#off-topic-tech

1 messages · Page 235 of 1

sharp matrix
#

Compared to other ue5 games, satisfactory is practically a gold standard in ue5 optimization now

verbal raft
#

hell naw

#

industria 2 is

sharp matrix
#

Well optimization is far worse in sn2 than satisfactory that much is pretty obvious in either case

verbal raft
#

looks so much better than freakin screenspace shadows

image.png image.png
verbal raft
#

though theres certainly work to be done in sn2

sharp matrix
verbal raft
#

and satisfactory looks terrible.. jace_smile

sharp matrix
#

And this is the settings the game is recommeding for my system

verbal raft
sharp matrix
verbal raft
sharp matrix
#

Seriously this is the future of gaming a guess, upscaling and fg, because lord knows whats going on under the hoodjace_smile

#

Nvidia would be pleased indeed

verbal raft
#

sn2 is the exception not the rule though

#

in every other game ive played i'd immideately notice that diference
not in sn2 tho

sharp matrix
#

See its more optimized for upscaling, its all clearly planned...m

verbal raft
sharp matrix
#

Sad but probably, why else would release specs for 60 fps that werent true, because it was true but only with upscaling

#

I can believe it, because noway is that true for native

verbal raft
#

seems plausable to me that the rec spec could get that much at native rez

#

with those settings

soft bloom
sharp matrix
#

Ultra++ is even worse, not even a 5090 can reach 60 fps at those settings, saw someone review it and only got around 40, so yeah...

#

As for recommended, considering ultra and ultra++ arent true i dont have alot of hope recommended is true either

verbal raft
#

epic buffs volumetric fog rez by an insane degree
which makes fog cost more than shadows and lumen combined

#

theres this video
and if its real then they werent using upscaling in the recomendations

image.png
#

(like the devs said they werent on DC many times..)

verbal raft
#

also wai a minute

#

its the 4070 S jace_smile

twin dew
#

Getting 55 FPS or so on RX6600, on Medium FullHD with Upscaling and Motion Blur disabled, in the water around the drop pod after entering the main map.

#

Restarting on High...

sharp matrix
verbal raft
twin dew
#

40-43 on High...

#

Anyways, back to Below Zero

sharp matrix
#

I found a place under water, if i dont move 60 fps is possible at dlaa high preset

twin dew
#

Which is why I specified that easily reachable place...

#

Because even above or below water made real difference.

#

And which was faster changed from Medium to High.

verbal raft
#

1440p high
45-55fps really here at native

image.png
#

and thanks to my tweaks shadows are flickering a ton i think

sharp matrix
verbal raft
sharp matrix
verbal raft
#

i feel zero pain

sharp matrix
#

Lol

verbal raft
#

have i mentioned yet that satisfactory performs worse actually ?

image.png
sharp matrix
#

Dont worry fg and upscaling is the futurehehe

twin dew
verbal raft
#

new save

sharp matrix
verbal raft
#

just a hypertube line for easy repeatability

#

cuy i was testing with someone who has a 7800 XT

sharp matrix
#

Btw new game in satisfactory easily 120 fps, meanwhile this game new game, pray for 60, pray hardjace_smile

verbal raft
sharp matrix
#

So about half the framerate of satisfactory, before anything is built

verbal raft
verbal raft
# sharp matrix Lol

i dont play at 1440p
even if i did i have no problems with quality upscaling

#

at 1440p

sharp matrix
#

Aka sn2 upscaling isnt saving this massive performance gapjace_smile

verbal raft
#

SF's performance varies wildly depending on where you are

sharp matrix
verbal raft
#

i already deleted this once today..

image.png
verbal raft
sharp matrix
#

Red forest looks visually really good, while performing quite a bit better, but satisfactory ue5 remains a well optimized game, sad most other ue5 games just arent up to this standard

#

Maybe css can give sn2 its engine, thatll fix it right?

verbal raft
#

this doesnt run nor look good

image.png
#

40fps and you get

#

blobs for shadows

image.png
#

and immideately obvious LOD poppin

pure karma
#

Could always be a cpu skill issue since my cpu is only recommended spec

#

dunno didn’t look at usages

sharp matrix
#

This is purely your faulty hardware, 1440p ultra settings, easily 80 fps..m

pure karma
verbal raft
pure karma
#

On high yea i get like 100

sharp matrix
#

Idk whats wrong with your system but might want to fix that

verbal raft
sharp matrix
verbal raft
#

do show

sharp matrix
#

Really what is wrong with your system

#

Im actually curious, 40 fps, yeah im easily double that

verbal raft
sharp matrix
#

see, im not ragebaiting, ultra setting 1440p, averaging around 80 fps here

verbal raft
sharp matrix
verbal raft
#

where are your screenshots of the settings ?

verbal raft
twin dew
pure karma
#

I get like 150 on ultra + lumen if thats what yall are coping over

verbal raft
#

lighting yes
shadow no

verbal raft
#

but i'll take that screenshot again without them

sharp matrix
#

like i said, all ultra settings, 1440p

#

this is dlaa btw, so basically native

verbal raft
sharp matrix
verbal raft
#

still terrible

pure karma
#

To be fair those 2 are set to default

#

I have i hard time believing that i get a better framerate there with my fatass train in that biome than you do without it

sharp matrix
verbal raft
sharp matrix
verbal raft
#

you still dont have GI enabled

sharp matrix
#

thats a optional unsupported feature, normally i dont have it turned on

verbal raft
sharp matrix
#

so medium or high for that?

verbal raft
#

high

verbal raft
# verbal raft

turning it off still runs terrible
and deletes all indirect illumination

image.png
sharp matrix
# verbal raft high

struggling around high 50 fps now, probably my system can handle lumen calculations better?

verbal raft
#

it should

#

you have a more capable card

sharp matrix
#

maybe

#

i thought lumen was on cpu, it doesnt have hardware support

twin dew
#

GPU shaders.

sharp matrix
#

that devs said that anyways i believe

verbal raft
#

software =/= CPU
it runs on GPU shader cores for the most part

twin dew
#

"SW" Lumen is on standard GPU shader units, as distance fields.
HW Lumen is on GPU RT-HW units.

sharp matrix
#

hmm i wonder does sn2 use lumen by default hmmm

verbal raft
#

yes

sharp matrix
#

idk is a ue5 game

verbal raft
#

it does
also VSM

#

super low rez VSM by default but its still better than satisfactories aproach

sharp matrix
#

hmmm maybe a difference of how this game and sn2 handles lumens, because so far even with lumen enabled, this still runs better, but maybe its not the same calculations hmmm

verbal raft
sharp matrix
#

that doesnt look right at all

verbal raft
#

that the devs are abusing this shadow method

sharp matrix
#

the shadows here look bad

twin dew
#

Or you enabling thing that hadn't been touched by devs and doesn't work right because of that?

