poolin was 27 MH/s, it already dropped to 2.6 MH/s

This hashrate didn't resurface anywhere else, so network slowed down a bit

Estimated fork time: 2019-11-30 20:06:17 CET

^ and blog.genesis-mining.com/the-big-controversy-what-is-asic-resistance

LOL

Too much voice given to people that don't understand hardware.

cloud mining is a big scam anyways

lol, RandomXmas

Someone needs to decorate Christmas tree with ARM boards and mine RandomX on them

I'm dreaming of a hot christmas

sech1: I managed to get 6KH on a dual E5-2670 with randomx-benchmark by running 2 instances and using cpu affinity. But not managed to get xmrig quite there - 5800H/s max - and that just by letting it managing NUMA

any tips?

<sech1> Probably post it in xmrig github, it clearly doesn't pick up NUMA layout properly

looks like I found the first false positive with RandomX Sniffer

Process AcroRd32.exe (PID 12052) may be mining RandomX on 3 thread(s)

What's it?

Acrobat Reader

found on the internet; I'm happy that are false positives

it's time to improve the heuristic

apparently, Acrobat Reader uses round to -infinity mode

sech1: so in principle, I should not need to mess with affinity / numa with XMRIG ? It should make correct number of threads and split correctly? (e.g. in this case 10 threads on each 8/16 core/thread cpu)

<sech1> Try to go with auto-config (delete all threads from your config)

First time I tried, I didn't have any thread config, and got about 5700 H/s - but now that I was double-checking with no settings, it correctly detects 2 numa nodes, total L2 and L3 cache, allocates 2 sets of huge pages etc - and yields 6300 H/s - which is better than I got from randomx-benchmark

funny how genesis mining blog is like everyone else, misreporting Monero's approach to PoW

Ooo the sniffer looks cool

here's what keeps supercomputers too busy to mine mobile.twitter.com/bencbartlett/status/1200515462473191425

Can anyone join the coffee chat today at 18 utc? (<1 hr)? PM me if you want to participate in the live discussion

last 100 blocks are all v12 now

I mean v11.12

IT'S HERE

fair play to all involved. serious achievment indeed.

seriously

Thanks tevador and sech1 :)

Tevador!!!!

congrats guys!🥳

didn't poolin stop mining?

yes, they stopped

they got a randomx block apparnetly

or announced that they would stop, I see they've found a block

Nicehash, not so impressive yet: nicehash.com/my/marketplace/RANDOMXMONERO

must have been a lucky block

All pools are still sub 300 MH/s combined, I expected more by now

this is just the initial drop due to unupdated miners

should be >500 MH/s tomorrow

304.65 MH/s on pools now

yeah looks like still a drop in hashratd

ok, so next to keep an eye on is 1980481, right?

epoch change

hrmmm indeed

explain the epoch change please?

thank you tevador sech1 and hyc

midipoet Dataset update, happens every 2048 blocks

how do you mean dataset y

should be smooth, miners will just stop for a few seconds to update dataset

2 GB RandomX dataset I mean

ah i understand, the calculations from which difficulty is comprised, and such

right?

No, it's like DAG in ETHash

had to google the DAG in ethash. it basically helps ensure the algo is memory-hard

yes, big chunk of memory to randomly read from during hash calculation

wtf

its prolly an OS that has all the optimizaitons preconfigured

like those crazy intel speedups that the one guy on here did

closed source

bound to happen

3.5% fee pool

from ryo devs

Why am I not surprised?

ryo is in the final agony before death

I guess

XMRig 10% slower, of course we can believe these numbers, right?

it's just like when Intel publishes their own benchmarks

except they at least put an asterisk there

oh, actually their numbers show 14% faster

no realistic at all

ah, they reduced XMRig by 5% because of 5% default dev fee lol

xmr-stak-rx with exactly the same code if 5% "faster"

*is

4.83/4.46 = 8.3%, this is more realistic on server Xeons

actually, it's achievable

I wonder if they're just bluffing or discovered the same optimization as I did

they are talking about OS optimizations, so I'd say the miner code is copied from XMRig

but 8% on new Ryzens - no, impossible

No OS optimizations can give this boost

unless it's silently overclocking CPU

disabled prefetch can give ~5%

that can theoretically be done by a kernel driver

it can be done in BIOS

but it's very CPU arch specific

it can be done even if the BIOS doesn't have the option

not for Ryzens

BIOS doesn't do anything but set a MSR value you tell it to set

hmm, AGESA 1.0.0.4B BIOS is out for my MB

I'll check if it has something interesting inside

I heard it has some prefetcher options, let's see... Flashing new BIOS now.

but this ragerx, lol

They have their own pool, you can't change it, so they can fake hashrates both in the miner and on the pool

I wouldn't be surprised if they paid only in Ryo

override pool url in dns to your own

that violates their DRM

j/k

Nope, can't find prefetcher options in new BIOS

sech1: which MB?

Asus Prime B450M-K

.network

hashrate didn't change either with new BIOS

so I guess Asus doesn't offer this option

it's a niche setting, most people won't use it

"BIOS doesn't do anything but set a MSR value you tell it to set" Can you prove it?

how to config xmrig to use dual cpu ? on windows ...

it only uses 1 cpu by default

sech1: is your optimization in current xmrig?

not yet

I am still pretty surprised that I am getting 5% more with xmrig on dual E5-2670 after the fork, compared with earlier benchmarks, while Ryzen is spot on identical

kico: on Linux it seems to automatically do the right thing

Is it moving a substantial fraction of c ?

Inge-, I see in windows I'm having a hard time to detect 2nd cpu

tried fidling with threads

numa field in config is only true/false it seems

what does xmrig say when starting up?

Inge-, i.imgur.com/QsxPHeh.png

it says numa 2

just noticed

but cpu load is 50%

my dual cpu system also has 50% cpu load

as there is not enough cache for running an RX thread on each CPU thread

oh ok makes sense

maybe that's as much as its going to make

I should try linux anyways

about 3kh on dual intel hexacore

I'm guessing it is using 12 threads?

if you press h for hashrate

x5650

yeah

so that is the max, based on your cache. It correctly allocates for 2 NUMA nodes

12t tried 24 but was worst

what hashrate you do you get per thread, approximately?

ok thanks a bunch Inge-

~240h/s

per thread

my E5-2670 gets about 400 H/s per thread

so 240 doesn't seem too unreasonable

yeah maybe that's it for this cpu hehehe

I will try linux just in case

you can also google randomx benchmarks - there are some similar cpu systems in one of the lists

great thanks a bunch for the help Inge- :)

no worries.

this is the "oficial" for benchmarks no ? i.imgur.com/QsxPHeh.png

ops

seems legit

DUAL XEON X5650 2743 300 W

mine is doing 2900

should be it :)

thanks!

kico: <3

rawttensox

heyo

kico: yeah, I think you are fine.

cheers :)

gingeropolous: have you used --enable-randomx for xmrchain.net monerod?

for monero-blockchain-explorer behined xmrchain.net*

Hashrate is half a gigameg now

sech1: You may add "RandomX ASIC" into coinbase signature for your solo miner.