<g​oodboy:matrix.org> wut dis, x.com/fluffypony/status/1957791643232289246

<g​oodboy:matrix.org> monero-project/research-lab #140

can we have a bridge to #monero-marketing?

<o​range_horizon:matrix.org> @Cindy I can't make the bridge, but i copied your suggestion for marketing and request for the bridge into #monero-marketing.

thx

<j​c1nc:matrix.org> What makes you choose IRC over Matrix? Just wondering since I’m new to this stuff

because it's more stable and easier to use?

and without 'blinking lights'

<d​iw_tim:matrix.org> Desperation from CfB. Many people are shorting Qubic, and I know someone in this chat who has made about $150k so far from shorting Qubic. Their scheme is falling apart and people are now taking advantage of Qubic.

No one cares about non-Monero shitcoins

<c​ountbleck:matrix.org> Qubic is notable for their attempt to threaten Monero, not for their attempt to spew bullshit

<c​ountbleck:matrix.org> lots of random scam tokens do that

<d​iw_tim:matrix.org> If the price falls enough, their incentives will become almost worthless, reducing the weight of their attack.

<t​orir:matrix.org> There are some mining rigs for rent that only appear to run while being rented, like this one: miningrigrentals.com/rigs/322626

<t​orir:matrix.org> You might have to check "show offline" in the main list to see them, since they are offline while not being rented. It's also not possible to verify that they aren't secretly mining outside of the MiningRigRentals server while offline.

<c​urry.guo:matrix.org> Is there any way to deserialize the raw_data that was constructed through wallet rpc transfer?

<r​adar69420:matrix.org> Hey guys really sorry if this is the wrong place. I want to trade my BTC for monero but dont really know how. Exchanges in my country dont have XMR. Can I get some advice on what ever "Atomic Swap" is, or a place I can learn?

<r​adar69420:matrix.org> Again, so sorry if im in the wrong place

<e​longated:matrix.org> Trocador.app

<e​longated:matrix.org> Bisq

<e​longated:matrix.org> Retoswap

<e​longated:matrix.org> Basicswapdex

<r​adar69420:matrix.org> Wow thankyou so much. Do you have a favorite, of these 3?

There is also unstoppablewallet which recently renamed to eigenwallet eigenwallet.org/download.html

Briefly Atomic Swaps are trustless, you either get your BTC refunded or the exact amount of XMR you expected. You have to download software that you need to navigate around to do this for you - and making sure that the offer you take has an acceptable exchange rate

Because you want to obtain xmr with a transparent coin (btc) you have to worry about its history - atomic swaps are the best way to do this ofc as there will be no way to have your funds withheld if source of funds is demanded which could occur on instant swap sites which sre most likely available for you

<h​bs:matrix.org> Some atomic swaps do not require specific software apart from web browser and monero wallet

ah true i forgot. Are there any BTC - xmr ones though? I only know of the BCH - xmr one @ axeswap.net from mainnet pat

<h​bs:matrix.org> Not sure you could do without a specific client when none of the chains have smart contract capabilities

<b​arthman132:matrix.org> I like haveno, but it’s not a replacement for local monero at least in my opinion.

<b​asses:matrix.org> it is different in how they operate

<b​asses:matrix.org> or work

<b​asses:matrix.org> it could be very well a better replacement than localmonero as with its design, could never be private or anonymous

<o​frnxmr:xmr.mx> LM was as private an anonymous as haveno

<o​frnxmr:xmr.mx> (Assuming you used it via onion or i2p)

<o​frnxmr:xmr.mx> well, close enough

<b​arthman132:matrix.org> I get that, but local monero was just so convenient and easy to use

<b​asses:matrix.org> sure it is very decentralized if you imagine so

ofrnxmr wasn't there a difference in who controls your wallet?

that is, did LM custody your wallet?

<o​frnxmr:xmr.mx> haveno's only decentralized in the network transversal

<o​frnxmr:xmr.mx> Nioc, honestly dont remember (and cant say, because it was proprietary). But i think it was something like they custodied a wallet that was protected by your password

<o​frnxmr:xmr.mx> Also @rando private and anonymous have nothing to do with decentralization

<o​frnxmr:xmr.mx> I never claimed LM to be decentralized. I said it haveno's privacy isnt much better

<b​asses:matrix.org> they can start logging and steal anyone's key whenever they want and it would be hard to tell if they are doing so

<o​frnxmr:xmr.mx> That would fall under non-custodial

<b​asses:matrix.org> I didn't use LM, but if anyone u had to create a wallet on LM to use it, right?

<o​frnxmr:xmr.mx> allegedly, LM could steal the funds frok your wallet. BUT iirc, LM claimed to not have access to your wallet

<b​asses:matrix.org> they could compromise this process from the start

LM gave you a subaddress to deposit to

<b​asses:matrix.org> trusting a website is always bad

<o​frnxmr:xmr.mx> Meaning, if you forgot your password, the wallet was locked permanantly. No way to know if true, since it was centralized & custodial

<o​frnxmr:xmr.mx> Nioc, they gave you either a) your own wallet or b) your own subaccount

yes

<o​frnxmr:xmr.mx> Because you could rotate subaddresses

<b​asses:matrix.org> anyways, LM is history now, I would suggest using our energy on improving haveno.