#

If that was with that UE feature enabled by console.

verbal raft
#

no thats default

#

default (CSM + DFS) / untuned VSM

image.png image.png
sharp matrix
verbal raft
#

it always should

sharp matrix
#

so sn2 uses lumen, does it use vsm or something else, idk

#

the shadows in sn2 i noticed look good, so it cant be a too primitive method

verbal raft
#

devs tweaked some stuff with it tho

#

well, made their own additions*

#

CVAR dump

image.png
verbal raft
sharp matrix
verbal raft
#

anyway thanks to the devs tweaking it occasionally has this undenoised megalights esque noise

#

maybe it let them get away with a lower cost but i dont really get why they'd do this

sharp matrix
#

Because weird stuff starts to happen inside of the bases,you build with lights and shadows, all of these strange and out of place artifacts start to show off

verbal raft
sharp matrix
#

But that could be because of the tweaks the devs did to the engine sn2 uses

verbal raft
verbal raft
sharp matrix
pure karma
#

what in the ever living flying fuck is this new UI tired_jace

image.png
pure karma
#

also somehow i dident think it could get worst but FH6 is even worst than Sub2 in the higher settings in therms of optimisation slop mostly cause of the vram usage

#

ironically tho that means the 1080ti gets an edge again even on cards like the 4060 and 5060 😂

#

so it gets to live another day

sharp matrix
twin dew
#

And I just cannot stop tweaking when I think of new failure cases that cannot happen on my deployment...
In this case DNS server(s) the OS uses changing.
Catch the error, reread OS nameserver config if DNS servers aren't set by program config, retry once.

        if (not config.custom_dns) and (not is_retry):
            logger.warning(
                "%s: %s: No working DNS servers. Rereading system DNS configuration and retrying.",
                set_entry.set_name,
                domain_entry.name,
            )
            logger.debug(
                "Old DNS server(s): %s",
                ", ".join(config.resolver.nameservers),
            )
            config.resolver.nameservers = Resolver().nameservers
            logger.debug(
                "New DNS server(s): %s",
                ", ".join(config.resolver.nameservers),
            )
            return query_dns(set_entry, domain_entry, config, True)
        <snip>```
#

The original version recreated new Resolver() for every query...

night girder
#

Is it me, or are big companies getting hacked more and more?

#

TeamPCP asking for 50.000 dollars for 4000 repositories with private code.

#

It feels that the frequency is going up. Might be the wrong perception on my part. Can't shake the feeling though.

warm jay
#

its definitly increasing. i think like at least 10 major security breach in less than a year

night girder
#

This week I read about a script Google security team found, probably written by AI that was able to bypass 2FA 😒

warm jay
#

oh didnt know about this one

#

i think there is also some major linux kernel vulnerabilities found recently, some in nvidia drivers

night girder
#

But they were able to stop it.

#

And based on the script itself and the comments, they believe it to be AI written.

warm jay
night girder
#

you can find screenshots in the article of the blog post, if you are interested in the "hallucinations" comments.

warm jay
#

the increase in security breach is very probably due to AI

#

i hope company get theirs things together and stop taking security lightly.

night girder
#

They will implement AI for security. While AI will also do the breaching.

sand saddle
#

What a lovely world we live in

pure karma
slow flint
#

We're actually all in the meta metaverse

visual tree
#

Hopefully, I will be the first example of self-substantiation and wake up from the matrix hehe

twin dew
#

SpaceX Starship Flight 12 (with v3 booster and ship) going for launch on 2026-05-21 21:30 UTC, so under two days from now.

sand saddle
#

Oo nice

oblique hamlet
#

are they trying to escape the matrix by flying outside it?

soft bloom
# 
lass ConfigException(Exception):
    def __init__(self, message: str, extra: list[str] = None):
        if extra is None:
            extra = []
        if not isinstance(extra, list):
            extra = [extra]
        super().__init__(message, extra)

first of all, this annotation is technically incorrect,
list[str] | None would be correct.
since None is not just ok, not just handled separately, but also a default.

  1. checking if extra is a list is a bit limiting. A better type would be Container, unless order has any meaning, in which case - Sequence. typing module has some types that are wider than std dtypes.

  2. I would argue that it's often easier to force container type as input instead of handling when it's not.

...which lead me to realize that annotation actually should be str | list[str] | None.
and so checking for container or sequence, logically, ssould catch str too.
so better to do reverse, and check if it's a string, and otherwise - leave as is.
But again - all this comes away if you only expect it to be either container of strings or None.

why container? because generally tuple is a better and faster way in 90% of cases. you need to append soemthing only so often

#

would be funny if I was wrong...
but you actually don't need to write + on new line - consecutive string literals, including f-string, will concatenate anyway

image.png
#

ok, sorry
that works if you encase them with ()

mental oriole
twin dew
twin dew
soft bloom
# mental oriole If you care about speed python is the wrong language.

if I don't care about speed at all might as well write in JS...

anyway, speed is a consideration. just not at highger priority than writing working, correct , and readable code.
in most cases you lose absolutely nothing by replacing [] with ()
and eventually you judge before typing that this specific tuple has no need to be mutable, so you write () initialy)

#

protective is good, in my opinion.
in right amounts.
that's how you avoid bugs in high traffic places

twin dew
#

But in the case of that exception, which is meant to cause exit of the program, speed really doesn't matter.

#

And it originally used .append()s to add those extra arguments.
Including when I rewrote to have that init.
I later changed the uses to do away with those.

#

But changed it to
def __init__(self, message: str, extra: list[str] | str | None = None):

#

The idea of that second argument is to allow for adding those debug output lines to when the program exits because of config file errors.

