goddamn. thats impressive that one 1920x does more h/s than my rig of 11 rx470s/480s

got a reply from TrendMicro, they are going to integrate RandomX Sniffer

*party time*

for anti-butnet

so far the only AV company that replied

Still, that was a pretty quick reply. excellent.

Snipa: github.com/tevador/RandomX/blob/master/src/superscalar.cpp

Yeah, I had misread the design doc a bit.

it's used to initialize the dataset (in fast mode), in slow mode it's used to calculate dataset items on the fly

Yeah, that's what wasn't clicking from the design doc. The wording threw me off.

gingeropolous: 1 8x vega rig - 9KH/s @1650w

gingeropolous: 1 3900x - 14000 H/s @170w

CPU E7-4830 v4 .. 15 kh/s 0watt

bleh

hm. I wonder how much diff between 3600 CL 16-19-19-19-39 and 3600 CL 16-16-16-36

except double the price

Amazon's Graviton 2 64core ARM chip realworldtech.com/forum/?threadid=188879&curpostid=188879

only 32MB L3 cache - once again insufficient cache per core

Seems like only AMD understands how bad of a bottleneck memory access truly is

The only increased L3 because of external I/O die

Higher mem latency

I suppose. but they're kicking Intel'ss ass now

Intel could get a comparable performance increase if they raised cache sizes to match. of course they can't do that with their monolithic dice

likewise ARM could too. Without it, these ARM chips will have perf inferior to AMD

Inge-: when I changed just my primary timings it made little to no difference

but I don't know how rated primary timings relate to the subtimings that are possible

mining with just 2 threads on my laptop hits 82C. not very promising

2 threads with 100% limit per thread vs 4 threads with 50% limit per thread, do you expect no difference between these configurations?

nioc: I expect the 16-16-16-36 is samsung b-die and supports far better timings

10 blocks in 5 minutes. seems like hashrate is still joining the network

total miners 27390

how is that being counted?

i dunno. i tried to get clear answers from some pool ops but didn't. its just the number at the top of column 5 here: miningpoolstats.stream/monero

so each pool publishes their count of miners? I suppose that's unique accounts/payout addresses

if anybody is mining simultaneously on multiple pools, they'd probably be double-counted.

dunno if that's a ensible thing to do

sensible

It's a number of different wallet addresses on pools

It was 40k right before the fork

surprisingly small per-thread difference in hash/s on 3900x and a many year old E5-2670 (v1)

performance has plateaued for a while now. improvements these days are primarily power efficiency, little else

one thread on 3900X is a virtual core

Physical Ryzen core can do 1200 h/s

is that different from a hyper thread on the Xeon?

Xeons run 1 thread per core

Ryzen 3000 run 2 threads per core

You disable SMT on xeons?

No

the E5-2670 is 8 core / 16 thread?

I mean on RandomX. Xeons don't have enough cache

ah

ah right

another one who got close to 16 kh/s on 3950X

same approach to threading - 28 threads

yes, because it's from the same reddit thread

found the smoking gun: i.imgur.com/furmZAm.png

memory dump from SRBMiner

uses librandomx

Is it first successful application of randomx-sniffer? Your message is too cryptic.

randomx-sniffer ca detect it, but my objective was to find out if they violate the license of librandomx

tevador: can you get Trend Micro do to a press release about adding the randomx sniffer and praising the community and you for having it available

I can ask, but I don't think they have integrated it yet

sgp_ or you can get in touch: security⊙tc

tevador: how did they contact you?

that's the only email I have anyways

I got a reply from a guy in Vulnerability Response that they have forwarded it to the development team

cool, thanks

I can email

I think it would be great publicity

yeah, could be good PR for TrendMicro also