<b​asses:matrix.org> or the next big thing if ever

<o​frnxmr:xmr.mx> As much as i like haveno, as long as 2 randos can sweeo the orderbook, i find it hard to trust

<o​frnxmr:xmr.mx> the only current mitigation is to password protect your offers. But then nobody can take your offer, lol.

<o​frnxmr:xmr.mx> I think my idea was to only require funding the deposit, then trade oartner sends trade request, and when you accept, the rest of the funds are moved. if you dont accept, you can lose your deposit. If rogue arbitrator, you can only lose your deposit

<b​arthman132:matrix.org> Remember if a wallet is custodial. the money isn’t really yours until you withdraw

<o​frnxmr:xmr.mx> thats what custodial means 🫡

<o​frnxmr:xmr.mx> Downside: takes an extra 20mins.

<o​frnxmr:xmr.mx> Upside: you dont have to risk your whole offer

<b​arthman132:matrix.org> When local monero was around I never used the wallet that they gave me in my account. I only used wallets I controlled

<o​frnxmr:xmr.mx> ?

<o​frnxmr:xmr.mx> when selling coins, you had to use the internal wallet

<o​frnxmr:xmr.mx> When buying coins, you _had_ to use an external wallet

<b​arthman132:matrix.org> Yeah I never sold monero on local monero. I just bought coins from there

<m​ilas900:matrix.org> So what’s the conclusion of now the 51% attack and what are the countermeasures

<o​frnxmr:xmr.mx> The conclusion is that numbers dont matter

<o​frnxmr:xmr.mx> 35 identifies as 51

<o​frnxmr:xmr.mx> Its 2028, and you can be whatever you want

there was no 51% attack

<o​frnxmr:xmr.mx> Yes there was

<o​frnxmr:xmr.mx> 35 is the new 51

<o​frnxmr:xmr.mx> And im george washington

<o​frnxmr:xmr.mx> Its 2028 <<>> damn, I need to set my clocks

<o​frnxmr:xmr.mx> Yeah. I didnt know it was 2028 until a few mins ago. Monero us 14000usd bte

<o​frnxmr:xmr.mx> btw*

<t​esttank:matrix.org> Don’t post any FUD pls

<o​frnxmr:xmr.mx> Tell that to someone who poated fud, sir/maam

<t​esttank:matrix.org> 14000$ for one Monero?? Why so bearish

<t​esttank:matrix.org> 140k before ye 2025 minimum

<o​frnxmr:xmr.mx> typo

<m​onero.arbo:matrix.org> lmao just figured out whatever Ubuntu LTS I originally installed on my system had dnssec explicitly disabled in resolved and that's why I've been getting DNS errors for years

<b​arthman132:matrix.org> They didn’t achieve a 51% hashrate, they got 53 out of 97 blocks at that time I believe. As for countermeasures. not much, because we’re sticking with pow and we just have to hope the price of monero goes up substantially to incentivize more miners to mine monero.

i wanted to estimate how it will affect my electric bill, just gotta find my meter maybe.. or i was thinking of buying a kill-a-watt to measure just the computer..

<e​longated:matrix.org> But NGU is bad

also how will the price go up if it's not on many exchanges? arbitrage?

<e​longated:matrix.org> Adoption

<e​longated:matrix.org> Easy dex, bidirectional atomic swaps

<e​longated:matrix.org> See evm chain swap mechanics, if we even get half of the functionality it will be good

<o​frnxmr:monero.social> More like 53/120 w/ 23 orphans

<o​frnxmr:monero.social> Price goes up when we stop selling for peanuts

<b​arthman132:matrix.org> I was talking about the 51% claim from like a week ago that the news articles latched onto.

<o​frnxmr:monero.social> yes

<b​arthman132:matrix.org> Didn’t they orphan over 8% of blocks that day?

<o​frnxmr:monero.social> When they mined "51/100" blocks, ~20 of those blocks were reorged. There were actually ~40 blocks

<o​frnxmr:monero.social> 20 blocks reorged is under 3%

<b​arthman132:matrix.org> Oh ok

<o​frnxmr:monero.social> think abt it like this. If you produce 10 blocks, and i also produce 10 blocks, there are 20 blocks

<o​frnxmr:monero.social> to say that i produced 10/10 blocks is lying. I produced 10/20

<b​arthman132:matrix.org> True

<o​frnxmr:monero.social> Even if, on chain, it shows that i produced those 10 blocks, it doesnt show that, during that period, there were 20 blocks produced

<b​arthman132:matrix.org> I mean even if Qubic did get 51% of blocks. It’s not really surprising, because they often get 30 to 40 percent of the hashrate pretty consistently. So it was only a matter of time that they would be able to achieve more than half of the blocks by mere chance.