#

Near end of the file:

        nftsets, config = initialize()
    except ConfigException as exception:
        logger.error(
            "Config Error: %s. Exiting.",
            exception.args[0],
        )
        for entry in exception.args[1]:
            logger.debug("%s", entry)
        sys.exit(1)```
soft bloom
#

speed is a bonus to correctly choosing datatype that fits the needs of the usecase - which is to hold 1 or more strings.
there's no need to have it mutable.
(oh, it was mutable? okay... but is not mutable anymore?)
and, as you pointed out - it is used in exception. simpler the code, fewer mutations - less chances there will be error while handling exception.

I mean, if you can make it simpler - do so. Unless needed otherwise.

and having stricter extra: tuple[str, ...] = () is doing just that.
you don't need any more checks inside (maybe force str conversion on all items). the only real change is swapping [ for ( where it is used, and adding those to cases of passing just one string.

twin dew
#

in the main().
So if someone uses the function(s) separately, they have to handle the errors themselves, and the errors don't cause exits.
For unit testing etc.

twin dew
#

That is the format it wants, assignment to None as default and then if in the actual function to set the "right" type right after.

#

Don't remember the reason the docpage gave for that format.

soft bloom
#

default list is bad because it's mutable.
idk why would it ever complain for empty tuple...

twin dew
#

Ok.

#

Yeah, that was it IIRC.
But I would rather support .append() still for future, even when the current code doesn't use it anymore.

#

As the performance difference really doesn't matter for that usecase.

soft bloom
#

thing about mutable values in default is that it lives as an object that can be mutated in one function call, and remember new value for the next...

twin dew
#

Ah.

soft bloom
#

just pointing at options here, ultimately it is a nitpick

twin dew
#

But I should change those set_settings and global_settings lists to tuples then.

soft bloom
#

I think the last time I used pylint was with fairly strict plugin for codestyle witha fairly long name...
and eventually I switched to ruff as it still highlights the stuff I care about without too niche warnings.
Still getting erminders to write docstrings hehe

#

also, I saw that dataclasses are not frozen. and maybe they don't need to be.
but just in case I will say that frozenset is an immutable set

#

thought sadly using it is a bit bloaty since you need to pass another iterable to it 🙁

twin dew
#

Those DomainEntry and NFTSetEntry do get changed all the time.

#

And same for those ip_set:s in both.

#

ip_set and next_update attributes keep getting changed in the code.

#

for that set, could use the functions to change the contents...
Which I didn't know before, read about those later.
For that datetime, would be annoying, but might be doable.

#

No, those are immutable and just return new objects.

soft bloom
#

sometimes, only sometimes
it's good to have object immutable but methods that handle common "copy and update" cases.
doesn't seem to be the case with DomainEntry though, judging by the code.
you change 2 of 3 attributes per update. and you don't need history/dump, which is where it could help in a way.

twin dew
#

Also, the combinations of those could be switched over to being tuples after the parsing is complete.
So not list[NFTSetEntry] but tuple(NFTSetEntry) and same for the NFTSetEntry containing list[DomainEntry] could be tuple(DomainEntry)...

#

Which actually matters when those are the stuff that is constantly being used for the programs lifetime.

#

Done, moment to get the new version here...

#

And changed the link in original with .conf to here.

visual tree
#

Ewww, it's SAP Hana test day today....

twin dew
#

Just made it use tuples from the start and do +=:s in the parsing.
Slower there, but doesn't really matter.

#

When that parsing part takes now 5 milliseconds from first debug log message from program start to end of parsing.

#

4ms from the start of parsing message to end of parsing message.

#

Of course gets worse and worse as you make more sets compared to list.append().

#

But I would expect that you would need insane amount of sets before that actually mattered and made the parsing phase too slow.

#

Ok, better to just convert back to being list at start and convert in the end...

#

So moment again...

visual tree
#

Stupid Workflow business capture software....

#

Been trying to install it for almost an hour

night girder
#

I first look up why it's warning me about the stuff. Then I just try to think about it for a while if I am doing something dumb.
But if it's about comments, or strings too long. I don't care honestly. It's a balance between common sense and learning why pylint is warning me.

#

If it's a warning about security or potential memory issues, that's a different story ofc.

#

Some things even complain about adding TODO's with some sort of logic "don't say you going to do it, just do it."

soft bloom
visual tree
#

Subnautica 2 has already sold 4 million copies

image0.jpg
night girder
#

But in the end, I reverted back to TODO's. It's just more easy for me. And the IDE supports it (can give me an overview of all my TODO's etc).

pure karma
#

i made 6.9M empty folders in the windows folder of one of the school computers just to make a point today

#

and yes the day was just that boring

night girder
#

Someone was able to pull this off while back hehe

image.png
#

17179.87 Petabyte cookies and site data 🤣

pure karma
#

edited probably?

night girder
#

Nah, bug.

pure karma
#

since i you you can HTML that menu dunno about the number

night girder
pure karma
#

generational whatsapp user

night girder
#

I mean, always possible it's fake these days 🤷‍♂️ But posted in small community.

night girder
#

I whish I could nuke it from my phone. But family, work related stuff keep using whatsapp sigh.

pure karma
#

also dafuq kinda dependacy is that

image.png
#

saw this earlier

night girder
#

What the ...

#

what is gaming services even? xbox/microsoft thing?

#

could be performance related.

pure karma
#

no idea probably part of game bar or something

night girder
#

baldur did mention it a few times but can't remember details.

pure karma
#

yea its needed for online stuff and forza is online only but the fact it dosent update on its own is crazy

twin dew
#

Yeah, that is MS Store thing from MS.

night girder
#

I think I read something about hackers bypassing DRM (denuvo).

#

Nothing is 100%

#

but the gaming service seems more orientated to performance and updates?

twin dew
#

Just thing for games that have MS account bindings in them for Xbox achievements etc.

#

AFAIK.

night girder
#

Does it also force you to have the service if you have forza on steam (which has it's own achievements?)

twin dew
#

Probably, up to the game dev if they use that MS library or not.

night girder
#

weird...

#

I got the xbox stuff removed afaik. So the game wouldn't work on my PC.

twin dew
#

Most games don't use that MS (XBox) Gaming Services thingie.

#

And same, I don't have it installed.

night girder
#

yeah, weird it's a hard requirement to play the game. And I don't find "why" it is that way.

#

As in the developers telling why it's required.

twin dew
#

No idea if it is required for non MS-store installs or not.

