hyc: plotting the delta between submitted timestamps, so it won't be very useful unless the signal is strong

in the case of the network I noticed that behavior on, the secret actor is ~60% of nethash so the signal is pretty clear

but there is always a tight fixed interval between the submitted block timestamps

hyc: re-read your message, sorry, yes - iterating the timestamp in addition

presumably each chip gets a unique timestamp and they all start from nonce 0 instead of splitting up nonce space like earlier miners

i sent you a dm with an example of the trait

so is randomx equally performant on intl and amd when it comes to GhZ?

i remember once i turned all the fancy prediction stuff off on my i7, it gets like 700 h/s

on each core

well, thread. i dunno if it would do well with same # cores but more cache

this is some manic thought that with intel getting clobbered by amd, monero can sweep in and somehow team up with intel to make a new CPU thats great at randomx and its put in super cheap computers and monero buys everyone in the world a computer....

ugh. then we'd have to maintain an OS. well, not really

solar: not AR network?

?

The network where 60% of nethash had this characteristic nonce pattern

not sure which coin AR is, so no

Arweave - the ones who paid for one of the bulletproof audits

the one I was looking at is an older coin with ASICs already, fairly well distributed until this recent increase

some next gen miner killed off the 3 older ASICS that this coin had previously

right.

but new ASIC has simar nonce pattern that you have seen appear on the monero network?

no, it has no nonce pattern, but it has a signature in block timestamp delta

it makes sense to use timestamp rolling for work distribution, especially given all the focus on nonce patterns in recent times

Aha. re-read what you wrote above.

I sent you a dm

gingeropolous my Ryzen can do 850 h/s on a single thread @ 4.1 GHz

862.1 h/s to be precise

1310.1 h/s with 2 threads on 1 core. No Intel CPU can do this.

damn, my beefy intel box is apparently down

i wanted to get some performance numbers :D

<Donkeybrrrrrr> an i7 8700 @4.3 is faster per thread when using max threads then a 3700 @4.1 using max threads

faster than 862 h/s per thread? I don't think so

Of course single thread running on a CPU core is faster than a virtual thread running on half a CPU core

<Donkeybrrrrrr> not on the 1 thread

Show me an intel CPU that beats 35KH/s

<Donkeybrrrrrr> ~740 per thread on the 8700

<Donkeybrrrrrr> the ram is 3200 xmp

you can't compare hashrates of single thread per CPU core and 2 threads per CPU core

the thing is, Intel CPUs suck at RandomX even in single thread, even at higher GHz

hm. the next gen threadrippers have enough L3 cache to utilize 64/128 cores...

But cooling is gonna be a biatch

<Donkeybrrrrrr> Dayum

enough L3 AND enough memory channels*

yep. will be a beast for workstations

Pretty happy with the 3970x.

Intel falling apart. going to produce a CPU at TSCM foundry

"The project is called Sierra Forest and it is about as unique an offering as you can expect from Intel, certainly not a mainstream CPU. Our sources said that Sierra Forest (SF) was a swarm of small core CPU/SoC that was meant to fight the upcoming wave of 32, 64, and 128+ core ARM server chips."

"So how many cores is a Sierra Forest ‘chip’? The largest one, called Sierra Forest-AP by our sources, was 512 cores,"

I talked with intel dude today, he was like "why the hell i did not listen to you and sell all my intel stocks to buy amd"