<o​frnxmr:monero.social> A 51% attacker can do 100% of blocks

<b​arthman132:matrix.org> Yeah, because they could just orphan all the other pools pretty much, but the owner of Qubic just wants easy cash. so like most scammers he will choose the path of least resistance. He pretty much got what he wanted. he gets publicity, his coin gets pumped up and he gets to dump. Then he rolls in the sweet cash.

if he orphans all the other pools, it will kill monero's value

which is bad for his plan because he wants to use XMR to buy more QUBIC

<b​arthman132:matrix.org> True and that’s why I believe that’s not really his goal and he’s moving on to dogecoin anyway in a couple months

<b​arthman132:matrix.org> Got to feed the endless hype machine to pump his coin

<o​frnxmr:monero.social> Moving to doge is just hype

<o​frnxmr:monero.social> If you want to be famous, why would you mention attacking ltc?

<b​arthman132:matrix.org> I mean he knows it’s impossible for him to realistically go after dogecoin, but it doesn’t matter and when hype overrides people’s brains. People will just believe whatever he says no matter what.

<b​asses:matrix.org> their attacks are based on hype, clout and fud

<o​frnxmr:monero.social> No it isnt

<s​yntheticbird:monero.social> dolus1, you'll need to trial and error, i'm not underclocking might get you a better hash per wat ratio

<o​frnxmr:monero.social> Doge miners are just opportunistic litecoin miners

<s​yntheticbird:monero.social> not sure*

<s​yntheticbird:monero.social> watt*

<s​yntheticbird:monero.social> COME ON MATRIX SEND THE MESSAGE ALREADY

<o​frnxmr:monero.social> Doge, as a chain, is essentially deadcode with no real users

matrix needs 20 seconds to send a message

don't rush it

<s​yntheticbird:monero.social> also dolus1 longevity is a factor for gpu because they are very complex, but CPU are literally indestructible

<d​olus1:matrix.org> i will do that after it's built

also dogecoin is fucking dead

<s​yntheticbird:monero.social> AM5 CPUs are designed to run with no longevity loss at 95°C

<s​yntheticbird:monero.social> or so AMD claim

<d​olus1:matrix.org> i'm 99% sure overclocking will reduce energy efficiency, sure i will get more hash rate but also use up far more kwh which i don't want to do

it's just a rugpull currency for elon musk at this point

and qubic wouldn't have much to ruin :P

<o​frnxmr:monero.social> yeah. Doge is garbage money

<b​arthman132:matrix.org> I mean sure, but with a dogecoin and litecoin are merged mined. it would be extremely difficult for him to take on both

<o​frnxmr:monero.social> No it isnt, @bart

<o​frnxmr:monero.social> The plan isnt for qubic themselves to mine doge

<d​olus1:matrix.org> according to the xmrig benchmark site the 7950x will net around 22-24kh/s so that is what i will try to achieve with it

<o​frnxmr:monero.social> Its to get litecoin miners to accept pay in qubic to mine doge

<s​yntheticbird:monero.social> lmao, these benchmarks are VEEEERRRRRRYYYYYYY synthetic

<s​yntheticbird:monero.social> you can expect less

<d​olus1:matrix.org> my current machine mines on average 5.5 kh/s (pc built in 2019 with ryzen 5 3600) so it will at least be a good improvement

<o​frnxmr:monero.social> Those benchmarks are like overclocking to the max

<o​frnxmr:monero.social> To try to get #1 on the leaderboard

<s​yntheticbird:monero.social> and also sorts of complete autism rabbit hole

<s​yntheticbird:monero.social> yeah literally benchmaxxing

<s​yntheticbird:monero.social> 22kH/s is what I achieve with my R9 9th gen

<s​yntheticbird:monero.social> and that's really the top

<s​yntheticbird:monero.social> I usually sit at around 19

<o​frnxmr:monero.social> Why so slow

<o​frnxmr:monero.social> My 3600 underclocked does at least 6

<e​ddie:oblak.be> can you do damage undervolting?

<d​olus1:matrix.org> it wasn't made for mining and has seen a lot of use

<s​yntheticbird:monero.social> no

<o​frnxmr:monero.social> No

<e​ddie:oblak.be> that's what I thought :p

<o​frnxmr:monero.social> You can go to low and have kernal panics

<e​ddie:oblak.be> thanks for confirming though

<o​frnxmr:monero.social> Or fail to boot

<e​ddie:oblak.be> never tried it

<s​yntheticbird:monero.social> yeah at best you'll fuck the schedule clock

<s​yntheticbird:monero.social> just reboot

<o​frnxmr:monero.social> Try it. These things are all way too high from factory

<e​ddie:oblak.be> it's just fiddling in the bios as I understood it?