#

Because Forza Horizon 6 is free to install from MS Store:
https://apps.microsoft.com/detail/9nr1r1xwlcnb?hl=en-US

Microsoft Store - Download apps, games & more for your Windows PC
Forza Horizon 6 - Free download and play on Windows | Microsoft Store

Discover the breathtaking landscapes of Japan in over 550 real-world cars and become a racing Legend at the Horizon Festival. Start your journey as a tourist and explore a world full of hit music and Japanese culture. Build a Valley Estate, acquire awe-inspiring homes, and display your prized car collection in fully Customizable Garages. Cruise ...

night girder
#

steam:

image.png
#

So tthat's probably why.

twin dew
#

Ok, so it is forced MS Account binding thing, no matter the PC store source.

night girder
#

It seems that way yes.

twin dew
#

But it is MS title.

night girder
#

Also ... 70 dollar game.

image.png
#

Well, microsoft just telling me this game is not for me 🤣

pure karma
#

you need to be always signed in to Xbox and need all the random stuff they want

twin dew
#

Well, it is MS first party title with online features (and microtransactions etc.) and it uses MS/Xbox account for all that.

#

And that MS Gaming Services is what on Windows provides lot of the common functionality for MS/Xbox accounts in games.

night girder
#

And I read that both sony and microsoft want to go back to exclusive titles.

#

"Along the way, we will reevaluate our approach to exclusivity, windowing, and AI, and share more as we learn and decide."

mental oriole
pure karma
jagged snow
#

Tried out a new distro, quite the fan

image.png
#

I've always been a debian/ubuntu user previously but I wanted to try out a fedora based distro for my gaming PC. I think I'll be sticking with it for the long haul

night girder
#

What did I read about vulnerabilities deep into Linux this week? It got fixed quickly if not mistaken.

#

CVE-2026-31431, CVE-2026-43284, CVE-2026-43500 and CVE-2026-46300.

#

Copy Fail, Dirty Frag en Fragnesia.

#

But on the other hand, there was also exploitation code available too. So quickly found leading to quickly fixed but also quickly abused.

twin dew
#

In one of the cases, the reporter published the exploit themselves before kernel and distro people had time to release fixes in coordinated manner.
In the other, someone failed and pushed the kernel patch to public repo too early and the exploit was reverse engineered before the coordinated release date.

night girder
#

And the investigators/security experts think it's probably AI finding these exploits.

#

I just hope these things are temporary. AI finding new exploits/vulnerabilities. People patching it up and that eventually AI will find less and less exploits.

twin dew
#

That reported and released was using AI, by company that sells AI for just that use.
So self-advertising.

night girder
#

They also start to use AI to check code for vulnerabilities before pushing to production, which can be a good thing.

#

Maybe now the code quality goes from "97%" to "99%".

#

But the fact remains, people with bad intention also use AI. So it's a race I guess 🤷‍♂️

night girder
uneven yacht
#

ive finally upgraded from HDD to SSD storage, now my server is all-flash instead of partial flash and partial spinning rust

20260522_032446.jpg 20260522_032718.jpg 20260522_041047.jpg
glossy glacier
#

Has to be the worst time to do it

soft bloom
#

aren't HDDs like much more realiable in the long run?

twin dew
#

Depends on write volume.

#

Where HDDs don't really care, only about stop-start cycles and runtime.

uneven yacht
#

thats when you measure MTBF

#

and also i dont care because im using RAID

twin dew
#

Now where is my 0.5mm drill...
I have microwave to fix...

One of spring end retainers for the lock mechanism springs is a tiny plastic hook extrusion, that has finally split off from the larger main part.
Need to get it back on.

twin dew
#

Damn, I'm out of metal wire of 0.5mm diameter... Only 0.7+mm stuff left...
Need to go buy more tomorrow, as even when I do have 0.7 and 0.8mm drills, the hole gets too large, not enough plastic left.

#

The cutline is on 2mm x 2mm section.

uneven yacht
#

far FAR exceeding the actual drives lifespan

twin dew
#

For SSDs, there is the other stat too, TBW and the daily version.
How many times the drive capacity you can write each day for the warranty duration to hit the expected write life max.

#

Where for enterprise drives you usually have 0.3 DWPD drives for mostly read stuff (and most consumer drives are here or even worse), then 1.0 DWPD stuff as generics, and 3.0+ DWPD stuff for write intensive.

twin dew
#

So found 0.5mm x 10m pianowire (spring steel) spool source nearby, need to get one tomorrow.

vocal thorn
pure karma
#

daily dose of this channel being invaded by war thunder players

bronze jasper
#

whats the good stream for the spacex launch?

pure karma
#

also decent news subnautica 2 got a optimisation patch already and if im not mistaken my performance has about doubled now so more good news for once

bronze jasper
#

i already beat it and am looking forward to the next chapter a yera from now

pure karma
#

yea i did a while back now but im just building random stuff and crashing out over all the bugs with the building

vocal thorn
twin dew
#

Flight 12 happened in last 24h.

wanton orchid
#

supposedly fixed since though

uneven yacht
#

Most read focused enterprise drives have 1 DWPD

#

And most write focused enterprise drives have 3 DWPD

#

Mine are the latter

twin dew
uneven yacht
#

Anything below 1 DWPD is rarely seen in the enterprise market at all

twin dew
#

Using QLC NAND etc.

uneven yacht
#

I've only seen 1 drive in the enterprise space that has a 0.3 DWPD rating

#

And that's the LC9

#

Which I'm not even sure is available for sale yet, and even if, I don't see any datacenters buying them

#

Maybe hyperscalers do, that one I don't know about

#

It has some kinda market that's for sure or kioxia wouldn't make it but it's not a wide one for sure

twin dew
#

Samsung BM1743 series is with 0.26DWPD for 5 years for example.

#

As PCIe 5.0 x4 SSD.

