible. The only planet where ROI doesn’t matter? It’s on a planet called regret where you are so deep into debt that you have convinced yourself that you invested for some other reason than profit. I’ve banged my head against the wall many a time around here with regulars and so when I see others who struggle to have a differing view around here? I need to say something. Not that my view counts for much. I’m a pretty neutral, open minded person. I’m happy to discuss the real risk vs. reward vs. hardware. You’re bringing something different to the conversation and I appreciate that. Most everyone here hasn’t even wanted to look at their investment costs, operating costs, ROI or anything that relates to shifting gears. The common advise is throw caution to the wind, disregard ROI or resellability of hardware and just get Chia specific drives for example that nobody would want should Chia end up going south at some point.

1 Like
9

In that case consider Flexfarmer as it reduces your power use :smiley:
Also GPU plotter will likely solve the debate.

1 Like
10

General purpose OS only. We also provide nodes for other future-resistant crypto projects.

11

@MisterSavage That’s the beauty of nature. As great George Carlin used to say “Bullshit is social glue that built many civilizations.”

We do not visit forums too much, if you waste time on pointless discussion there will be no evolution of your beautiful mind. We built webpages, and let people come. Who seeks, always finds the way.

We spent first 30 years to become masters. You always owe your success to the previous generation, and it is unwritten law to share it. We build academies, but there is raging poverty for over 40 years.

You cannot do anything about it “Most people wanna be right instead of doing it right.”

Facts are right there, yet people rather stick to their “profitable” bullshit, and fiercely defend it. Since bullshit lists runs pretty quickly out, 2nd stage is mocking, taunting.

So far we have found great masterminds over there:

  1. the "free" energy guy that needs no math, because his mighty brain can do it on-the-fly while earning thousands.
  2. The “15 min wonder” that spent 50k on HW while his plotting is 15min per 150W, and it will take 30 years ROI.

And there will be more, wait once they start to flock and attack in numbers :wink:

HR Psycho department did study how to help more efficiently poor souls. Sadly, most people suffer from many fears, and they simply runaway from their problems. Common Joe McFuck believes his horrible boss steals from him, the Big Brother is suppressing shis geniality, so they try to escape this hell by trying to look smart in other fields at any costs no matter what. They used to socialise in pubs, but they have no money anymore so we built Internet for them. Every forum you vist, it is same.

We see it in every field of life inside the Matrix. And we used to be part of poor group, but refused to feed our poverty, and rebelled. Ever since we guide those who want to do it right. There is not many of them, but we are getting filthy rich together.

After all, the whole purpose of life is to be happy and live rich. It looks like most prefer the hell of rat race work…eat…whine…sleep…rinse…repeat…move risk of life to others…live happily after from social system.

Last year we met a Feminazi IT expert…before admins banned her…she told proudly everyone about fucking men, and then demanded help because her 700Mhs rig didn’t work (she proudly bragged about sher hashrate as it was something special…1st sign of poverty) Yet, she was unable to resolve simple task as why my Windows PC with switch doesn’t work. They are all geniuses, easy to spot.

12

@Chris22 GPU plotting is interesting concept once SSD storage costs kill the HDDs. It is expected somewhere by 2025, but we highly doubt that SSD price drops from usual 100$/TB down to current ~16-20$/TB of mechanical HDDs.

Nobody sells new tech, if the old tech barely starts to make money. Businesses are focused on profit, not the usual just-for-fun-kicks most do.

We have bought first pioneer in SSD storage, and it cannot copy reliably at 100MBs. Of course, it was cheapest high capacity NVMe crap back then. SATA SSD can do easily 300-500MBs, saturating 6Gbps SATA protocol.

Kioxia already offers 100TB SSD, but do you wanna pay 50k$?

It will take many HWgenerations before GPU tech reaches level of zero to none evolution, and they start to work on efficiency. Current experiments we saw is about 40W/plot for 250Wh GPU based rig. Price for rig still cannot beat the cheapest consumer rig such as HR IT demonstrated above by.

If you are the smart one, and spent lots of money on 10Gb networking. With lots of luck you will upload plots at 8min each if your mega server with expanders can do 200MBs sustainable write speed (we never saw such a HDD in range of 16$/TB we usually buy, but that is story for another time)

13

Plotting time doesn’t really matter, but storage capacity does. If I had to replot after the new format (-20% size), I may plot k33 instead of k32

1 Like
14

@Potatom and you know exactly when comes the new format, so you prepare 5-10 years in advance right? Russians used to do it as well, they had 5-10 years plans. We all know how CCCP ended up.

in 5-10 years plotting of k33 will take the same or even less time than k32 now. Numbers always tell the story. In 5-10 years, there will be >40TB HDDs, or finally replaced by SSD/future storage systems.

In 5-10 years your current space will be a joke, and easily replaced by 50-100% less storage devices. We regularly swap high efficiency/space for old storage drives because what matters is always profit/ROI right here, right now.

Just in last year, 8-12TB was replaced by 18TB. Density/wattage ratio improving all the time.

Also known as Say’s law. Efficiency is the key to progressive economy. “Planing” volatile tech industry is the same as predicting price of cryptocurrencies :wink:

it makes no business sense whatsoever.

Or you can do it crypto-kids way…mining/farming, and hoping there will be a wave of retardation after next halving, and it will somehow pay for your Kamikaze “investments”.

Visions of Lambos, houses, and easy/safe profits. Uncle Warren & Charlie must laugh so hard that their pacemakers must scream like crazy ROFL

15

Higher k-value plots have less reads. I would love to fill drives with 1 plot only, but the current plotters don’t allow for efficient plotting time of those higher k values.

It’s not a waste of time to look into making things differently than other people. That’s how you get advantage over competitors and thus better ROI for a business. But $30k isn’t enough to start taking it seriously, especially when there are hundreds of millions of dollars in drives securing the chain :slight_smile:

16

@aurelius how did you figure out higher k-plots have fewer reads?

hashing consensus of 101GB of data is for example 20 IOPS, and 1000 inodes; 224GB of data is 41IOPS/2050 inodes.

Or do you think you plot larger plots, and farmer will check only first 101GB but you get 2.05x higher rewards?

It makes no sense. Can you collaborate, and share your cpu load, iops readings, Watts/plot, please?

17

I’m not talking about the amount of data read to find the proof, I’m talking about passing the filter, and since some data is cached the disk doesn’t even get accessed unless there’s a proof to find, thus less querying.

3 Likes
18

How do you cache 600TB of random data?

19

You should re-read my reply, I never said that the plots are cached, I’m talking about disk access and you’re smart enough to understand what I mean. I would appreciate if you didn’t try to argue with me in bad faith by twisting my words.

2 Likes
20

Exactly, so how did you find out better reads?

Each plot has got 7 tables; each table contains x^k entries, where k = space plotted.

Cache is as big as queries^k.

‘top’ says 10% of 32GB is used. I am pretty sure same amount of RAM will be used for 10’000 k32 or 4’500 k33 plots or 100TB.

And same amount ionodes access will be used. Unless, you cracked Chia algo, and changed fundamental principle of cryptography.

How did you calculate less reads is still not answered. It sounds as typical “feeling tuning” we hear in our automotive department.

Usual Johny McDonald scientific proof of theory “man, I drove behind my friend. Engine has more torque, with no smoke”