<o​frnxmr:monero.social> My sys is way more stable and runs cooler with a slight undervolt

<s​yntheticbird:monero.social> YEAH BUT ACTUALLY LE TURBO GAMING MODE IS LE IMPORTANT OR SOMETHING

<o​frnxmr:monero.social> I dont remember how much i lowered it though

<d​olus1:matrix.org> i'd like to undervolt to about 2/3 - 1/2 normal voltage and run a number of 7950 systems in parallel

<s​yntheticbird:monero.social> gl with your endeavor

<d​olus1:matrix.org> although at a certain point i might be better off just getting a single rig with an epyc

<d​olus1:matrix.org> i'm not sure yet

<s​yntheticbird:monero.social> epyc 100%

<e​ddie:oblak.be> afaik epycs are the most efficient

<s​yntheticbird:monero.social> yeah

<d​olus1:matrix.org> good to know then i will probably look into getting one of those

<o​frnxmr:monero.social> They cost the most fir the hardware

<s​yntheticbird:monero.social> but as usual, any good hardware is enterprise hardware and therefore very expensive

<d​olus1:matrix.org> although the mobos for them are pretty expensive compared to normal ones

<o​frnxmr:monero.social> So roi is a longer

<d​olus1:matrix.org> i don't care as much about roi

<d​olus1:matrix.org> i want efficiency mostly and not too high consumption

<d​olus1:matrix.org> so epyc it is i guess

<e​ddie:oblak.be> from what I theoretically calculated they return most / watt

<o​frnxmr:monero.social> Idk about now, but isnt 3900 still one of the most efficient

<s​yntheticbird:monero.social> At winter I don't want efficiency I want heat but ig different needs

<e​ddie:oblak.be> as I see it the hardware cost is not the problem, it's the recurring cost for electricity that kills profits

<o​frnxmr:monero.social> Hardware costs more than power in the short-med term

<s​yntheticbird:monero.social> last year's winter i put my pc under my bed just so that hot air goes into me during the night

<o​frnxmr:monero.social> Like 10000 for a server is worse than $35/mth

<d​olus1:matrix.org> honestly unless you have free electricity via solar or something idk if mining xmr is profitable at all right now

<s​yntheticbird:monero.social> you get it right

<o​frnxmr:monero.social> Its as profitable as the price you sell it for

<s​yntheticbird:monero.social> it's only profitable if you have free electricity

<s​yntheticbird:monero.social> yeah or this

<d​olus1:matrix.org> i could install a couple solar panels to offset the cost

<d​olus1:matrix.org> maybe

<o​frnxmr:monero.social> if you market sell to mexc while they have withdrawals closed, sounds like a you problem

<s​yntheticbird:monero.social> big investment imo but I would be excited to hear about it

<e​ddie:oblak.be> installing solar panels just to mine will not help

<e​ddie:oblak.be> if you calculate the cost of that solar installation in

<s​yntheticbird:monero.social> solution: steal the solar panel

<s​yntheticbird:monero.social> solution: steal the hardware

<s​yntheticbird:monero.social> solution: steal money

<o​frnxmr:monero.social> Borrow*

<d​olus1:matrix.org> order ski mask + airsoft gun for FREE HARDWARE???

<e​ddie:oblak.be> just go near highway and wait until hardware falls of truck

<d​olus1:matrix.org> so now i need to start looking for an epyc.....

<d​olus1:matrix.org> 🤔

<d​olus1:matrix.org> maybe secondhand from a data center?

<s​yntheticbird:monero.social> poke me in dm if you find one (you won't)

<e​ddie:oblak.be> you can find good deals online if you look around a bit

<b​arthman132:matrix.org> Tell me these so called good deals

<s​yntheticbird:monero.social> most of the time epyc cpus being sold online secondhand are really from companies being broke or individuals having the ability to sell it. Datacenter directly will never sell it to personal

<d​olus1:matrix.org> epyc 7302 for €270?

<d​olus1:matrix.org> maybe

<s​yntheticbird:monero.social> disgusting

<s​yntheticbird:monero.social> garbage tier cpu

<s​yntheticbird:monero.social> you want at least 64 cores 128 threads

<d​olus1:matrix.org> idk anything about epycs

<e​ddie:oblak.be> barthman132: serversupply.com/PROCESSORS/AMD%20E…2.3GHz/AMD/100-000000077_374915.htm

<d​olus1:matrix.org> alright 64 cores it is

<s​yntheticbird:monero.social> yeah i would have guess, that's fine everyone have its first time

<e​ddie:oblak.be> I have never used it in real life but it seems balanced not to high power consumption, still good possible hashrate

<s​yntheticbird:monero.social> mobo will be more expensive than consumer tier

<s​yntheticbird:monero.social> for not so much gain

<s​yntheticbird:monero.social> in hashrate

<e​ddie:oblak.be> you can still go for a 48 core then

<s​yntheticbird:monero.social> true

<d​olus1:matrix.org> Epyc 9654 (96 core) for €3.550?

<e​ddie:oblak.be> these EPYC-ROME series seem also nice

<s​yntheticbird:monero.social> my brother in christ this is the last gen one, also double check i think this one have its max clock capped at 3GHz or something

<r​avfx:xmr.mx> Randomx like fast ram too

<r​avfx:xmr.mx> high mhz + tight timing make a real difference

<r​avfx:xmr.mx> you can't do that on epyc

<e​ddie:oblak.be> if you can spend that money do it

<s​yntheticbird:monero.social> stfu AMD will never make me buy a threaripper

<d​olus1:matrix.org> are there epyc compatible mobos with ddr5 ram slots?