#

All the manufacturers have some.
They just aren't usually in the headlines except for the "new largest ever drive" news from misinformed tech media.

uneven yacht
#

The Samsung what? I haven't even heard about the BM line

twin dew
#

Those Samsung enterprise/datacenter SSDs aren't in the news.
And not for sale in retail channels.

uneven yacht
#

61TB?! Who the hell is this for

twin dew
#

15.36TB - 128.88TB per drive.

uneven yacht
#

At this point just get 2 or 4 drives brother, it'll be cheaper

uneven yacht
twin dew
#

Like I said, for actual enterprise use.
For when you need to store lot of data, that needs to be fast to read.
But isn't replaced much.

uneven yacht
#

That's what HDDs are for

twin dew
#

Those aren't fast to read.

uneven yacht
#

They're really fast at scale

twin dew
#

Seek times are a bitch.

uneven yacht
#

A server full of them can easily reach 2 or 4GB/s

twin dew
#

They are fast in sequential reads.

uneven yacht
#

A SAN array can do 10 or 20GB/s

twin dew
#

While one of those PM1743:s is rated for "up to 14.2GB sequential reads" with 1780k 4k random-read IOPS.

uneven yacht
#

This is an incredibly niche market

twin dew
#

If the demand wasn't there in someones usage, they wouldn't exist.
They are niche drives, but every enterprise SSD maker has a model.

uneven yacht
#

Well yeah that makes me think

#

Who's buying them, someone has to or they wouldn't exist

twin dew
#

I did say those 1DWPD drives are the general use drives.
And those 0.3DWPD QLC drives are specific niche.

#

Like the 3DWPD "cache" drives, or even higher ratings, are another niche.

uneven yacht
#

But what hyperscaler would have a lot of data that fits that niche

#

Google is the only one I can think of

twin dew
#

You just cannot stuff the same amount of storage as HDDs into same space and power envelope either.

#

And if you think how many HDDs you would need to match the speed of even one of those, it gets even worse.

uneven yacht
#

Not sure why, but they seem pretty enthusiastic to use them

#

Surprisingly, that does not apply for ProLiant servers

#

Tbh I should ask HPE about that now that I think about it

uneven yacht
#

Which is what makes me think hyperscaler, but then also, which hyperscaler would have that much read only data

twin dew
#

No idea.

uneven yacht
#

Google popped up in my mind because they have massive databases that store past events for their multitude of search and recommendation algorithms

twin dew
#

But the world is full of data that isn't replaced much.
But might be read often.
Or need to randomly be read fast.

uneven yacht
#

Past events don't change, they could do even with ROM because that database never changes in any other way than just growing

twin dew
#

But basically (almost) all the various user cloud storage would happily live on those drives etc.

#

As basically no-one replaces all the data every three days.

twin dew
#

Anyways, realized I could just use that 1.0mm spring steel wire to make a metal replacement hook.
The main body was large enough to take it.
Also, that original plastic hook was always in rotational torque...
That spring is straight one, but the hooks were 90 degrees offset from each other.
And the other one is part of the metal door stamping.

#

Also, "classic" electronics, including DIP format microcontroller.

20260523_144657.jpg 20260523_100101.jpg
twin dew
#

Googles new "don't list pages that highjack back button" search policy has made Google Search basically useless.
Expecting that they have something wrong in that detection with there being barely any results, ever.

#

For search of "Review" I get two pages of results.

pure karma
#

mine works as normal for now anyway

glossy glacier
twin dew
#

That was what they said they would blacklist.
But seems that something has gone wrong with that detection and most of Internet is currently blacklisted.

glossy glacier
#

good, i hated that on microsoft sites >.>

twin dew
#

https://www.youtube.com/watch?v=2kxanBYTAaY

YouTube
Scott Manley
Starship Flight 12 - V3 Debuts with Max Power, Fatal Flips, Fast La...

Taking care of your health just got easier – start here with my sponsor Zocdoc: https://Zocdoc.com/scottmanley

Flight 12 was the debut of Starship V3, powered by Raptor V3 and not to forget launch mount 2. There were a lot of changes informed by previous successes and failures, and... a handful of new successes and failures. There were rapto...

▶ Play video
soft bloom
pure karma
#

"click now for hot volcanos in your area" jace_smile

sterile plinth
#

I'm deciding which games to move to my 2nd SSD to free up space, which kind of games would be least affected (in performance) by being moved to a slower SSD?

#

should I try to keep my open-world games with big worlds in my fast SSD since they'd benefit the most?

sand saddle
#

As long as it's ssd game load times will have very similar load times

languid gulch
#

screenshot of the Starship booster completely missing a grid fin during staging

2026-05-23_200239.png
jagged snow
#

They removed that on purpose

#

They switched to 3 larger ones to give them more tower clearance on catch

#

And the 4th one ended up being mostly useless after boostback because it spent the entire time in the booster's wake

#

Overall the only real issues on this flight were the v-rap failure on the ship(unknown cause) and fuel starvation on the boostback due to an incorrect flip direction most likely caused by the new interstage.

tall forum
#

will satisfactory run reasonably well with 9070xt and ryzen 5 9600x 1440p ultra

soft bloom
#

https://www.youtube.com/watch?v=1zoWUW29TNA
Yay, microplastics are contributing to global warming,
not cool
(pun intended)

and they alter precipitation

YouTube
Anton Petrov
Microplastics Just Upgraded From a Health Risk to a Planetary Threat

Support this channel on Patreon to help me make this a full time job: https://www.patreon.com/whatdamath (Unreleased videos, extra footage, DMs, no ads)
Alternatively, PayPal donations can be sent here: http://paypal.me/whatdamath
Get a Wonderful Person Tee: https://teespring.com/stores/whatdamath
More cool designs are on Amazon: https://amzn.to...

▶ Play video
languid gulch
#

don't worry, the AMOC will stop soon, and then a mini local ice age will make people forget about things for a bit

glossy glacier
twin dew
#

From the CPU side XD
But basically any computer can struggle with Satisfactory when the save gets way above UObject limit.

wintry galleon
#

https://youtu.be/fadF8oZC-JY?si=4hBkgZHxdZfdHP4M

YouTube
Xu Diace - Xinyu Carton Machinery Co.,Ltd
XY-WJ-1600-80 economic type 2ply corrugated cardboard production li...

🏢Manufacturer : DONGGUANG XINYU CARTON MACHINERY CO.,LTD
🙋🏻‍♂️Sales Director : Mr.Diace Xu
📲Hotline(WhatsApp/WeChat): 008615127793682
008615127798270
📧Email : xy_diacexu@vip.163.com
hbcartonmachinery@gmail.com
#xinyu #carton #flexoprinter #casemaker #corrugatedbox #packaging #slitter #scorer #ambalaj #Emballage...

