blah blah blah. and they still trot out the BS about committing to change the PoW every 6 months

total miners 28296

i am a bot

gingeropolous: START AUTO/open[$6]:,A*/2+close(-)SYNC/CANCEL:PRG("b9ac4ad02f")://8+

OK, you're confirmed not a bot.

we may have Graviton 2 benchmark results incoming twitter.com/_msw_/status/1204953215730479104

so whats the deal with mining software in the app stores?

tevador reddit.com/r/MoneroMining/comments/e9kjx6/randomx_malware

gui integration would be cool

Adding a malware sniffer in the GUI ? Or am I missing some context ?

the reddit link. but yeah thats the idea.

not any malware, just specifically randomx mining malware

The vast majority of people with malware would not be running the GUI, would they ?

And pretty much the only way they'd end up downloading the GUI instead of an AV is if they already found it's monero related malware.

well, a good botnet always checks to see if the host is already compromised , and will take measures to insure that it is not overtaken by a rival botnet

and in some sense, monero is an opt in botnet

well, any p2p network software really

tbh I find it out of scope for the GUI :P

AV software should add this

Or a separate simple detector Qt UI.

usually miners won't run wallets on the workers

nor nodes

would be useful for users who mine on they're main machine/laptop

their*

If you have malware, the likelihood it's one with randomx is pretty small, no ?

So sure, it might help people who have malware that uses randomx and who also use the monero GUI, but that's going to be a pretty small population.

I have not heard back from Trend Micro about a press release

double_facepalm.jpg eng.ambcrypto.com/monero-profitabil…ps-to-new-lows-post-randomx-upgrade

Zen2 optimization manual, finally developer.amd.com/wp-content/resources/56305_SOG_3.00_PUB.pdf

randomx sniffer is a proof of concept, it's supposed to be integrated into AV software

that profitability article: bullshit-o-meter off scale

sech1: interesting to see what you get out of that manual.

RISC-V JIT for RandomX should not be particularly difficult to implement

that announcement is about Samsung using it in modems. not in application processors.

No phone will be running a mobilephone OS on their Risc-v cores

no end users will be able to execute code on those processors

the other problem with targeting any application code to run on RISC-V is deciding which instruction sets to support. or instruction set extensions.

the base set of instructions is quite bare.

i cant wait till we can put whatever OS we want on our phones, like PCs.

hopefully crypto extensions will become a standard soon

on all architectures

(Broadcom will be the last one to adopt it

still no idea why they omitted it. perhaps they had their own proprietary crypto engine

I see no crypto extensions in the RISC-V architecture docs

9640 h/s in Windows

faster than in Linux, even without 1 GB pages

I found a way to write MSR registers there

So Windows is somehow 20 h/s faster, hmm...

restarted XMRig and it's 9655 h/s now, no idea why :D

knowing windows, it's probably losing clock interrupts

so you're measured rate keeps increasing because its clock has slowed down

*your

benching on windows is always a pain, trying to isolate background processes, etc

No, I think it's just the old behavior where hashrate varies between launches

I had this on Windows since forever

yeah, very non-repeatable platform

but damn, setting MSR registers is actually easy, and I only use tools from Microsoft themselves :D

I'll prepare a guide soon

yep, works like a charm

just tested after reboot

9655 h/s again

cool

can you test it?

It works on my machine (c)

I don't have windows on my AMD rigs

we need MSRs for Intel also

Do you mean additional MSRs to that 0x1a4?

hrm, how to set 1 gb on 4 opterons

MSR for Intel are not that important because they all can just disable prefetchers in BIOS

running as sudo didn't do it

you need to have 3x1GB per NUMA node

3x1GB pages

and of course enough free memory for it

i gots it

can you assign NUMA nodes on kernel cmd line?

can you assign hugepages per node...

also, sech1, when my 64 thread thing runs, it only autoconfigs 56 threads

Turn off HT assist in BIOS. It uses some 4 MB L3 cache per CPU, so you get fewer threads in autoconfig

smash bios

hrm, so it seems to have loaded the 1gb thing if i run with sudo

but is it actually doing it? do i actually need to mod grub and other sys files etc

if it works with sudo and prints "allocated 3072 MB ... 3/3 +JIT" or something like this, you're good

if cat /proc/meminfo shows you have 1GB pages, then you're probably fine

on my laptop it wouldn't do it without modding grub config

grub config is more robust because it allocates 1GB pages early during boot process

yeah im at the same hr on my 6276s

even though proc/meminfo shows 1 gb pages

guess i gotta figure out how to mod grub config

if meinfo shows 1GB pages, doesn't sound like it will make much diff

how much free RAM are you showing at the moment?

16G

i got 64 total. maybe i gotta disable the HT

it shouldn't have had any trouble allocating a few 1GB pages then

sech1: my laptop's BIOS doesn't have the option to disable prefetch

then just set MSR 0x1A4 to 6

or run this: github.com/xmrig/xmrig/blob/dev/scripts/randomx_boost.sh

with sudo of course

windows

I will modify your ryzen script and test it

on a completely different note, I donated my randomX XMR to openprivacy

interesting article on Marvell ThunderX2 nextplatform.com/2019/12/10/looking…marvells-future-thunderx-processors

claims their ThunderX3 will be significantly more efficient than AMD Rome

all on TSMC 7nm

Nice, nice

We need to push RandomX tests to servethehome.com when they get it for testing

ah, definitely

They do Chess benchmarking, so why not mining benchmarking?

one guy on the monero mining TG claims about 15.5KH/s on 3900x

It's possible with MSR mod

I've seen 14+ kh/s even without the mod

But 15.5 kh/s... It's 1290 h/s per core, very fast

It should be running at 4.2 GHz or faster

yeah it was way past efficiency

More likely 4.3 GHz or faster + super tuned RAM

his regular setting was 1.175v 14.4KH/S

and yes, 3400-14-14-14-28 with IF at 1700

I'm waiting for the guy with 3950X who got almost 17 kh/s

He said he would try MSR mod tomorrow

3400? so he still has headroom, getting to 3600 / 1800

Not all Ryzens can handle 3600/1800

And RandomX is very sensitive to FCLK instability

I suppose. his seems to be doing well though

I sent a note to servethehome contact page

17.4 kh/s on 3950X: imgur.com/a/9nHQDkO