<r​avfx:xmr.mx> yeah, but timings are extremely loose

<r​avfx:xmr.mx> like JEDEC loose

well of course

for a RAM-intensive algorithm, faster RAM does mean faster execution

<s​yntheticbird:monero.social> of course, but the slots are different, you'll be looking for RDIMM instead of UDIMM

<r​avfx:xmr.mx> your going to lost 25% of the hashrate just because of the RAM

<s​yntheticbird:monero.social> different timing, different features, different capacity

<r​avfx:xmr.mx> tighter timing is as important

<s​yntheticbird:monero.social> akshually sir wendell just released a video about the max pricey H14SSL-N from supermicro and it has 12 channel DIMM so it will be fast

<r​avfx:xmr.mx> because randomx is literally trashing

<r​avfx:xmr.mx> not when you have to change row and column at almost every single request

<s​yntheticbird:monero.social> I don't listen to counterarguments sorry

<k​ayabanerve:matrix.org> ChadBird

<r​avfx:xmr.mx> you have to expect CL52 to 54 for DDR5 6400

<k​ayabanerve:matrix.org> Or is SyntheticChad 🤔

<s​yntheticbird:monero.social> I prefer ChadBird

<s​yntheticbird:monero.social> epic

<d​olus1:matrix.org> i found another epyc here that seems pretty good for its price

<k​ayabanerve:matrix.org> I do too, even if it's less obvious

<s​yntheticbird:monero.social> RDIMM i suppose?

<d​olus1:matrix.org> a 9754 (128 core) for €3.230

<r​avfx:xmr.mx> JEDEC memories

<r​avfx:xmr.mx> That's for what theses things are rated for.

<r​avfx:xmr.mx> Except if they finally allow you to select XMP or EXPO profiles (or allow manual timing configuration). But I never saw that on a supermicro board

<r​avfx:xmr.mx> or other server stuff

<d​olus1:matrix.org> €3.900 with VAT 🙄

<d​olus1:matrix.org> still affordable i think

<s​yntheticbird:monero.social> amd.com/en/products/processors/serv…-and-8004-series/amd-epyc-9754.html

<s​yntheticbird:monero.social> 12 memory channels, max clock at 3.1 GHz and theorical max ram speed at 4800MT so I wouldn't call it a good choice

<d​olus1:matrix.org> hmmm

<s​yntheticbird:monero.social> I advice poking around with RavFX because he have much more money than me

<r​avfx:xmr.mx> 4800mts need CL24 or less to be in the "potable range"

<s​yntheticbird:monero.social> and therefore much more experience

<s​yntheticbird:monero.social> CL24 DDR5 literally doesn't exist

<r​avfx:xmr.mx> By potable I mean equal or faster than a Pentium II with CL1 PC100 SDRAM as far as precharging, row and column changing, etc, etc

<s​yntheticbird:monero.social> you wrote that on your 8086 powered computer

<r​avfx:xmr.mx> non, my 3950x

<s​yntheticbird:monero.social> I hate 3rd gen

<s​yntheticbird:monero.social> so I hate you

<s​yntheticbird:monero.social> cordially

<r​avfx:xmr.mx> People just have no idea how memory work.

<r​avfx:xmr.mx> For the AM5 Ryzen, you want 6000MTs with CL30

<r​avfx:xmr.mx> that's what I call optimal

<r​avfx:xmr.mx> Also dual ranks and only two sticks

<s​yntheticbird:monero.social> Give me a single manufacturer of 6000MT CL30

<s​yntheticbird:monero.social> I'll wait

Hanzuei

<r​avfx:xmr.mx> corsair.com/us/en/p/memory/cmk32gx5…memory-kit-black-cmk32gx5m2b6000c30

<r​avfx:xmr.mx> Here first result on tha search engine

<r​avfx:xmr.mx> we are in 2025, not 2023

<s​yntheticbird:monero.social> FUCK YOU

<s​yntheticbird:monero.social> 👍️

<r​avfx:xmr.mx> That shitty timing iteration, append at literally every new memory generation

<r​avfx:xmr.mx> new memory arrive, it's all shit timing (initially)

<r​avfx:xmr.mx> Then it ramp up to the "optimal minimum timing" after a while

<s​yntheticbird:monero.social> You'll release le DDR7 CL400 tomorrow morning and you'll be happy

<s​yntheticbird:monero.social> I'm using CL54 btw

<r​avfx:xmr.mx> I'm using 3600-16-16-16-38 @ 3800-14-16-16-36-*

<r​avfx:xmr.mx> Yes they get toasty too, I have an extra fan aimed at the sticks lol

<r​avfx:xmr.mx> -* mean all other timing are manually set too but I won't reboot to look

<s​yntheticbird:monero.social> Absolute memory

<r​avfx:xmr.mx> It's about the max I can do on that 3950x (the memory have to stay in sync with the infinity fabric, else it add latency)

<r​avfx:xmr.mx> and the infinity frabric don't run faster than 1900

<r​avfx:xmr.mx> matrix.monero.social/_matrix/media/…oad/xmr.mx/kvdEzCBqMnMIOyhibcDUZLzT