▶ Play video
twin dew
#

Well done Google... Well done:

image.png
#

Just completely useless as search engine currently.

bronze jasper
median zodiac
inland onyx
vocal thorn
#

Yep that’s a thing

#

Congrats if it works

night girder
#

And AI mode was just the beginning. Google has now announced more changes to the search engine at its own developer conference, I/O, and the message is clear. Google no longer wants you to Google. The company is going to make it increasingly difficult for you to click on blue links and wants you to get answers directly from Google itself. That is fine for Google, but the rest of the internet, and therefore its visitors, will reap the bitter consequences.

#

duckduckgo is the way to go. I guess.

twin dew
#

Just switched to DuckDuckGo mostly.
Just often still go to Google first just from habit.

night girder
#

If you need to google, you can she !g

twin dew
#

DuckDuckGo in turn has their AI mode button just intentionally placed to be misclicked on.

night girder
#

in the search query and duckduckgo goes to google

#

It has?

image.png
#

And you can disable it in settings.

image.png
twin dew
#

Mostly that top right when you have input something already.

image.png image.png
night girder
#

You can even turn off search assit.

#

I don't have that 🤷‍♂️ Firefox + Windows + DDG addon.

#

I can disable it in the settings, but firefox nukes cookies/cache. So settings are always lost.

#

Also here it's clean.

image.png
twin dew
#

Just the only point that was "bad" in current default experience so far there.

night girder
#

Ah. Well default doesn't mean a lot, to me.

#

The thing that matters most is, how much do you let me change the default jace_smile

#

Same with Firefox. It has AI, but atleast a kill switch so ... eh better than most?

#

Duckduckgo has different issue imo.

vocal thorn
tall forum
pure karma
#

and thats with framegen

#

always depends how much you build and how dense you build tho ofc

twin dew
visual tree
tall forum
#

also what about ram is that very important?

#

im going with ddr5-6000 32gb cl28

#

all with a b650M gaming plus wifi mobo

mental oriole
visual tree
#

Same

uneven yacht
#

How well multi threaded is this game? I never really checked and now I don't have a save to check that on

twin dew
#

No idea about Factorios multithreading...

umbral martenBOT
#

Hello there fellow Pioneer!
You're currently chatting in off-topic channels.

Please note, the entire off-topic category is reserved for not Satisfactory related chattery. To comply with your FICSIT Inc. contract, please make use of the appropriate channels in the Satisfactory category. Thank you!

Command executed by: baldurfi
uneven yacht
#

Cause I wanna upgrade my CPU but im not too sure how many of the games I play would benefit from that

#

For now I've basically only got cities skylines 2 and beamng

#

And I'm thinking if this could be the 3rd game

twin dew
#

Satisfactory has two main threads, one for game processing, one for rendering.
And then variable number of worker threads.
But usually you get CPU limited by one of the main threads hitting single-core limits, so no idea how much going past 6 cores helps if any.

Really likes X3D cache.

uneven yacht
#

So basically not multi threaded at all

twin dew
#

Can use tons of theads, but those coordinating threads are the limiter usually.

uneven yacht
#

I'm on 16 cores right now and I'm debating going 32

twin dew
#

Because getting fully independent things is HARD.

uneven yacht
#

Plenty games done did it already

#

I mean I guess it's also a matter of budget

twin dew
#

When the issue is factory calculations?
Where stuff connects to stuff?
And you basically have to go from either start to end, or end to start for each chain.
But those chains interconnect depending on what the player has done.

uneven yacht
#

I mean yeah I guess that only leaves chunk based multi threading which is hard to do and could not even lead to good results depending on the nature and speed of the calculations being done

#

But also no, a city is ALL interconnected and cities skylines 2 can absolutely utilize 64 threads

twin dew
uneven yacht
#

So there are ways of doing it, it's just a question of budget and time allocation to developing it

twin dew
uneven yacht
#

They do actually

#

But a lot of the time physical stuff turns into background numbers before turning back physical, which allows for a transition between threads

#

In satisfactory that could be belt networks being separated by machines

#

Machines only hold items as pure data not a physical game object so it's far simpler dividing jobs between threads using that as the separator

#

But as I said, multi threading is not a very fast thing to develop, so in most cases and I assume also here it's a matter of dev time and budget

#

Well anyways, damn it, still only at 2 games that are gonna benefit from more cores D:

#

I guess Minecraft too, c2me plateus around 80 threads so I'd be fine there too

cyan crescent
#

Who tf needs a 600hz monitor? Why does that exist?

uneven yacht
#

Because it's the obvious next step after 500Hz

cyan crescent
#

I fail to see the reason past like 400hz

uneven yacht
#

Having a really really high refresh rate monitor has no purpose, there's barely any buyers

cyan crescent
#

So why make it then? Seems like a waste

uneven yacht
#

Having the highest refresh rate monitor is what sells

twin dew
#

Because marketing trumps all and that is one of the current Number Wars?

cyan crescent
#

I dont see a need past 240

twin dew
#

Like core count wars (that finally stabilized to around 8 cores) and RAM amount wars on phones.

#

Even if that is 8 very shitty cores, instead of fewer better cores in same die area which would be better in actual use.

cyan crescent
#

You dont need super high ram on phone. Around 12-16GB is enough

twin dew
#

But you just cannot have less cores than 8...

cyan crescent
#

I have 12 cores atm.

uneven yacht
#

That's what finds customers, reviewers, eSports companies, paranoid people, all those are gonna buy what is currently the highest refresh rate monitor

uneven yacht
#

What idiot invented E cores

#

They're missing half of the fuckin instructions

twin dew
cyan crescent
#

Im basically looking at what i can use now and in the future. Currently in the process of swapping out my monitor. Anything past 1440p 240hz i doubt my 7900xtx could handle so i just wont go higher and when i upgrade my gpu, i dont need higher fps than that. I will just crank graphics.

uneven yacht
twin dew
#

Because marketing sees putting out phone with less than octa-core as suicide.
Because of those core count wars that ended at that number.

cyan crescent
cyan crescent
twin dew
# uneven yacht What idiot invented E cores

Just Intel Atom cores.
Same idea as with ARM CPUs having fast cores, and efficient cores.
For Atom, they only lack AVX512 currently compared to the P-cores (classic Core line).

