Are there any plans for an intermediate fork before Seraphis? I heard something about BP++ and PoW change fork

sech1: yes, BP++, PoW change and maybe Rucknium's output selection algo changes seems doable before Seraphis

I think BP++ is currently getting audited, plowsof? There is a Rust implementation so it should not be too difficult to port it to C++ for someone who has experience with that.

Regarding PoW change, it will need at least 6 months gap between "final code merged" and the fork itself

Judging by my previous experience

This is the remnants of the old P2Pool (before March 18th fork, 5 months ago): old.p2pool.observer

So I'd say at least 6 months after "new versions of xmrig/xmrig-proxy/p2pool" are released

Or we'll lose half the hashrate during the fork again

BP++ and Rucknium's changes can be merged later than that, of course. PoW change just impacts more people, not only the people running nodes.

Awaiting update from Cypherstack reg the new bp++ paper (how to proceed / if price or scope would have to change)

sech1: so it might make sense to start looking into potential RandomX tweaks now

if it's the first thing that has to be ready

yes, I have some ideas for tweaks

for example, CFROUND tweak that can boost Ryzen hashrate 10% without changing ASIC resistance

Plus program size tweaks, and scratchpad read/write points inside the program. Zen4 can run 10-20% larger programs with the same hashrate, it's not fully loaded with current parameters and algorithm.

Do you have a rough idea how much of the current network hash rate is based off Zen4 ?

Not much

It's just an example

Zen2/Zen3 CPUs have similar properties

and all Ryzen CPUs are probably more than 50% of RandomX hashrate

*Ryzen and EPYC

I write about Zen4 since it's my daily PC now

To guard against ASICs, it's probably best to optimize for the best arch in terms of hash/watt, right ?

yes

best arch in hash/watt on desktop is Zen4 right now (Zen4 EPYC for servers)

But if we do that and that arch is a small part of the network hash rate, it works against us. There must be some threshold at which it becomes a good thing to do...

7950X can be tuned to 150 h/s/watt

CFROUND tweak will boost _all_ Ryzens, even the first ones

Sure, I mean in general.

the other tweak (program size and scratchpad access points) will boost Zen4 most, Zen2/Zen3 will get smaller boost

No idea about Intel CPUs, but the tweak is generic enough to boost them too

by "boost" I mean total RandomX instructions executed per second, not the raw hashrate

because program size tweak will not increase hashrate, but it shouldn't decrease it too

Jesus, almost 4 years of randomx... How time flies :S

A nice fork to eat some ASICs

old. / old-mini. on p2pool are impressive, but it's mainly one/two miners keeping it alive % wise. Indeed making a LOT of noise about upgrades and having software available well ahead of time seems to be key

Hmm. That's an interesting thing for merge mined coins. When to switch exactly...

3 months advance notice was clearly not enough for p2pool

or was it 2.5 months? lol

first p2pool fork announcement was made in the beginning of January, fork happened on March 18th, and we still have miners who didn't update

announcement was different than software

which came in later

Software was released on January 31st

so 1.5 months :D

and it was February in between

yes, but people new in advance they needed to update

as long as they read reddit :D

hahahaha

"NSA had 2 months notice, as long as they monitored my filesystem"

this is what I mean with making noise then releasing the software, with also longer time and making more noise :D

people will still not update tbh but should keep that to a minimum

yeap. Especially people mining with explorer.exe or svchost.exe :D

botnets are also people!

Only if they're incorporated.

basically monero software will come out and THEN only software developers of other tools might start looking into upgrading their own tools

I just realized we never had an actual PoW change with RandomX. Fork in August 2022 only required pool supdate (view tags stuff), and hashrate didn't drop.

Except for minexmr miner migration which happened at the same time

so expect the actual time for some people to be half of the time than intended

I don't need monero software release. Merged PR is enough

you don't

as for RandomX, even merged PR in RandomX repo is enough

but will "centralized wallet XXX" even look? :D

or "closed source wallet with 300 coin support"

They mostly use wallet2 library

Yes, last August we had many obscure wallets get behind

(anyhow, just keep in mind the time gap, and halve that for maybe the intended duration)

some wallets didn't update for months

pow change wise I already looked at expected API and upgrade paths myself, but can't call everyone else responsible

Don't know about all miners, but P2Pool miners usually update to the newest version in 2-3 months (70% of hashrate)

and even faster

first month is 50%, then it slowly goes up to 70%

83%/70% for v3.5 on main/mini atm

main is also at "70%" because of that one big miner on old pool

numbers somewhat biased on main as it's done on dynamic pplns window, one day I should also publish the stats across time via an endpoint

then 8.8% also are in v3.1 which was the most recent version when the fork happened

mini has better numbers there

So 8.8% never updated after the fork

the kind of people "it works? don't touch it"

old pool miners are even more hardcore "it works/don't touch" people

lemme play with the statistics, I'll make some charts

<k​arano:poddery.com> i am facing problem with Gupax

<k​arano:poddery.com> its exiting in p2pool

<k​arano:poddery.com> my node is starting successfully