<r​avfx:xmr.mx> matrix.monero.social/_matrix/media/…oad/xmr.mx/LBrtjDcNEManUJrCCGAmkXTX

<r​avfx:xmr.mx> vs

<r​avfx:xmr.mx> matrix.monero.social/_matrix/media/…oad/xmr.mx/azfdmFRQKwiQMeQEvbGCkoOv

<r​avfx:xmr.mx> matrix.monero.social/_matrix/media/…oad/xmr.mx/kdIcoZuHrrBnxAltHUlVtIBk

<r​avfx:xmr.mx> ^ Memory bandwidth of said Epyc is like 3.5X more than my ryzen

<d​olus1:matrix.org> damn so the epyc is buns?

<r​avfx:xmr.mx> Epycs are great

<d​olus1:matrix.org> do i just go for a ryzen

<r​avfx:xmr.mx> But usecases matter.

<r​avfx:xmr.mx> For the price of the Epyc, you can get a lot of Ryzen (if the purpose is only for mining)

<d​olus1:matrix.org> which ones are the most efficient in terms of hashes per watt

<r​avfx:xmr.mx> I use the Epyc for other shit. So mining I do on it too, but that's only a bonus

<r​avfx:xmr.mx> well, if you look at previous screens, That Ryzen use ~120W while the Epyc use ~180W while doing a lot less hr

<r​avfx:xmr.mx> But Epyc with more core might be plus or less similar (I think).

<r​avfx:xmr.mx> Like if you compare Zen 5 with 96 cores Epyc vs a 9950X

<r​avfx:xmr.mx> but have to find proper comparison

<r​avfx:xmr.mx> wait non

<r​avfx:xmr.mx> TDP of the 96 cores part is like 320-400

<d​olus1:matrix.org> 400 watts on 96 cores?

<r​avfx:xmr.mx> so I don't know, have to see how hash it produce to see if it's worth it

<r​avfx:xmr.mx> yeah, that's the rating look like

<d​olus1:matrix.org> that's over 9 kwh per day 😟

<r​avfx:xmr.mx> So I think the Ryzen might use less as it will produce a lot more hr (per core)

<r​avfx:xmr.mx> But if you add many ryzen, take into account that you also have many board

<d​olus1:matrix.org> i'm going for efficiency and trying not to break the bank so ryzen is probably best for me

<r​avfx:xmr.mx> so you have to find the mainboard that have the less integrated useless froufrou

<r​avfx:xmr.mx> and smaller as possible, like ITX or something

<d​olus1:matrix.org> + a board for an epyc is 4x the price of a board for ryzen

<r​avfx:xmr.mx> And the Price of a 96 cores Epyc...

<d​olus1:matrix.org> cursory google search shows me an epyc with 96 cores is about €3.550, while 16 core ryzen is about €700. if i wanted to stack ryzens for 96 cores i'd be looking at €4-5k which is more expensive

<d​olus1:matrix.org> not to mention the mobos, psu, storage, ram

<d​olus1:matrix.org> so yeah a single rig with an epyc is cheaper to build per core

<d​olus1:matrix.org> and it's not cose

<d​olus1:matrix.org> and it's not close

<r​avfx:xmr.mx> But it's going to produce less hr than the equivalent Ryzen core

<r​avfx:xmr.mx> which Epyc that was?

<d​olus1:matrix.org> probably also far lower power consumption

<d​olus1:matrix.org> 9654

<r​avfx:xmr.mx> ok, yeah, that's the one I actually got my 350-400W figure

<d​olus1:matrix.org> how much hash can i expect to get on it fresh?

<r​avfx:xmr.mx> If you go to the epyc route.

<r​avfx:xmr.mx> get 6400MTs memory that fit it, and you HAVE to fill all the channels

<r​avfx:xmr.mx> 84K

<d​olus1:matrix.org> sick

<r​avfx:xmr.mx> according to randomx benchmark

<e​ddie:oblak.be> I think if you don't want to break the bank, the AMD Ryzen 7 5800X 8-Core gives you the best bang for the buck

<d​olus1:matrix.org> i take those benchmarks with a pinch of salt

<r​avfx:xmr.mx> slower than my 6 years old Ryzen (per core)

<r​avfx:xmr.mx> as far as mining go

<d​olus1:matrix.org> so if bench says 84k i will say 60-70k

dolus1 I ansered in the other channel, for the 7950X the mobo to get is B650M-HDV/M.2 AMD AM5 microATX Motherboard

<a​lexandre:uii.pt> I've been thinking about upgrading my ryzen 7 1700, was actually looking into that, if it's the one i'm thinking (last one with AM4)

<r​avfx:xmr.mx> hope no

<d​olus1:matrix.org> alright

to get anything near the top bench scores you need to know how to tune the RAM subtimings

<d​olus1:matrix.org> also factoring in energy usage?

<e​ddie:oblak.be> yes

5900X better bang for the buck than 5800X

<d​olus1:matrix.org> i can get ryzen 5800 for €140

smae cost for mobo, ram etc

<d​olus1:matrix.org> very cheap!