#

Existed for long time as separate low end product line.

uneven yacht
cyan crescent
#

No idea. I just needed a low power cpu with an intel nic on the mobo for my nas when i take it on deployment.

#

Efficiency and low heat matters greatly in my use case

wanton orchid
twin dew
#

And those E-cores for most part work fine for actual independent number crunching code.
They just cause fails in games and like when Windows puts threads on the wrong cores.

#

Allowed Intel to catch up and often bypass AMD CPUs in actual multithreaded work workloads.

uneven yacht
#

But that doesn't change that they're pretty shit

#

AMD did a much better job at them

#

With the c architectures

twin dew
#

Well, those aren't different architecture, just different physical implementation of exact same core.

#

Which would be very hard do to as the first implementation.
They take debugged working core and reimplement it in smaller area by rejiggering where everything is exactly.

#

While for the initial implementation each team has separate areas they do their stuff in.

uneven yacht
#

It is, it's a completely different way of doing the same work as the full size core, there's a reason why Zen5 is a different thing from Zen5c, they quite literally are physically completely different, same work, smaller die area, lower power requirement, lower clock speed

#

So yes it is a different architecture

#

It's not any part of what Zen5 or Zen4 is

#

It had to be designed from scratch instead of tiling what they invented with their regular Zen line

twin dew
#

No, like I said, different physical implementation of same architecture.
To get that smaller die area by redoing where exactly everything is, from the initial implementation where stuff is laid out in clear blocks, with straight edges between them in polygons, to mixed jumble.
But also implemented on lower power node, which limits the max clock speed.

#

And as they are the exact same architecture, the code doesn't see any differences between the implementations.

#

Except for parts where there are intentional changes, like some versions having to do AVX512 in two passes instead of one etc.

#

Which is also done on the Zen5 APUs.

uneven yacht
#

Yeah zen5c reverted it's AVX512 implementation back to Zen4, I hope Zen6c will have single pass AVX512

#

Oh wait Zen6 will have bigger CCDs won't it?

#

Oh I can't wait for the EPYCs

#

I will be finally able to get above 64 cores at high clock speeds

twin dew
#

But point is, at the logical connection level, Zen5 APUs are different from Zen5(c), but Zen5 and Zen5c are basically the same outside of that AVX256/512 unit.
They are the same architecture, just different exact physical die implementations.
Where the tested to work logical level stuff is reimplemented as tight as possible as jumbled "mess".

uneven yacht
#

but that doesnt change the fact that it is a lot different from Zen5, it is a different microarchitecture

twin dew
#

Point is that microarchitecture is high level thing.
And at that level they are identical.
It isn't about the exact physical die implementation, which is different between those two.

#

Intel too has multiple different dies each generation, each implementing same microarchitecture.
With varying amounts of cores per die.
And then even larger changes in the supported extensions etc. in the Xeon dies between low, middle and high core count dies.

#

Not even accounting for the various disabling they do per SKU.

#

And in other direction, Intel P-cores and Intel E-cores on same die are different microarchitectures.
And some ARM CPUs have up to 4 different microarchitectures in same die.

#

Sometimes they use same exact arch for multiple speed levels, sometimes not and just microarchs from same family.

cyan crescent
#

I recently installed bazzite linux onto my Asus Rog Zephyrus G14 Ga402rk and it reported a battery health of 78%. How good or bad is that and at what percentage should i look into getting it replaced?

pure karma
#

Im surprised no one here has talked about unreal engine 6 yet

twin dew
#

Which seems to be just rebranding of UE 5.x whatever would have been next x.

pure karma
#

Probably just hit 6 normally

#

Since last i checked they were at like 5.8

twin dew
#

And UE4 went to 4.28 or more.

#

That second number doesn't stop at 9.

twin dew
#

And I'm continuing to add horrors to that thingie of mine:


    if code == 0:
        return output

    if (code == -1) and (error[0: error.find("\n")] == "Error: Could not process rule: No such file or directory"):
        logger.warning("NFTables failure: Tried to remove IP(s) from a set, that had already been removed by something else.")
        for line in command.split("\n"):
            code, output, error = nft.cmd(line)
            if code == -1 and (error[0: error.find("\n")] == "Error: Could not process rule: No such file or directory"):
                logger.debug(
                    "Command: %s",
                    error[error.find("\n")+1:error.rfind(";")+1],
                )
        return ""```
#

I could just change that delete to destroy for nftables to not error out.

#

Ah, sorry, not visible in this bit.

#

But if something else is messing with the sets, that is something the user should be aware off.

pure karma
#

I cant tell if thats code or an error log so 👍

twin dew
#

Detect specific error, output log entry and rerun all the individual commands.
And print the one(s) that fail again.
As the first set wasn't processed at all by NFT, as when part of atomic transaction fails parsing, nothing is actually changed.

#

Noticed the case, because I have a set that is meant to be updated just before use, and has timeout for the values.
That I was also running as low TTL test case, and started getting errors from NFTables.

#

And I'm stupid...
Why I'm using find("\n") when the string is fixed length to that char when it matches?

mental oriole
#

Made some new cool updates on my message bus 🙂

namespace courier
{
    
    template<>
    void handleObjectMessage(sample::timer::delete_timer& timer, const sample::generic::deltaTime&time)
    {
        timer.remainingTime-= time.dt;
        if(timer.remainingTime < 0)
        {
            std::cout << "Requesting to delete object id[" << (size_t)timer.id <<"]" << std::endl;
            sample::timer::ptr->schedule(timer.id, sample::generic::deleteObject{});
        }
    }
    template<>
    void handleObjectMessage(sample::generic::object& obj, const sample::generic::deleteObject&)
    {
        std::cout << "Handling courier delete object request for id[" << (size_t)obj.id <<"]" << std::endl;
        sample::timer::ptr->remove(obj.id);
    }
}

object ignores all events except the deleteObject event
delete_timer ignores all events except the "deltatime" event. when it's timer reaches 0 it schedules a delete object message at the end of the update frame.
Other objects can listen to other events if they exist. Events are ignored by default template implementation.

verbal raft
pure karma
#

granted iv only done basic JS so who am i to review code

sterile plinth
#

im buying a new laptop (sub700), is there anything important I should be looking out for in terms of specs/features or just the usual cpu and display

mental oriole
# pure karma this is clean code i aprove

The idea is to keep things simple.
Next up is to make a better id to index lookup when sending an event directly to an object.
(Note everything is decoupled atm, so the only thing the object knows it itself and the id it's connected to)

twin dew
#

Not for the exact model, but same chassis and about similar CPU and GPU, or check that the cooler is the same between the models and the CPU and GPU are same or better.
In some cases laptop lineup for same chassis can have multiple coolers and worse models get worse coolers.