<d​olus1:matrix.org> (i think)

<e​ddie:oblak.be> yeah it's one of the cheapest

look at total buils cost

<r​avfx:xmr.mx> 84K on a 96 cores chip from 2025 = 0.875K per core

<r​avfx:xmr.mx> My 2019 cores Ryzen do 1.18K per cores

*build

<d​olus1:matrix.org> at that price point im thinking more about the running cost than upfront lol

<r​avfx:xmr.mx> I wish we could adjust memory timing on Epyc

<r​avfx:xmr.mx> that is literally all the difference

memory timing on my 3900X = plus 30% HR

<e​ddie:oblak.be> you can take the AMD Ryzen 7 9700X 8-Core Processor, which has TDP of 65 watts

<e​ddie:oblak.be> it's a bit more expensive

<d​olus1:matrix.org> the problem with such a cheap processor is you still need a dedicated rig for it (mobo, psu, storage, ram)

IIRC

<r​avfx:xmr.mx> you have to include one power supply and one motherboard and two memory sticks per ryzen too

TDP doesn't mean how much it will draw

<r​avfx:xmr.mx> assuming you PXE boot or something and don't need video

<d​olus1:matrix.org> so there is a "golden ratio" of cpu price to price of all parts combined

<e​ddie:oblak.be> of course

<r​avfx:xmr.mx> it mean how much they expect them to "dissipate" afaik

<d​olus1:matrix.org> for example is it worth it to get a cheap cpu for 140 when the parts to support it cost another 300?

cost/HR

^^^^^

<r​avfx:xmr.mx> but CPU are 100% efficient so it's plus or less that (assuming your not using PBO shit

<e​ddie:oblak.be> but I assume the cost of psu/disk/etc between 9700X & 5800X can't be that big of a difference?

<d​olus1:matrix.org> maybe it is better to get a high end ryzen instead of low-mid end epyc

PBO? NFW

<r​avfx:xmr.mx> you don't need disks if you learn to PXE

static overclock and undervolt

<r​avfx:xmr.mx> Maybe I should spawn a CCS for a PXE mining system

<r​avfx:xmr.mx> I wonder if people want that

<e​ddie:oblak.be> so everything in RAM or?

<r​avfx:xmr.mx> you can boot like ton of machine from the network and they just start to min

<d​olus1:matrix.org> maybe ryzen 9 9950x?

<r​avfx:xmr.mx> it boot kernel+initrd from network

<r​avfx:xmr.mx> Then initrd mount a network from with NFS

<r​avfx:xmr.mx> Then you run the rest from NFS

<r​avfx:xmr.mx> BUT

<r​avfx:xmr.mx> you only need xmrig soooo, an initrd with xmrig would be fine too

<d​olus1:matrix.org> it is the highest non threadripper ryzen cpu on xmrig benchmark

<r​avfx:xmr.mx> 9950X is the fastest normal Ryzen yes

ROI for best cost for HR = long time :)

<e​ddie:oblak.be> I have one on its way, curious ^^

I have 3 mining rn that cost 0

2 others I built for other things with mining in mind

plus more

nice heat in the winter :)

<r​avfx:xmr.mx> that make it 1.7K per cores

<r​avfx:xmr.mx> (1.18K per core == my 3950x)

<r​avfx:xmr.mx> (0.78K per core == Epyc 9654)

<r​avfx:xmr.mx> Don't forget to price memory that have at **maximum** 10ns CAS

<r​avfx:xmr.mx> example of 10ns CAS ::

<r​avfx:xmr.mx> SDRAM PC-100 CL1

<r​avfx:xmr.mx> DDR-200 CL1

<r​avfx:xmr.mx> DDR-400 CL2

<r​avfx:xmr.mx> DDR2-800 CL4

<r​avfx:xmr.mx> DDR3-1600 CL8

<r​avfx:xmr.mx> DDR4-3200 CL16

<r​avfx:xmr.mx> DDR5-6400 CL32

<r​avfx:xmr.mx> If you can get CL digit lower, it's better

<r​avfx:xmr.mx> To calcule nanosecond CAS, do

<r​avfx:xmr.mx> CL * 2000 / MTS (example 32 * 2000 / 6400)

certain subtimings are more important that primary

*than

<r​avfx:xmr.mx> Usually the rest follow nioc

<r​avfx:xmr.mx> but yes, there important. Ideally you tune them manually, some other make big impact

<r​avfx:xmr.mx> Got a big boost when I increased the delay between the refresh cycle

trfc

don't use all your mem OC budget on primary is all imma saying

<r​avfx:xmr.mx> trfc are the delay between refresh cycle!

<r​avfx:xmr.mx> wait non, trfc is how much time you take to refresh the ram

and there are others which I can't remember rn

<r​avfx:xmr.mx> that one make a difference, along with the other one that define the amount of time you wait between each refresh

also setting hugepages is huge on linux, windows are auto set

<a​lexandre:uii.pt> is that a requirement or just increases performance? my laptop is apparently DDR5-5600 CL46

<r​avfx:xmr.mx> But yeah, the secondary timing are all imporant but usually stick rated for good CL will have also good other primaries and secondary

yep

<r​avfx:xmr.mx> It's impororant if your primary load is "ram trashing" aka RandomX

^^ravfx

<r​avfx:xmr.mx> Laptop and thin PC are also often limited to JEDEC timing sadly

increase performance a​lexandre

<r​avfx:xmr.mx> so they will be less good at mining and there is no really good way to fix that, they just don't allow better for theses

<r​avfx:xmr.mx> Tighter timing increase performance but games benchmarks do show only a small difference, its really not the end of the world

<r​avfx:xmr.mx> But for Mining it make a huge difference

<a​lexandre:uii.pt> i see, would the "version" of the ram matter much? or just timing? (ddr4 vs ddr5)

<r​avfx:xmr.mx> For mining you want 10ns CAS equivalent (and hope the other timing of that ram you get is as good)

ddr4 vs ddr5 is cpu / system dependent, I have no experience with ddr5

I do have some DDR3 running rn lol

<r​avfx:xmr.mx> Speed is a thing but speed is there because DDR transfer data at the rising and falling edge of clock plus they burst (send what's at the next address)

<r​avfx:xmr.mx> Each new ddr Generation seam to double the amount they send in burst.

<r​avfx:xmr.mx> So newer generation is BETTER

<r​avfx:xmr.mx> BUT, you lost a lot of that performance if your primary use case consist of constantly changing memory row and columns

still gets surprisingly good HR

<r​avfx:xmr.mx> because while your ram can throw a lot of data back to back, if you request something in another location in the ram, the system have to wait for the ram to select that new location

<r​avfx:xmr.mx> Think about it as a big Excel sheet

<r​avfx:xmr.mx> everything on the right of the cell you select, on the same line, is blazing fast

<r​avfx:xmr.mx> If you ask something on the same row but not right after the cell you are at right now, it's slower

<r​avfx:xmr.mx> If you ask something on another row, it's a LOT slower

<r​avfx:xmr.mx> Timings define the performance of that mecanic

<r​avfx:xmr.mx> (the changing columns and rows thing)

<r​avfx:xmr.mx> While the speed of the memory and the ddr generation define how fast it is when you need to move a lot of data back to back (in the same row and using following columns)

<r​avfx:xmr.mx> Most software get huge advantage with burst because often program are in the same locations.. And game have bigs assets that are usually loaded back to back.

<r​avfx:xmr.mx> RandomX is... Random

ermm. actually, RandomX is random cuz it has random code

randomly-genrated code

<r​avfx:xmr.mx> It also seam to trash the memory... and a lot

<r​avfx:xmr.mx> it's why Timing make a huge difference

that interacts with randomly-generated RAM i guess

<r​avfx:xmr.mx> BTW : youtu.be/7J7X7aZvMXQ

<r​avfx:xmr.mx> If your interrested to know how memory work

<r​avfx:xmr.mx> Great animations

<r​avfx:xmr.mx> [How does Computer Memory Work?]

RandomX is literally its own CPU, which is very interesting

i don't think it has any illegal instruction combinations, anything is legal?

(since it's dealing with randomly generated code)

tell me if i'm dumb :P

<r​avfx:xmr.mx> Well, if it was doing illegal instruction, it would crash ;)

yeah the RandomX CPU would crash... to..... to something?

would it just halt execution? :o

<r​avfx:xmr.mx> halt execution

<r​avfx:xmr.mx> CPU would product a fault and the OS would take care of killing the process

no i mean the RandomX CPU :P

which is being virtualized inside the host CPU

<r​avfx:xmr.mx> You can test it.

<r​avfx:xmr.mx> you use DDR3 right?

<r​avfx:xmr.mx> compile something using -march=znver4 and run it, you see

through forms of either interpretation in high-level code or compiled into host code

<d​olus1:matrix.org> I found an epyc 7763 processor for €1.333

<r​avfx:xmr.mx> oh wait no, that's the cat who use ddr3

<d​olus1:matrix.org> bench is set at about 0.1 MH

<r​avfx:xmr.mx> The bench are widely variable

using MH makes it feel minimal

<r​avfx:xmr.mx> more that i'm looking at it, more bullshit it is

just say 100KH/s

<d​olus1:matrix.org> yeah it says here using 2 cpu's

<d​olus1:matrix.org> so more like 40-50 kh/s

<r​avfx:xmr.mx> ah there, two CPU, would make it like that yeah

<c​hiefprivacy:matrix.org> Best OS to use to mine XMR?

<r​avfx:xmr.mx> Linux, just easier for that purpose

<r​avfx:xmr.mx> And you don't need to run a complete GUI just to mine XMR..

<t​orir:matrix.org> Cindy: RandomX is designed so every possible sequence of bytes is valid code. github.com/tevador/RandomX/blob/mas…er/doc/design.md#21-instruction-set "The interpretation of the instruction word bits was chosen so that any 8-byte word is a valid instruction."