#

Hoping this helps for specific thing when that signal handler is entered while logging module is being run...
Need to find way to test for that, possibly by making test thingie that just runs logging in loop, because I forgot to copy the stack trace etc. for the one time I hit the issue.
Where instead of the script quitting, I got long stack trace and it just continued.

        try:
            logger.raiseExceptions = False
            logger.info(
                "%s(%s) received. Exiting.",
                Signals(signum).name,
                signum,
            )
            if nftsets and config:
                flush_set_elements(nftsets, config.nft, config.dry_run)
        finally:
            sys.exit(0)

    signal(SIGINT, handler)
    signal(SIGTERM, handler)```
#

then found in docs that logging module is not reliable inside signal handlers because of locking issues.

twin dew
#

Or might need to move to that 1 second sleep busy checking for a flag.

#

Ok, I was able to replicate the issue reliably...
So now to see how I can reliably fix it.

# 
Traceback (most recent call last):
  File "/usr/lib/python3.13/logging/__init__.py", line 1154, in emit
    stream.write(msg + self.terminator)
    ~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^
RuntimeError: reentrant call inside <_io.BufferedWriter name='<stderr>'>

During handling of the above exception, another exception occurred:
<snip>```
#

Yeah, the exact issue where if that signal handler is called while logging module is doing another thing, you get that issue when signal handler tries to use logging module.

#

So lets see how it can be fixed...

#

If you are implementing asynchronous signal handlers using the signal module, you may not be able to use logging from within such handlers. This is because lock implementations in the threading module are not always re-entrant, and so cannot be invoked from such signal handlers.

twin dew
#

Ok, this works:

        sys.exit(0)

    signal(SIGINT, handler)
    signal(SIGTERM, handler)

<snip>

    logging.raiseExceptions = False

    try:
        main_loop(nftsets, config)
    except SystemExit:
        logger.info("Exiting.")
        flush_set_elements(nftsets, config.nft, config.dry_run)
        sys.exit(0)```
Where that final sys.exit(0) is superfluous.
#

That raiseExceptions = False to get rid of logging module printing those --- Logging error --- entries with its internal exceptions.

#

Whops, that might not be needed, test version had extra stuff still in handler.

#

Yup.

#

And changed that last one to just reraise the same caught exception.

#

And sorry for the spam...

night girder
#

Only thing I wonder, how can time go below 0 😁

#

Oh fuck. You are right. Negative time does exist.

mental oriole
#

Easier to just handle it as negative is just expired mode 🙂

#

The idea of my object structure is to keep object sizes to a minimum, as that allows for more of the same object to fit inside one cache line, aka it's faster to enumerate.

#

And events to all objects -> fast.
Event to singular object -> slower (lookup & cache unfriendly) but less frequent so acceptable.

twin dew
#

And now the backwards compatible to original version should be done too.
And all the important changes backported to my preferred version, which isn't fully compatible with old configs or behavior (flushing sets vs. removing just IPs the program has added).

#

Now I just need to finish couple of systemd unit files, before I can send the thing back to original author.

#

Well, two lines to change in that defaults object to get it config compatible.
And then 23 lines changed/added/removed to make the behavior same for three different things.

quick estuary
#

guys is it normal that my pc uses like 9gb ram without anything open just being on

night girder
#

check task manager, and see if anything is using "a lot" of memory.

quick estuary
#

theres only discord and nvidia container using more than 300mb

#

also could you help me with another problem

twin dew
#

And what does Performance, Memory, say about "In use" and "Available"?

quick estuary
#

like 10gb using and 21.7gb available

night girder
#

sounds about right to me.

#

And I have a game open so.

image.png
#

game takes about 2.8GB.

quick estuary
#

Ye but like i dont get how my friend has 8 gb ram and can have discord open with a game and i have nothing open and it uses 9gb

#

but ye ig its fine

broken shore
#

you have browser opened ?

twin dew
#

Because not used RAM is useless?
And I would expect swap use to be much higher on that 8GB RAM computer.

quick estuary
twin dew
#

I'm at 17.2GB used currently on desktop.
But I have about 20 tabs open etc.

night girder
twin dew
#

Windows.

broken shore
#

check that the process is shut, it can stay like in "cache" and still use some memory though
check your on-start opening apps, some are simply useless sometimes
and baldur is right, your swap* usage must be lower

quick estuary
#

ok ,thx guys

twin dew
#

But do note that all those "300MB RAM" electron apps do add up.

broken shore
#

like wallpaper engine etc they dont seem like much but do stack a lot

twin dew
#

Even Notepad is now using 100 (Private) / 200MB (total working set) RAM.

#

Dropbox app using 330/515MB.

#

Where that second number is with memory that is shared with other apps, from using same libraries etc.
So only counts once for all of them.

glossy glacier
#

Had to check a "defect" PC that either didn't boot or randomly turned off today

#

Main suspect was the PSU with the off chance of mainboard

#

It was neither

#

It was the ||power button, randomly closing the circuit very shortly. Enough to trigger an ahci shutdown but not long enough to power it on||

pure karma
ashen patio
mental oriole
pure karma
#

Yea windows run fine on like 4GB’s of ram but “needs 10GB+” cause its greedy elevatedone

visual tree
#

Just why 😭

image0.jpg
#

Ugliest Ferrari in history

#

Apparently, iPhone designer was behind interior design

mental oriole
pure karma
#

generational blunder on my part onyl getting a 12GB GPU tbh

#

atleast its a easy upgrade but not a cheap one and definitely not one for anytime soon

oblique hamlet
#

have you checked if linux uses less resources for you?

jagged snow
pure karma
#

but yea for ram generally

languid gulch
#

for some reason whenever someone writes out 12GB, i read it as 128GB

twin dew
#

He meant buying new GPU with more VRAM.
Not adding more to same GPU.